| Dokumendiregister | Riigimetsa Majandamise Keskus |
| Viit | 3-6.1/77 |
| Registreeritud | 05.01.2024 |
| Sünkroonitud | 31.12.2025 |
| Liik | Kiri |
| Funktsioon | 3-6.1 |
| Sari | Looduskaitse ja jahinduse alane kirjavahetus |
| Toimik | |
| Juurdepääsupiirang | Avalik |
| Juurdepääsupiirang | |
| Adressaat | Tartu Ülikool |
| Saabumis/saatmisviis | Tartu Ülikool |
| Vastutaja | Kaupo Kohv |
| Originaal | Ava uues aknas |
JÄÄKSOODE VEEREŽIIMI TAASTAMISE KOMPLEKSUURINGU LÕPPARUANNE 1. PROJEKTI KESTUS Algus
(kuu/aasta): 24.04.2017 Lõpp:
(kuu/aasta) 01.09.2023
2. PROJEKTI TAOTLEJA (teadusasutus): Tartu Ülikool
Telefon: +372 7 375826
Aadress: Ülikooli 18, 50090 Tartu
Registrikood: 74001073
Panga rekvisiidid: SEB Pank AS, Tornimäe 2, 15010 TALLINN, arvelduskonto (IBAN): EE281010102000234007 , SWIFT/BIC: EEUHEE2X , käibemaksukohustuslase nr (VAT number): EE100030417 , tehingupartneri kood (TP kood): 605201 3. PROJEKTI JUHT: Ain Kull
(Ees- ja perekonnanimi) kaasprofessor, PhD (Amet, teaduskraad)
4. PROJEKTI PÕHITÄITJAD ARUANDEPERIOODI VÄLTEL Projekti põhitäitjad:
Ees- ja perekonnanimi Teaduskraad Ametikoht 1. Ain Kull PhD loodusgeograafia kaasprofessor 2. Valentina Sagris PhD geoinformaatika teadur 3. Edgar Karofeld PhD rakendusökoloogia kaasprofessor 4. Kai Vellak PhD taimeökoloogia kaasprofessor 5. Alar Läänelaid PhD maastikuökoloogia emeriitdotsent 6. Gert Veber PhD loodusgeograafia teadur 7. Marko Kohv PhD rakendusgeoloogia teadur 8. Mae Uri Dipl./BSc spetsialist (keemik) 9. Edgar Sepp MSc, doktorant geoinformaatika spetsialist 10. Martin Maddison PhD keskkonnatehnoloogia kaasprofessor 11. Ivika Ostonen-Märtin PhD juureökoloogia professor 12. Kristina Sohar PhD loodusgeograafia teadur 13. Iuliia Burdun PhD doktorant, kaitsnud PhD 2020 14. Tauri Tampuu PhD doktorant, kaitsnud PhD 2022 Projektiga seotud abitööjõud: 1. Kärt Erikson BSc/MSc magistrant (2023) / doktorant (2023 sept.) 2. Birgit Viru MSc/PhD doktorant, kaitsnud PhD 2020 5. PROJEKTI KULUD ARUANDEPERIOODIL 2023.a. 69267,18 eurot
Kokku
Töötasud (põhitäitjad +abitööjõud) 40702.70 Sotsiaalmaks 13408.58 Töötuskindlustusmaks 325.08 Ostetud teenused 4879.82 Lähetuskulud 4750.46 Materjalid, tarvikud, masinad, seadmed 5087.24 Muud kulud 113.30 Kokku 69267.18
Ostetud teenuste selgitus 4879.82 Mulla- ja veekeemia analüüsid biogeokeemia
laboris Lähetuskulude selgitus 4750.46 Kõik lähetused on seotud uurimisaladel
gaasi- ning veeproovide regulaarse kogumisega, drooniseire ja taimkatteseirega
Materjalide, tarvikute, masinate ja seadmete selgitus
5087.24 Fotosünteetiliselt aktiivse kiirguse mõõtmise andurid ja temperatuuriandurid. Soetati mõõteseadmetele patareisid ja akusid, seadmete hooldusmaterjale, mõõdulinte, teipe jmt. tarvikuid
Muude kulude selgitus 113.3 Kummikud, töökindad välitöödeks. Kulurida ei kajasta Tartu Ülikooli üldkulueraldist (20%) RMK-lt 2023.a. esitatud aruannete eest (arvestuslik summa 16328.22), mis kajastub eelarves pärast aruande heakskiitmist ja lepingutasu laekumist tartu Ülikoolile.
6. PROJEKTI TÄITMISE LÕPPARUANNE Rakendusuuringu „Ammendatud turbamaardlate vee-režiimi taastamise kompleksuuringu metoodika väljatöötamine ja uuringu läbiviimine“ eesmärgiks oli perioodil 2017 – 2023 luua jääksoode seisundi ja korrastamisjärgsete muutuste seiramise metoodika, rajada valimisse kuuluvas viies jääksoos seirealad ning viia läbi kogu perioodi hõlmav kompleksseire. Lõpparuandes antakse ülevaade projekti raames 2017.a. aprillist kuni 2023.a. septembrini Laiuse, Kõima, Maima, Kildemaa ja Ess-soo jääksoodes läbi viidud tegevustest ja esmastest tulemustest ning tuuakse välja peamised seire käigus tehtud tähelepanekud korrastamistööde edukust mõjutavatest teguritest. Koondaruandes käsitletud teemade detailsem analüüs (eeskätt metoodika osas) on esitatud aruande lõpus viidatud lisana esitatud artiklites ja lisamaterjalides. Seirealade rajamine, seire ja korrastamistööde ajajoon Eelneva ruumianalüüsi ning välitööde tulemuste põhjal valiti RMK poolt korrastatavate jääksoode hulgast seiratavateks aladeks Laiuse, Ess-soo, Maima, Kõima ja Kildemaa jääksood. Neist Maima ja Ess-soo moodustavad väga sarnase paari, kus on esindatud mahajäetud freesturbaväljad ja iseseisvalt taimestunud turbavõtuaugud ning nende vahel kuivemad metsastuvad tervikud ning Ess-soos ka freesturbavälja laiendamiseks eelkuivendusega kuid algse rabataimestiku eemaldamiseta ala. Selle paari puhul oli eesmärgiks korrastamistööde käigus võrrelda Lääne-Eesti ja Kagu-Eesti erinevusi (ilmastik, pealiskorraga seotud mõjutused turba- ning veeomadustele) järgnevate töötlustega aladel:
a) võrdlusalad (korrastamistööde käigus mõjutamata veerežiim ja taimestik); b) madalaveeline veekogu loodusliku taimestumisega; c) alad pinnaspaisudega stabiliseeritava veerežiimiga, kus taimestik areneb iseseisvalt; d) alad lausalise kraavide täitmisega stabiliseeritava veerežiimiga, kus taimestik areneb iseseisvalt; e) alad pinnaspaisudega stabiliseeritava veerežiimiga, kus turbasambla fragmentide siirdamisega
kiirendatakse taimestumist; f) alad lausalise kraavide täitmisega stabiliseeritava veerežiimiga, kus turbasambla fragmentide
siirdamisega kiirendatakse taimestumist. Laiuse jääksoo puhul oli kobraste tegevuse tulemusena lõunapoolses jääksoo osas kujunenud madalaveeline veekogu, põhjapoolses osas aga mahajäetud freesturbaväljal suhteliselt suure ida- läänesuunalise kõrgusgradiendiga vähe kuni mõõdukalt taimestunud ala. Korrastamistööde käigus säilitati loodusliku taimestumise teel soostuv veekogu, põhjapoolne freesturbaala jagati aga neljaks erineva veetasemega osaks, kus idapoolses osas on maapinna suhtes kõige sügavam veetase ning läänepoolses osas veetase maapinnale lähedane ning keskel võrdlusala. Kõikjal peale võrdlusala eemaldati puurinne ning veetase stabiliseeriti kraavidele rajatud pinnaspaisudega. Läänepoolsel ala rajati veetaseme tõstmise tõttu laienenud veepeegliga kraavidega alal ka raiejääkidest lainerahusti. Taimestiku puhul eeldati looduslikku arengut. Kildemaa jääksoo hõlmas nii mahajäetud freesturbavälja kui rabapoolses osas ka freesturbavälja laiendamiseks eelkuivendusega kuid algse rabataimestiku eemaldamiseta ala, mis on lausaliselt
puurindega kaetud (sarnane Ess-soo vastavale alale). Korrastamismeetmena oli kavandatud kraavide sulgemine pinnaspaisudega, tihedama puistu raadamine ning taimestiku iseseisev areng stabiliseeritud veerežiimi tingimustes. Kõima jääksoos oli korrastamisalal nii turbavõtuaukudega ala (sarnane Maimaja Ess-soo vastava tüübiga) kui ka eelkuivendusega ala (sarnane Kildemaa ja Ess-soo vastavale tüübile kuid oluliselt kõrgema veetasemega ning vähem metsastunud). Kõima uurimisala oli lausaliselt rabaliikidega taimestunud ja vajas korrastamistööde käigus vaid pinnaspaisudega kraavide sulgemist veetaseme taastamiseks ning tervikutel ja kraavide servades puurinde eemaldamist/harvendamist. 2017.a. suvel viidi seiratavates jääksoodes läbi turbalasundi sondeerimine, mille käigus hinnati turbalasundi tüseduse ruumilist varieeruvust ning määrati organoleptiliselt turba tüüp ning lagunemisaste (joonis 1). Jääklasundi omaduste ning taimestumise iseloomu järgi valiti igas jääksoos (v.a. Laiuse) kaks algseisu kõige paremini esindavat piirkonda võrdlusaladeks, kus kogu uuringu raames muudatusi ei tehta ja mille suhtes võrreldakse korrastatavate alade muutusi.
Joonis 1. Üldvaade võrdlusaladele Kõima (vasakul) ja Maima (paremal) jääksoos ning vastavalt kumbagi jääksoo võrdlusalade A ja B turbaprofiilidele. 2017.a. augustis rajati kõigil uuringualadel võrdlusalad (alad mis jäävad muutumatuks ka korrastamistööde käigus ehk referentsalad). Neile aladele installeeriti turbaveevaatluskaevud (Ess-soo 3 tk, Laiuse 2 tk, Maima 2 tk, Kõima 2 tk, Kildemaa 2 tk), veetaseme mõõtekaevud ning gaasivoogude mõõtmise püsivahendid. Samal kuul alustati igakuiste vee- ja gaasiproovide kogumist. Igakuiselt (Laiuse ja Ess-soo puhul ka kaks korda kuus) koguti võrdlusaladelt gaasiproovid (CO2, N2O, CH4), mõõdeti vaatluskaevudes ning kraavides veetase, portatiivsete seadmetega O2 sisaldus (mg/l) ning küllastatustase (O2%), pH, konduktiivsus (µS/cm), ORP (mV) ja koguti veeproovid laboratoorseteks analüüsideks. Laboratoorselt määrati igakuiselt vaatluskaevudest ning võrdlusaladega piirnevatest kraavidest kogutud veeproovidest üldsüsiniku ja üldlämmastiku, lahustunud üldsüsiniku, lahustunud orgaanilise süsiniku, lahustunud anorgaanilise süsiniku ning lahustunud üldlämmastiku sisaldus. 2018.a. suvel alustati taimkatte maapealse seire välitöid kõigis viies jääksoos (Kõima, Maima, Laiuse, Kildema ja Ess-soo), kus võrdlusaladel ja erineva planeeritava korrastamismeetodiga aladel märgistati 1x1 m püsiruudud (kokku 156), võeti nende nurgapostide koordinaadid (RTK, kasutati ka drooniseirel ankurpunktidena). Taimkatteseire püsiruudud fotografeeriti ja neil teostati taimkatte analüüs (üldkatvus, eri rinnete ja taimeliikide esinemine ja katvus). 2017-2020. a. osaleti jääksoode korrastamisprojektide koostamisel ja anti sisend projekteerijatele. 2019.a. suvel-sügisel korrastati Laiuse ja Kõima jääksood. 2020.a. rajati korrastatud Laiuse ja Kõima jääksoodes täiendavad püsiproovialad, installeeriti korrastatud aladele täiendavad piesomeetrid ning vaatluskaevud ja rajati täiendavad gaasivoogude mõõtmise alad (sh. täiendavad ujuvkambrid kraavidele ning veekogule). Neil aladel alustati igakuist seiret ning proovialadele rajatud ülevooludel veeseiret. 2020.a suvel-sügisel korrastati Maima jääksoo, oktoobris installeeriti korrastatud aladele täiendavad piesomeetrid ning vaatluskaevud ja rajati täiendavad gaasivoogude mõõtmise alad. 2021.a. suvel parandati Laiuse jääksoos eraldusvalli kõrge veetasemega ala ja kontrollala vahel ning korrigeeriti kahel alal (reguleerimatute) ülevoolude kõrgust. 2021.a. sügisel-talvel korrastati Ess-soo jääksoo ning alates 2021.a. novembrist alustati
korrastamisjärgset seiret värskelt rajatud ülevooludest. 2022.a. kevadel installeeriti Ess-soos korrastatud aladele täiendavad piesomeetrid, vaatluskaevud ja rajati täiendavad gaasivoogude mõõtmise alad ning alustati korralist igakuist seiret. 2023.a. augustis alustati Kildemaa jääksoo korrastamist. Uuringuperioodi ilmastiku ülevaade Ilmastik mõjutab oluliselt soode veerežiimi aastatevahelist muutlikkust ning seeläbi nii veega toitainete ärakannet, ökosüsteemi gaasivahetust kui ka taimestiku arengut. Looduslikus seisundis sood on suhteliselt suure puhverdamisvõimega, jääksood aga vähese puhverdamisvõimega ning eriti tundlikud on ilmastiku suhtes värskelt korrastatud alad. Kuna aastate lõikes on ilmastik olnud väga erinev, tuleb seda silmas pidada ka aastatevahelisi veetaseme, gaasivoo ning ärakande väärtusi võrreldes ning korrastamistööde üldise edukuse hindamisel. Kui korrastamiseelsel perioodil oli väga põuane vaid 2018 aasta (mis järgnes keskmisest vihmasemale 2017 lõpule), siis korrastamisjärgne periood oli oluliselt kuivem nii 2021, 2022 kui ka 2023 juulini. Uurimisperioodi iseloomustab pikaajalisest keskmisest kõrgem õhutemperatuur (joonis 2). Eriti soojad olid 2019 ja 2020 aastad, mil talvekuudel oli kuu keskmine temperatuur normist isegi kuni 6 kraadi soojem. Soojemad talvekuud tähendasid sagedasi sulaperioode ja kevadel väiksemat lumeveevaru, mistõttu soo veetase sõltus juba varakevadest alates peamiselt sademete hulgast ning päikesekiirguse intensiivsusest.
Joonis 2. Kuu keskmise õhutemperatuuri (joonisel heleroheliste tulpadena) erinevus ( ̊C) uurimisperioodil võrreldes pikaajalise keskmise kliimanormiga (joongraafik). Samblafragmentidega jääksoode korrastamine toimus 2020.a. (Maima) ja 2021.a. (Ess-soo), seetõttu on eriti oluline tähelepanu pöörata perioodi 2021-2023 ilmastikule. Talv algas suhteliselt varakult 2021.a. keskmisest külmema detsembriga kuid maapind külmus varase lumikatte (novembris) tõttu vaid osaliselt ja keskmisest soojem talve jätk (joonis 2) soodustas nii lume sulamist kui külmumata pinnasest gaasivoo eritumist. Korrastamisalade seisukohast oli aga kõige olulisem sulailmadega lumeveevaru kahanemine ja (pool)külmunud pinnasel tekkinud lombid, mis tuule tekitatud lainetusega uhtusid samblafargmentide katteks laotatud põhu vaaludesse. Talvistele keskmiselähedastele sajuhulkadele järgnes aga 2022.a. erakordselt kuiv märts (joonis 3) ning kogu järgneva aasta jooksul oli iga kuu sademete hulk ligi 40% väiksem pikaajalisest kuu sademete normist. Vaid kahel järgneval talvekuul oli sademeid keskmisest enam, kuid 2023.a. kevad algas taas väga tugeva põuaga kui sademeid langes kuude lõikes vaid 30-40 pikaajalisest normist.
Joonis 3. Kuu keskmise sademete summa (joonisel tumesiniste tulpadena) erinevus (%)uuringuperioodil võrreldes pikaajalise kuu keskmise sademete summaga (joongraafik; kuu sademete summa millimeetrites). Keskmisest kõrgem õhutemperatuur, erakordselt väike sademete hulk (132 mm normist vähem 2022.a.) ja keskmisest päikeselisem ilm (eriti 2023 aprillist juunini; joonis 4) tingis intensiivse evapotranspiratsiooni tõttu maist alates kiire veetaseme alanemise (auramine ületas sademete hulka juba märtsist) ja kuivastressi 2021.a. samblafragmentide laotamisega korrastatud uurimisaladel Ess-soo jääksoos, aga ka 2020.a. sarnaselt korrastatud Maima uurimisaladel.
Joonis 4. Kuu keskmine päikesepaiste kestuse summa (joonisel kollaste tulpadena) erinevus (%) uuringuperioodil võrreldes pikaajalise kuu keskmise päikesepaistega tundide summaga (joongraafik; kuu päikesepaistega tundide summa). Kuivale 2022. aastale järgnenud väga kuiv, päikeseline ja suhteliselt tuuline 2023.a. kevad tõi kaasa väga kiire veetaseme alanemise ning samblafragmentide kasvu pidurdumise või isegi kohati hukkumise. Maima jääksoos Ala 9 (kood P) kannatas 2020.a. suhteliselt hästi kasvama läinud turbasammal tugeva talvise külmakohrutuse all. Külmakohrutusest on Maimas iga talv olnud tugevalt mõjutatud ka võrdlusala 1 (Ala 6-2) ja Ala 8 (kood O). Ess-soos on külmakohrutuse mõju olnud väiksem, kõige enam on seal mõjutatud olnud ala 4 (kood F).
TULEMUSED Pinna- ja turbavesi Nii turbavees kui jääksoo kraavides on süsinik ja lämmastik valdavalt lahustunud vormis (vastavalt 92% ja 92% üldsüsinikust ja üldlämmastikust), lahustunud ja lahustumata vormid aga omavahel tugevalt korreleeritud. Lämmastiku ja süsinikusisaldus oli mõõtmisperioodil kõrgem turbavees, kraavides oli pindmise äravoolu ja sademetevee tõttu toimunud vähesel määral lahjendumine. Küll aga on nii kraavi- kui turbavees lahustunud süsiniku (DC) ja lahustunud lämmastiku (DN) suhe väga heaks turba mineraliseerumise indikaatoriks: DC/DN suhe on seda kõrgem ja regressioonvõrrandi seos tugevam, mida enam on ala häiritud (joonis 5).
Maima
y = 11.953Ln(x) + 27.244 R2 = 0.6754
10
20
30
40
50
60
70
0.0 1.0 2.0 3.0 4.0 5.0
Dissolved N (mg/l)
D is
so lv
ed C
(m g/
l)
Kildemaa
y = 19.46Ln(x) + 39.445 R2 = 0.7029
0
10
20
30
40
50
60
70
0.0 1.0 2.0 3.0 4.0 5.0 Dissolved N (mg/l)
D is
so lv
ed C
(m g/
l)
Joonis 5. Lahustunud süsiniku ja lahustunud üldlämmastiku vaheline seos Kildemaa ja Maima jääksoo näitel. Veekvaliteedi ruumilise autokorrelatsiooni hindamiseks koguti veeproovid nii tootmisväljakutevahelistest kraavidest, piirdekraavidest kui väljavooludest 2018 ja 2019 mais ning 2019 septembris. Veeproovide tulemused näitavad tugevat autokorrelatsiooni lahustunud süsiniku kontsentratsiooni (joonis 6) ja mõõdukat korrelatsiooni lahustunud üldlämmastiku (joonis 7) osas. Lahustunud süsinik on kõigi seirealade puhul orgaanilise süsinikuna. Vaid Maima jääksoos esines kevadel ning sügisel pikema vihmaperioodi järel anorgaanilist süsinikku (karbonaadina) Ala 5 (kood B) ja 6-1 (võrdlusala 2) seirekaevudes. Korrastamisjärgselt on 2021-2023 aastatel Ala 5 seirekaevus põhjaveeline toide suurenenud ning anorgaanilise süsiniku (DIC) ja lahustunud lämmastiku (DN) osakaal suurenenud. Eeldatavasti on see seotud pinnaspaisude jaoks liiga sügavalt ekskavaatoriga materjali võtmisel veepidemeks olnud hästilagunenud turbakihi rikkumisest.
Joonis 6. Lahustunud süsiniku (DC) kontsentratsioon (mg/l C) jääksoo kraavi- ja turbavees Ess-soo ning Kildemaa uurimisalade näitel.
Joonis 7. Lahustunud lämmastiku (DN) kontsentratsioon jääksoo kraavivees. Lahustunud lämmastiku sisaldus on väga madal (joonis 7) nii kraavides kui turbavees. Kui DC kontsentratsioon sõltus taimestumisest mõõdukalt ja olulisem oli kraavituse seisukord, siis toitainevaeses keskkonnas on lahustunud lämmastiku kontsentratsioon selgelt madalam taimestunud kraavides ning eriti madal piirkondades kus maapind on lausaliselt kaetud taimestikuga. Korrastamistööde mõju toitainete ja lahustunud süsiniku ärakandele on lühiajaline kui tööde käigus ei mõjutata põhjaveetoitelisust. Maima jääksoos suurenes lahustunud lämmastiku kontsentratsioon (joonis 8) kõige enam alal 5 (Maima B) ja 10 (Maima D). P ja E alal võib osaliselt kõrgemat DN fooni korrastamisjärgselt selgitada ka samblafragmentide laotamise järel laguneva orgaanilise ainega (kattepõhk ning surnud fragmendid).
Joonis 8. Lahustunud lämmastiku kontsentratsiooni muutus uuringuperioodi jooksul Maima jääksoos. Kuigi ka Laiuse jääksoo on toitainerikkam ja osaliselt põhjaveelise toitega, on seal korrastamisjärgsel ajal (pärast 2019 sügist) lahustunud lämmastiku sisaldus kerges langustrendis (joonis 9) ja selgitatav kiiresti areneva taimkattega ja vähese äravooluga. Kui vesi on pikema viibeajaga, siis tarbitakse lämmastikku nii taimestiku poolt kui sõltuvalt redokspotentsiaalist denitrifikatsiooni/nitrifikatsiooni protsessides. Võrdlusalal kus muutused on olnud väikesed, on ka DN püsinud stabiilsena.
Joonis 9. Lahustunud lämmastiku kontsentratsiooni muutus uuringuperioodi jooksul Maima jääksoos. Ess-soo ja Kildemaa uurimisaladel on DN sisaldus vees madal ja aastatevahelised erinevused statistiliselt ebaolulised. Küll aga on selgelt tuvastatav aastaajaline käik kõrgema kontsentratsiooniga suvekuudel, ning kuivematel aastatel (2018, 2022, 2023).
Joonis 10. Lahustunud lämmastiku kontsentratsiooni muutus uuringuperioodi jooksul Ess-soo ja Kildemaa jääksoos. Lahustunud orgaanilise süsiniku sisaldus on Maima jääksoos võrdlusalal 1 jäänud muutumatuks, võrdlusalal Maima 2 langes korrastamistööde käigus põhjaveelise toite lisandumisel väga madalale (joonis 11), aga teistel aladel on sarnane võrdlusalaga 1.
Joonis 10. Lahustunud orgaanilise süsiniku kontsentratsiooni muutus uuringuperioodi jooksul Maima jääksoos. Sarnase selge aastase käiguga kuid olulise trendita on DOC sisalduse käik ka teiste uurimisalade puhul (Laiuse, Ess-soo, Kildemaa, Kõima). Ess-soost võetud puursüdamiku 8-kuune laboratoorne inkubatsioonikatse näitab, et temperatuuri seos DOC-ga pole erinevalt CO2 voost lineaarne, aga 25 C ületav temperatuur suurendab oluliselt DOC teket turbas, olles samas sõltuvuses veetasemest/aereeritusest (Palviainen et al., 2023). Laiuse jääksoos tuleb esile korrastamistööde järgselt madalaveelise veekogu DOC sisalduse langus ja stabiliseerumine 60 mg/li tasandi lähedal.
Joonis 11. Lahustunud orgaanilise süsiniku kontsentratsiooni muutus uuringuperioodi jooksul Laiuse jääksoos.
Vaatamata suhteliselt kõrgemale kontsentratsioonile nii DN kui DOC osas, ei ole ärakanne korrastatud jääksoost suur kuna vee äravool korrastatud aladel on viimastel aastatel (ülevoolu rajamisest saati) olnud vaid lühikesel perioodil talviste sulade ajal ning kevadel lume sulamise järel, mil kontsentratsioonid on keskmisest madalamad. Laiuse jääksoo ülevoolude puhul on äravool vaid märtsis-aprillis, läänepoolses ülevoolus (madalaveelise veekogu ja Lehtmetsa raba vesi) kuni 4 kuud (märtsist juunini). Sarnane on äravoolu periood ka Maima ning Ess-soo puhul. Kõima edelapoolse kraavi äravoolu pole võimalik hinnata kuna vesi valgub ühtlaselt metsa alla. Kirdepoolses äravoolus liigub vesi novembrist maini. Täpse äravoolu koguse hindamine on takistatud kuna Ess-soos viis 30. augusti sadu ülevoolu kõrvalt pinnase ja mitmel sügiskuul puudus äravoolu mõõtmine, Laiuse läänepoolsel ülevoolul muutis kobras V-ülevoolu kuju ja suurust ning Maimal on suure veetaseme kõikumise tõttu olnud vaja vähemalt kaks korda aastas ülevoolu kõrgust reguleerida. Vooluhulga ja kontsentratsiooni järgi hinnates on süsiniku ärakanne DOC kujul jääksoodest vahemikus 62-87 kg/ha*aastas. Mullastik Korrastamistöödega seotud muutused mulla keemilistes omadustes on väga väikesed ja üldjuhul statistiliselt ebaolulised (joonis 12). Ainus oluline muutus on seotud Maima jääksoo mulla happesusega, kus ilmselt on põhjuseks vettpidava turbakihi häirimine ja selle tulemusena suurem põhjavee sissevool alale (eriti Ala 5 (B), aga ka 2 (L), 10 (D) ning 11 (E). Teiste parameetrite osas olulisi muutusi ei toimunud, aga pinnasetööde tõttu suurenes ruumiline varieeruvus. Samblafragmentide laotamisega alal tõusis pindmises kihis süsinikusisaldus keskmiselt ligi 1% võrra, kuid pole selge kas seda tingis täiendav orgaanilise aine lisandumine (sammal, põhk) või eelnevalt osaliselt mineraliseerunud pinnase koorimine.
Joonis 12. Mulla pH, üldfosfori ja üldlämmastiku sisalduse muutus korrastamistööde käigus. Maima jääksoos mulla pH muutuse ja põhjaveelise toitumuse suurenemise vahelist seost kinnitab ka lahustunud anorgaanilise lämmastiku (DN) sisalduse suurenemine poorivees ning kraavides (joonis 8). DN sisaldus on suurenenud samadel aladel (B, D, E) kus tõusis mulla pH sisaldus ning poorivee karbonaatiooni sisaldus, aga muutus ei avaldu võrdlusalal ega selle juures kraavi vees. Jääksoode mullaanalüüsi andmeid kasutati üleriigilise suuremõõtkavalise mulla fosforisisalduse kaardi koostamisel. Valminud kaart on GIS andmestikuna vabavaraliseks kasutamiseks ja metoodika osas detailsemalt kirjeldatud artiklis: Kull, Anne; Kikas, Tambet; Penu, Priit; Kull, Ain (2023). Modeling Topsoil Phosphorus—From Observation-Based Statistical Approach to Land-Use and Soil-Based High- Resolution Mapping. Agronomy, 13 (5), 1183. DOI: 10.3390/agronomy13051183. Biomassi lagunemiskatsed Jääksoodes viidi läbi standardiseeritud teekotikatse rohelise ning punase (rooibos) teega ning Laiuse ja Ess-soo jääksoodes korrastamisjärgselt maa-aluse ja maapealse biomassi lagunemiskatsed. Standardiseeritud teekottide (punane e. rooibos ja rohelise tee) katse esmased tulemused Laiuse jääksoos alustatud eksperimendist lubavad oodata selget seost nii veetasemega kui taimestikuga (joonis 13). Esimese aasta massikadu on Maima ja Kildemaa jääksoos punase tee puhul sarnane Laiuse jääksoos teekottide massikaoga, kuid erinevus rohelise ja punase tee vahel on väiksem. Kildemaa jääksoos on lagunemine mõnevõrra kiirem kui Maima uurimisaladel, eriti rohelise tee osas.
Joonis 13. Vasakul teekottide paigutuse skeem katsealadel, parempoolsel joonisel punase ja rohelise tee jääkmass 3 kuu, 6 kuu, 1 aasta, 1,5 aasta ja 2 aasta pärast Laiuse võrdlusalal (control), rabametsas (Raba), kuivenduse mõjuga rabametsa servas (Kuivendatud mets) ja pinnaspaisudega korrastatud keskmise veetasemega uurimisalal (Keskmine veetase; korrastamisprojektis Ala 2, uurimisala kood Laiuse E) ning alumisel joonisel jääkmass esimese aasta lõpuks Maima ning Kildemaa võrdlusaladel. Rohelise tee lämmastikusisaldus on kõrgem (3-5%) ja imiteerib peenjuurte lagunemist ning on happelises pinnases suhteliselt suure hajuvusega. Punane tee imiteerib rohkem okaste varist ning selle lagunemine on erineva taimestiku ning veerežiimiga aladel ühtlasem. See viitab ka voortevahelises Laiuse jääksoos (turba pH 2.5-3.5, mediaan 3.1) lagundavate mikroorganismide ühtlast aktiivsust erinevates kooslustes ja rohelise teega võrreldes suhteliselt madalamat leostumiskadu, eriti esimese 6 kuu jooksul. Tulemuste põhjalikum analüüs koos kõigi keskkonnategurite (temperatuur, veetase, mullakeemia, sademed jmt) toimub koostöös Iiri ning Rootsi teadlastega ja võrdluses nende sarnaste katsete andmetega. Sarnaselt 2021.a. varakevadel Laiuse jääksoos alustatud maa-aluse ja maapealse biomassi lagunemiskatsetega laiendati 2022.a. katset värskelt korrastatud Ess-soo alale. Lagunemiskatsesse lisati standardiseeritud teekotikatsele ka eraldi proovid männi ja sookase ning villpea, jõhvika ja mustika/sinika peenjuurte ning varisega. Lagunemiskatsed (vahetult maapinnal ning 5-10 cm sügavusel turbas) rajati kuivemal ja märjemal võrdlusalal, turbasambla fragmentide laotamisega alal, pinnaspaisudega tõstetud veetasemega alal ning suletud kuivenduskraavidega rabametsa alal (joonis 14). Katse on korduste arvu järgi planeeritud kolmeaastasena.
Joonis 14. Lagunemiskatse rajamine Ess-soos uurimisalal 2022. aastal. Vasakpoolsel joonisel proovide paigaldamine alale nr. 11 (kood C) pinnaspaisudega suletud kraavidega alal ning parempoolsel joonisel proovide paigaldamine suletud kuivenduskraavidega rabametsa alal. Kaugseire Arvestades jääksoode suurt pindala, raskesti ligipääsetavust, alasisest suurt heterogeensust ning korrastamistööde puhul ka võimalikku kiiret taimkatte arengu dünaamikat, on kaugseire potentsiaalselt hea vahend seisundi hindamiseks. Käesoleva uuringu raames hinnati nii optilise seire (droon ja satelliit) kui radarkaugseire (satelliit) rakendamise võimalusi. Drooniseire peamiseks eeliseks on väga hea lahutusvõime ja võimalus lennata vastavalt vajadusele ning ilmastikuoludele. RGB kaameraga droonid on praeguseks kujunenud laiatarbekaubaks ja pildi kvaliteet on väga hea. Peamised RGB kaameraga droonide kasutamisega seotud metoodilised küsimused puudutavad erinevate aastate lõikes homogeensete aegridade saavutamist, sest vaatamata päikesekiirgusandurite ja kalibreeritud peegeldusplaatide kasutamisele on drooniseireks liiga suurte (eriti Ess-soo ja Maima) alade puhul probleemiks suur kiirgusspektri ajaline varieeruvus. Lennuaja jooksul muutuvad valgusolud ja kiirgusspekter kahandab piltide põhjal automatiseeritud taimkatteklassifitseerimise edukust erinevate ülelendude vahel, aga ka isegi sama päeva lendude osas kui kiirgusintensiivsus jõuab pika lennuaja jooksul oluliselt muutuda. Paremate sensoritega (kiirgusspektri andurid nii üles kui allasuunatuna) droonid, kalibreeritud peegeldusplaadid, georefereeritud ankurpunktid jmt. muudab aga lendamise kalliks ja töömahukaks (sh. kameraalne järeltöötlus). Maima ning Kõima jääksoo korrastamise eelse drooniandmestiku põhjal hinnati erinevate masinõppe algoritmide rakendatavust ja nende maakatte klassifitseerimise täpsust. MarjanSadat Barekaty leidis oma magistritöös Maima jääksoo põhjal, et nii Random Forest (RF), Support Vector Machine (SVM) ja K-Nearest Neighbours (KNN) meetod annavad suhteliselt sarnase tulemuse RGB kaameraga drooniandmestiku puhul. Kõrgeim kaalutud keskmine F1-skoor saadi RF vaikemudeliga kombineerituna vegetatsiooniindeksitega (0,59), sellele järgnesid KNN (0,58) ja SVM (0,57) kombineerituna vegetatsiooniindeksite ja MinMaxScaleriga. Pildi suurem pikslitihedus ei parandanud klassifitseerimise tulemust. Klassifitseerimist raskendas oluliselt UAV ortofoto kõrgest ruumilisest lahutusest tingitud müra ja maakatteklasside mitte tasakaalus olev koosseis (erinevate liikide/koosluste ruumiline esinemine ebavõrdne, mis on aga looduses tavapärane olukord). Teistele uuringutele tuginedes saaks ilmselt klassifitseerimistulemusi parandada kasutades objektipõhist pildianalüüsi (OBIA), mis töötaks paremini puurinde ning mättaid moodustavate taimede puhul ning lisades kalibreeritud multispektraalsed andmed ning lisatunnused (nt. LIDAR andmed). Sarnaselt droonipiltide töötlemise ja sellelt taimkatte tuvastamise metoodikale on võimalik automatiseeritult tuvastada ka taimestumise osakaalu taimkatteruutude fotode alusel. Uuringu käigus arendati QGIS tarkvara baasil fototuvastussüsteemi, et kõrge lahutusega fotodelt (joonis 15 a ja b) RGB kanalites automatiseeritult eristada kasvama läinud turbasambla fragmentide pindalalist katvust. Selleks fotografeeriti standardselt kõrguselt 1x2 m raami jäävad ruudud (100 tk), neist 33 kasutati õpetusalana ja 67 ala automaattuvastuse alana ning neist omakorda 33 lisaks käsitsi klassifitseeritavate kontrollaladena (joonis 15 c).
Joonis 15. Kasvama läinud turbasambla fragmentide tuvastamiseks kasutatud fotod (a ja b), mis georefereeriti ja transformeeriti ortofotodeks. Käsitsi klassifitseeritud alad kasvavate turbasamblafragmentidega (15c) on kujutatud roostepruunide areaalidega. Näited hästi tuvastatavatest fragmentidest (15 d) ja raskesti tuvastatavatest fragmentidest (15 e). Automaatne klassifitseerimine osutus tõhusaks punaka, lillaka, roheka ja rohekaskollaka tooniga turbasammalde puhul (summaarne tuvastamistõhusus 78%; joonis 15d) kuid tõhusus jäi madalamaks kollakaspruuni tooniga sammalde puhul, kus tuvastamist segasid õlgedele ning lagunevatele taimejäänustele sarnased spektraalsed omadused (joonis 15e). Samuti oli raskusi üksikute väga väikeste hajusalt paiknevate või osaliselt õlgedega kaetud väikeste fragmentide tuvastamisega. Sarnaselt drooniandmestiku töötlemisele on ka tavafotode töötlemise puhul eelduseks pildistamine sarnastes valgusoludes, suur õpetusandmestik ja suhteliselt väike eristatavate klasside arv. Suurem klasside arv või Random Forest/Bagging algoritmide kasutamine tekitab rohkem segaklasse, mille sisu on raskesti tõlgendatav. Lisaks RGB kaamerale katsetati Laiuse testalal ka infrapunakaameraga (IR) drooniseiret, et ühest küljest parandada RGB kaameraga kombineeritult taimkatteklasside eristamise võimet ja teiseks hinnata taimestumise edukust maapinna temperatuuri alusel (suvine kõrge pinnatemperatuur on hüpoteesikohaselt taimestumisele oluline takistus) ning maapinna erineva soojenemise kaudu (kaks ülelendu IR kaameraga hommikul jahtunud maapinnaga ning pärastlõunal maksimaalselt soojenenud maapinnaga ajal) välja töötada maapinna niiskuse arvutamise metoodika. Paralleelselt IR droonilennule viidi läbi ka maapinnal kontaktmeetodil pinnatemperatuuri ja mullaniiskuse (m3/m3) mõõtmine (joonis 16). IR kaameraga testiti ka erineva lennukõrguse mõju 5 m kõrguse muuduga vahemikus 35-150 meetrit, sobivaimaks lennukõrguseks on taimkattestruktuuri määramiseks 70-80 m, maapinna temperatuuri kontrasti järgi niiskuse hindamiseks piisab ka 150 m lennukõrgusest.
a) b)
c) d) e)
Joonis 16. IR kaameraga mõõdetud maapinna temperatuur (23.aug.2018, kl. 15) ja samal ajal maapinnal kontaktmeetodil mõõdetud mullaniiskus (iga lilla ja kollane punkt tähistab mõõtepunkti). Jahutuseta laiatarbe infrapunakaamera droonidele osutus kogu uurimisala katva komposiitpildi koostamiseks ebatäpseks (vt. joonis 15 vasakpoolse kujutise lennusuunast sõltuvat triibulisust) ja mõjutab seeläbi lõpptulemust. Samas temperatuurikontrasti väärtused (pärastlõunasest temperatuuri komposiitpildist lahutatud hommikune temperatuuri komposiitpilt) korreleerusid mõõdetud mullaniiskuse väärtustega. Termokaameraga droon sobib suurepäraselt ka allikaliste kohtade või hilissügisest kevadeni lekkivate pinnaspaisude tuvastamiseks. Arvestades seda, et droonipildi alusel on väga keeruline (ja/või kulukas) koostada aastateülest homogeenset aegrida, on taimestikuseire puhul kõige tõhusam drooniseire kasutamine üldise taimestumise hindamiseks dominantliikide/koosluste alusel ning nende piiride pikemaajalise muutumise jälgimiseks. Lausalise kaardistamise aluseks võiks olla k-means meetodil loodud aluskaart (selle loomine ei eelda eelnevat ala seiret), mille klassidele antakse sisu georefereeritud väliuuringute abil. Eristatavate klasside arv sõltub kasutatud lähteandmestikust, jäädes Sentinel satelliidi optiliste kanalite ja indeksite kasutamisel enamasti 5-7 klassi vahemikku, drooniseire RGB andmete puhul 7-9 klassi ning multispektraalsete kanalite kasutamisel 10-12 klassi piiresse. K-means meetodil loodud dominantklasside arvu määramine on empiiriline, eeskätt ekspertteadmistel põhinev ning vajab reeglina 3-4 erineva versiooni loomist, mille puhul statistiliselt eristunud klassid sisustatakse georefereeritud välitööandmestiku alusel uurimisalal. Neist versioonidest valitakse lõpuks välitingimustes reaalselt tuvastatavate ja looduses eristuvate klasside alusel sobivaima klasside arvuga aluskaart. Seega on drooniseire kõige paremini kasutatav a) ala (visuaalse) eelhinnangu ja seirealade esindusliku paigutuse koostamiseks, b) väiksemate alade detailseks sagedaseks võrdlemiseks (nt. veepiiri või mingi taimestikuareaali aastaajaline dünaamika), c) termokaameraga vee liikumise ja allikaliste kohtade ning paisude lekete tuvastamine, d) sisend ajas dünaamilise ruumilise kasvuhoonegaaside mudeli jaoks taimestiku katvuse muutuse alusel (eeldab vähemalt 3-4 perioodi katmist igal aastal: varakevadine lumesulamine, kevaduvine tärkamine, suvine rohtse biomassi maksimum, sügisene samblarinde seisundi hindamise aeg). Küll aga eeldab selline detailsusaste drooniseire puhul suurt arvutusjõudlust, ajakulu ning arvestatavat rahalist ressurssi. Optiline satelliitseire tagab samaaegselt suure ala katvuse, kuid on väikse ruumilise lahutusega (piksel u. 5-30 m vahemikus) ja ei saa valida ilmastiku järgi sobivat ülelennu aega. Arvestades ülelendude sagedust ja meie laiuskraadil tavapärast pilvisust, on Sentinel-2 missiooni näitel kuu kohta keskmiselt kasutada 1-2 päeva kvaliteetset kujutist (valdavalt pilvevaba) huvipakkuvast alast. Sügisel ja talvel võib esineda kuid, mil kvaliteetset kujutist ei saadagi. Korrastamata jääksoode puhul on see piisav kuna muutused on üldjuhul väikesed (erandiks kevadeti üleujutatavad alad), aga korrastamisjärgseks seireks on see aastaajalise arengu dünaamika hindamiseks ebapiisav. Küll aga sobib selline sagedus pikaajaliseks (paljude aastate üleseks, st. enam kui 10-aastase perioodi muutuste) kindla fenofaasi või aastaaja alusel (madalsoo ja rohundirikka ala puhul kesksuvine, turbasammaldega aladel sügisene periood) hindamiseks. Ülelendude sagedus aga omakorda on seotud kaetava ala suurusega (piksli suurusega) – nii näiteks saab MODIS missiooni Terra ja AQUA satelliitide abil arvutada maapinna ööpäevase temperatuuri amplituudi, aga piksli suurus ulatub kilomeetrini ja huvipakkuva ala sisu kipub hägustuma kuna hõlmab nii freesturbaväljakuid, kraave kui servas ümbritsevat ala (joonis 17).
Jääksoo korrastamine 08-10.20219
Joonis 17. Päevane maapinna temperatuur (°C) Laiuse korrastamisalal (Laiuse 1) ja looduslikus seisundis rabametsas (Laiuse_natural) Terra satelliidi andmestiku alusel aastatel 2017-2021.
Samas on sel viisil aastane pidev temperatuurikäik uuritavalt alalt tagatud ja seda saab kasutada näiteks sisendina mullahingamise (Rsoil) või ökosüsteemi hingamise (Reco) modelleerimiseks nagu näidatud jääksoode näitel Burdun et al., 2021 poolt. Vaatamata madalale ruumilisele lahutusele on selline maapinna temperatuur sisendandmestikuna parem kui lähimas ilmajaamas mõõdetud õhutemperatuuri või maapinna temperatuuri vahetu kasutamine, kuna ilmajaam asub mineraalpinnasel, kus termiline režiim on soomuldadest erinev. Ökosüsteemihingamise modelleerimiseks nii vahetult mõõdetud kui kaugseire andmete alusel on sobilik järgmine valem (Riutta et al., 2007; Järveoja et al., 2016):
Metaanivoo hindamine satelliidi andmetel põhineva maapinna temperatuuri alusel ei anna häid tulemusi kuna metanogenees on seotud sügavama anaeroobse turbakihiga ning aereeritud tsooni temperatuur pigem soosib metaani oksüdeerimist/metanotroofide poolt tarbimist ja kahandab metaanivoogu ning selle seost sügavama kihi termiliste omadustega. Kui looduslikus soos metaanivoog ligikaudu järgib aastast temperatuurikäiku (mõningase ajalise nihkega), siis jääksoodes on seos nõrk ja olulisem on sademete hulk ning poorides vee küllastatu hapnikuga. Neid näitajaid paraku praeguste teadmiste kohaselt pinnakihist sügavamal kaugseire vahenditega piisava ruumilise lahutuse ning ajasammuga ei ole võimalik tuletada. Teataval määral võimaldab seda satelliitradarandmestik (SAR), kuid ka seal on avalikult kasutatava andmestiku lainepikkus sobiv vaid väga õhukese pinnakihi kirjeldamiseks.
Joonis 18. Mõõdetud ja satelliidi maapinnatemperatuuri andmete alusel modelleeritud Reco looduslikes soodes (Männikjärve, Linnusaare), kuivendusega jääksoode osas (Kõima 1, Kildemaa 2) ja jääksoo freesturbaväljadel (allikas: Burdun et al., 2021). Optilise kaugseire abil jääksoode korrastamistööde järgse arengu kirjeldamiseks on tavapärase nähtava valguse spektriosa (RGB) kõrval otstarbekas kasutada erinevate spektriosade alusel koostatud indekseid. Kuna jääksoode seisund, korrastamismeetodid (veekogu, metsastamine, rohttaimedega madalsoo-suunaline korrastamine, samblafragmentide laotamine, pinnaspaisude kasutamine isetaimestumisega jne.) on alade lõikes varieeruvad, on vajalik erinevaid indekseid kasutada. Madalaveeliste taimestuvate veekogude puhul annab parima tulemuse NRG indeks, taimestumata veekogu piiritlemiseks aga NDPI. Avavett ja väga niisket pinnast kajastavad paremini NRG ja NGR indeksid, kuid NGR puuduseks on see, et ei suuda edasi anda infot kuivema taimestumata turbaga piirkondade kohta (mis jääksoo korrastamise seisukohast on oluline määratleda). Rohundirikka taimestikuga jääksoo, metsastunud/metsastatud jääksoo kirjeldamiseks sobib hästi laialt kasutatav taimkatteindeks NDVI. Joonis 19 illustreerib 2020.a. korrastatud Maima jääksoo erineva taimestumismääraga (ja korrastamisviisiga) alade ning seda ümbritseva looduslähedase rabataimestikuga ala näitel erinevate indeksite võimekust seisundit kirjeldada.
Joonis 19. Sentinel-2 satelliidi andmete alusel arvutatud indeksid Maima korrastatud jääksoo näitel (21.09.2023). RGB (Red/Green/Blue) iseloomustab tavapilti nähtavas spektriosas, NRG (nIR/R/G) indeksit kus sinine spektriosa on asendatud lähisinfrapunaga, NDVI (normalized difference vegetation index) taimkatet kajastav indeks, NGR (nIR/G/R) sarnane NRG indeksiga niiskuse kirjeldamiseks, NDPI (Normalized Difference Pond Index; (mIR1- Green)/(mIR1+Green) ja NNR (nIR/nIR/Red).
Satelliidiseire andmete alusel kiire hinnangu andmiseks korrastamise edukuse kohta lühiajalise perioodi alusel (mõned aastad) on takistuseks väheste pilvevabade kaadrite esinemine. Atmosfääri läbipaistvus (eriti pilvisus, veeaur) mõjutab oluliselt kõigi optilise seire kanalite alusel arvutatud indeksite väärtust ja võib mõjutada arvutatud ajalisi trende. On üsna sage, et kogu kuu lõikes pole ühtegi hea lahutusega (piksel 10m või väiksem) pilti kogu uurimisala kohta ning erineva pilvisusega tehtud piltide alusel komposiitpilt lahendab probleemi vaid osaliselt. Joonis 20 iseloomustab Maima jääksoo näitel 2022 sügisest (kuiva pika põuase suvega aasta) ja 2023 sügiseni (kuiva kevadsuvega aasta) näitel ühe aasta jooksul RGB, NDVI ja NRG indeksite aastaajalist dünaamikat. Tähelepanu tuleks pöörata Ala 1 (kood M), 5 (B), 7 (N) kiirele taimestumisele valdavalt pilliroo, villpea ja tarnadega ning samblafragmentide laotamisega kuid kõrge veetaseme all kannatavate alade 3 (K) ja 4 (C) kokkuveoteeäärse tsooni muutustele ning normaaltingimustes sobiliku ala 9 (P) seisundi muutusele.
Joonis 20. Korrastatud Maima jääksoo seisundi muutus iga kuu parima kvaliteediga (pilvevabama) pildi alusel RGB (vasakpoolne veerg), NDVI (keskmine veerg) ja NRG (parempoolne veerg) näitel 2022 sügisest alates kuni 2023.a. sügiseni. NDVI mustja ja punakad toonid iseloomustavad rohelise taimestikuta (ja/või veega ning tehispinnasega alasid, tumeroheline lausaliselt taimestunud alasid).
Joonis 20. järg
Joonis 20. järg Joonis 20 illustreerib hästi kuidas 2023.a. väga kuiva suve järel kahanes juulini veega kaetud ala, aga septembris oli taastunud liiga kõrge veetase peaaegu kevadise seisuni (eriti ilmekas Ala 6-1, 4, 1, 7 ja 11 näitel, eriti NRG indeksiga väljendatuna). Seejuures pilvevabade piltide puudumise tõttu ei tule kaugseire andmetest välja, et muutus toimus lühikese aja jooksul just vahetult pärast augustikuise pildi tegemist ning järgmiste ülelendude ajal oli taevas lausalise pilvkattega.
Satelliidi radarandmestiku (SAR) puhul on eeliseks selle vähene sõltuvus ilmastikust või pilvisusest, kuid ülelendude sagedus on väike ja aluspinna koherentsuse muudu alusel pindalaline lahutusvõime tagasihoidlik (enamasti vajalik hektarile lähenev pindala, et sisulisi muutusi ajas eristada). Kõrgusmuudu kaudu niiskusrežiimi muutumise hindamine DInSAR (järjestikuste kujutiste faaside alusel arvutamise meetod) on soos võimalik (nn. soo hingamise mõõtmine) ja enamasti üsna täpne (mõõdetav millimeetrites), kuid probleemiks on ülelendude sagedus, sest erandlikel juhtudel võib kahe pildi vahelisel perioodil maapinna kõrguse muut sadude tõttu ületada faasi ulatust (Sentinel 1 C-band puhul u. 2.5 cm) ja sel juhul tegelik kõrgusmuut jääb teadmata arvu faaside võrra ekslikuks (Tampuu et al., 2023). SAR andmestikku on võimalik kasutada muutuste tuvastamiseks ka koherentsuse kaudu. Sel juhul on soodes vertikaalne-vertikaalne polarisatsioon muutuste kirjeldamiseks tõhusam kui vertikaalne- horisontaalne polarisatsiooni kasutamine, kuna viimasel on just jääksoodes suurem hajuvus (joonis 21).
Joonis 21. Sünteetilise apertuurradari (SAR) erinevate polarisatsioonide hajuvus (nii tõusva kui laskuva suhtelise orbiidi RON alusel 6-päevase sammuga andmestiku põhjal lumevabal perioodil) loodusliku lageraba, jääksoo ning kasutuses oleva freesturbavälja võrdluses (Tampuu et al., 2020). Maima jääksoo uurimisperioodi hõlmav koherentsuse muutuses endisel freesturbaväljal ja turbavõtuaukudega alal võrreldes loodusliku taustaalaga tuleb väga selgelt esile järsk muutus freesturbaväljal alates 2020 a. lõpust (joonis 22), mil veetase järsult freesturbaväljakutel tõusis ning seejärel kajastuvad 2022.a. kuiv suvine-sügisene ning 2023 kuiv suvine periood kasvava koherentsusena (veega kaetud ala kahaneb). Looduslik ning turbavõtuaukudega ala reageerivad 2022 põuale aga vastandsuunalisena (kuiva turbasambla niiskus ja vastavalt elektrijuhtivus kahaneb).
Joonis 22. SAR kahe suhtelise orbiidi (RON 58 ja 80) alusel vertikaalne-vertikaalne polarisatsiooniga jääksoo korrastamisega seotud muudatuste tuvastamine Maima jääksoos endisel freesturbaväljal (Maima_frees), turbaaukude piirkonnas (serv) ning raba looduslikul taustaalal (looduslik).
Ka Ess-soo uurimisalal on täheldatavad sarnased muutused SAR andmestiku alusel (joonised 23, 24).
Joonis 23. Ess-soo SAR pilt suhteliselt orbiidilt RON 160 kevadel kõrgema veetasemega perioodil 1. märtsil 2022. Sinakad toonid iseloomustavad madalat koherentsust (puurinne, vaba veepind) ning kollakad ja punakad toonid suuremat koherentsust. Mustad piirjooned tähistavad Ess-soo erinevaid korrastamisalasid, millest on välja jäetud eraldavad pinnaspaisud, kokkuveotee ning kraavid ja üleminekulised tsoonid.
Joonis 22. SAR andmete alusel vertikaalne-vertikaalne polarisatsiooniga jääksoo korrastamisega seotud muudatuste tuvastamine Maima ja Ess-soo jääksoos. Ülemine joonis iseloomustab 2020.a lõpus järsu veetaseme tõusu tõttu suurt muutust Maima jääksoos, kuid 2021.a. sügisel Ess-soos sarnast õleujutust ei esinenud ning muutus koherentsuses on tagasihoidlikum. Alumine joonis iseloomustab korrastatud alase väga sarnast sünkroonsust maapinna niiskuse muutuses põua tõttu 2022 ja 2023.a., kuid toob ka välja erineva suhtelise orbiidi (RON) valiku olulisuse niiskuse kirjeldamise seisukohast.
Veetaseme dünaamika Veetaset, kasvuhoonegaaside voogu ning Maimas ja eriti Ess-soos värskelt korrastatud aladel samblafragmentide kasvama minekut (ka laiemalt alade taimestumist) mõjutas väga tugevalt 2022.a. ja 2023.a. ilmastik. Kui Maima jääksoos tõusis pärast korrastamist 2020.a. lõpus ja 2021.a. veetase sammalde kasvuks ebasoodsalt kõrgeks, siis Ess-soo korrastamisele järgnes kaks väga kuiva suve. Talv algas suhteliselt varakult 2021.a. keskmisest külmema detsembriga kuid maapind külmus varase lumikatte (novembris) tõttu vaid osaliselt ja keskmisest soojem talve jätk (joonis 3) soodustas nii lume sulamist kui külmumata pinnasest gaasivoo eritumist. Korrastamisalade seisukohast oli aga kõige olulisem sulailmadega lumeveevaru kahanemine ja (pool)külmunud pinnasel tekkinud lombid, mis tuule tekitatud lainetusega uhtusid samblafargmentide katteks laotatud põhu ning osaliselt ka samblafragmendid vaaludesse. Talvistele keskmiselähedastele sajuhulkadele järgnes aga 2022.a. erakordselt kuiv märts (joonis 3) ning kogu järgneva aasta jooksul oli iga kuu sademete hulk ligi 40% väiksem pikaajalisest kuu sademete normist. 2023.a. kevadsuvi osutus aga veelgi kuivemaks ja veetase alanes taas väga kiiresti, langedes Maima jääksoo võrdlusalal ning samblafragmentide laotamisega pinnaspaisudega alal 9 (kood P) ligi 60 cm sügavusele maapinna suhtes. Juulis alanud sademed küll tõstsid veetaset, aga optimaalse tasemeni (-20 cm) jõudis see alles septembris (joonis 23).
Joonis 23. Kuu keskmise veetaseme dünaamika Maima jääksoo korrastatud aladel ning võrdlusalal. Alade tähises „sph“ näitab turbasamblafragmentide laotamist, „Pais“ ala korrastamist ainult pinnaspaisude rajamisega, „Täis“ lausaliselt pinnasega täidetud kraave. Halli varjutusega ala indikeerib eelistatud veetaseme vahemikku korrastatud alal (veetase maapinna suhtes vahemikus 0...-20 cm). Laiuse jääksool on Lehtmetsa raba näol suur tagamaa madalaveelisel veekogul ning mõningane põhjavee toide, mis koostoimes Lehtmetsa peakraavil toimetavate kobrastega tagasid suhteliselt hea veetaseme stabiilsuse kogu korrastatud ala ulatuses (v.a. kõige kõrgema maapinnaga väike eraldatud idapoolne nurk) ja veetase oli kogu aasta ulatuses vahemikus 0...-40 cm (joonis 24). Sellest tulenevalt algas 2022 aastal ja jätkus 2023.a. jõudsalt ülepinnaline taimestumine Laiuse kesksel korrastamisalal (kood Laiuse E) ning läänepoolsel alal (Laiuse W), kuid jäi puudulikuks kõige kuivemal väikesel idapoolsel alal. Samuti laienes keskmiselt 4.4 meetri võrra veekogu suunas taimestunud vöönd madalaveelise veekogu põhja-, edela- ja lõunaservas, mis on madalamad ja laugema kaldaga. Kõima jääksoos on küll veetase tänu suurele looduslikule puhverdavale tagamaale ning juba algselt lausalisele samblakattele optimaalse lähedal, aga nii 2021. kui 2022.a. on veetase ilmastikust tingituna augustiks langenud madalamale kui eelnevatel aastatel. Seevastu Kõima turbavõtuaukude veetase on oluliselt tõusnud (eriti gradiendiga korrastamisala edelaosa suunas) ja turbavõtuaukude vahelised tervikud on muutunud niiskemaks, veetase kõrgem (Kõima S tervik; joonis 24) kui võrdlusalal ja kvalitatiivselt on märgatav ala lääne- ning edelaosas tervikute servades turbasambla laienemist aukudest tervikule, kanarbiku ja samblike hääbumist ning nokkheina ja villpea lisandumist.
Joonis 24. Kuu keskmise veetaseme dünaamika Kõima ja Lause jääksoo korrastatud aladel ning võrdlusalal. Kasvuhoonegaaside voog Korrastamise käigus saavutatud kõrge veetase on kahandanud turba lagunemise kiirust ja süsihappegaasi lendumist korrastatud aladelt nii Kõima, Laiuse, Ess-soo kui Maima jääksoos. Peamine mõju on Maima ja Ess-soo alal saavutatud turba lagunemise aeglustumise kaudu, Laiuse jääksoos aga ka kiiresti arenema hakanud taimestiku tõttu (peamiselt karusammal, jõhvikas, pilliroog, lääneoas ka turbasammal). Juba algselt lausalise taimkattega Kõima jääksoos gaasivoo osas statistiliselt olulisi muutusi ei ole, pigem on muutused selgitatavad aastate vahelisest ilmastiku erinevusest. Kõima jääksoo puhul on turbavõtuaukudes kõrgema veetaseme tõttu edelapoolses osas lopsakalt arenemas älvestele iseloomulikud turbasambla liigid ning kohati laieneb turbasammal ka madalamatele terviku osadele. Enamasti on tervikud siiski aeglase taimestumisega ja gaasivoogu mõjutab enam turba niiskusrežiimi muutus. Maima jääksoo kontrollala nr. 2 on aastaringselt lausaliselt 30-50 cm paksuse veekihiga kaetud ja jäetud antud analüüsist välja kuna ei vasta enam kontrollala kriteeriumitele. Kontrollala nr. 1 on samuti korrastamistööde järel märjemaks muutunud (eriti kevadel ja sügisel), mistõttu ka põuasel 2022 ja 2023.a. suvel oli seal veetase sarnane uuringuperioodi algusega, aga 2023.a. ei avaldunud see mõju veel taimestiku arengus väljaspool kraavi servasid ning ala 8 (O) piirdevalliga külgnevat peenart, kus on intensiivne jõhvika areaali laienemine. Kogu endise freesturbavälja ulatuses on domineeriv mullahingamine, autotroofne hingamine ja taimede fotosüntees on aastase voo mõttes enamasti tagasihoidlik. Erandi moodustavad pillirooga kattuvad alad (Ala 1 (M), 5 (B), 7 (N) ja turbasamblaga endised turbavõtuaugud (ala 12 (G)), kus keskpäevane ökosüsteemi CO2 sidumine (NEE, Net Ecosystem Exchange) võib ulatuda pilliroo puhul -192 mg CO2-C m2 h-1 ja turbasamblal -77 mg CO2-C m2 h-1. Enamasti jääb siiski aeglase taimestumise, laotatud põhu ja surnud samblafragmentide tõttu NEE isegi suvekuudel Maimas emissiooni poolele. Kui 2021.a. oli samblafragmentidega korrastatud aladel süsihappegaasi emissioon ligi poole väiksem kui kontrollalal ning lausalise kraavide täitmisega alal omakorda väiksem kui pinnaspaisudega suletud kraavidega alal, siis 2022.a. sellist erinevust ei esinenud ja vaid suuremalt jaolt veega üleujutatuks jäänud alad (C ja K) olid teistest väiksema emissiooniga, kuid 2023.a. oli ka neil aladel voog ülejäänuga sarnasemaks muutunud. Sellest tulenevalt on ökosüsteemi hingamine (Reco) jätkuvalt hea indikaator süsihappegaasi emissiooni väljendamiseks (joonis 25), mis toob kombineeritult välja nii mullahingamise kui taimestiku arengu mõju. Ökosüsteemi hingamine jäi vaatamata kahele järjestikusele soojale kuivale suvele valdavalt samale tasemele kui eelnevatel aastatel. 2021.a. veega kaetud aladel aga 2022 ja 2023.a. põuastel suvedel vesi soojenes kiiresti ja veetase alanes, jättes maapinna kohati mudaga kaetuks ja suurendades süsihappegaasi voogu. Erandlik on joonisel ala B (paisudega suletud kraavid, veega osaliselt üleujutatud), kus 2022.a. suvine Reco CO2-C piik on seotud intensiivse pilliroo kasvuga ning taime hingamine kombineerub sooja mudaja pinnase emissiooniga. Lisaks mõjutas üleujutatud alade voogu ka surnud kanarbiku jmt. lagunemine. 2023.a. taimestumise tõttu päevane NEE suurenes sel alal ligi 20 mg CO2-C m2 h-1 võrra ja surnud taimede lagunemine on aeglustunud.
Joonis 25. Ökosüsteemi hingamine (Reco) Maima jääksoos. Märge „veega“ iseloomustab korrastamise järgselt üleujutatud ala, „norm“ tähistab normaalse veerežiimiga ala, kus veetase jäi valdavalt maapinnast sügavamale. Aasta CO2 bilanss oli 2022.a. sambla fragmentidega korrastatud üleujutatud aladel emiteeriv (0.46 t/ha C), kraavidel pinnaspaisudega korrastatud aladel 0.86 t/ha C ning koos fragmentide laotamisega aladel 0.75 t/ha C. Turbaaukudes aga toimus tänu ohtrale päikesekiirgusele ning optimaalse lähedasele veetasemele (kohev sammal liigub sünkroonselt veetaseme muutusega, veetase 0...-15 cm) sidumine NEE -1.04 t/ga C. 2023.a. kuni augustini samblafragmentidega korrastatud aladel päevane NEE suurenes sidumise suunas ligi 10 mg CO2-C m2 h-1 võrra, aga enamikul suvekuudel jäi endiselt emiteerivaks isegi päevasel fotosünteesi toimumise ajal väga hõreda taimestiku tõttu. Kõima jääksoos on võrdlusala emiteeriv (1.9 t/ha C), turbaaukude vaheline tervik emiteerib 2.6 t/ha C, samas kui turbaaugu emissioon on 1.1 t/ha C ning kuiva suve tõttu oli ka looduslähedases seisus rabaosa emiteeriv (0.6 t/ha C) ning 2023.a. suvekuudel emissioon kasvas 2022.a. võrreldes veelgi. Laiuse jääksoos algas 2022.a. suve teises pooles ja jätkus kogu 2023.a. kiire taimkatte levik varasemalt palja turbaga alal. Kasvuala laiendasid kõige jõudsamalt karusammal ja jõhvikas, kraavides pilliroog, tarnad ning valge vesiroos. Läänepoolses osas kus turbaaukudele rajati lainetõkked, laienes kiiresti pilliroo ning hundinuiaga kaetud ala, tervikutel ja madalamates niiskemates lohkudes turbasammal. Kiire taimkatte muutuse tõttu allus mõõtmisandmestik modelleerimisele gaasimõõtmisrõngaste lõikes erinevalt (R2 0.43-0.95). Laiuse 1 (võrdlusala) on läbi kõigi aastate olnud CO2 emiteerija, Laiuse idapoolne (Laiuse E) korrastamisala oli 2021.a. emiteeriv, kuid 2022.a. saavutas sidumise jõhvikaga kaetud alal ning pillirooga taimestunud alal. Kõige märjemal alal (Laiuse W) on 2022.a. süsinikuneutraalsed või siduvad kõik taimestunud alad (joonis 26). Kraavide ning Laiuse madalaveelise veekogu süsinikubilanss on positiivne, keskmine emissioon 0.42 t/ha C. Seisva veega kraavides võib küll suve alguses vetika vohamise tõttu mõnel kuul süsihappegaasi sidumine olla intensiivne, aga suve teises pooles algab tekkinud biomassi lagunemine ja eritub nii süsihappegaasi kui metaani. 2023.a. suurenes päevane NEE sidumine kõigis korrastatud jääksoo osades 10-30 mg CO2-C m2 h-1 võrra võrreldes 2022.a., kõige enam pillirooga taimestunud pinnaspaisuga suletud kraavil.
Joonis 26. CO2 bilanss korrastatud Laiuse jääksoos. Laiuse 1 on võrdlusala, Laiuse E keskne korrastamisala võrdlusalast idas ning Laiuse W kontrollalast läänes paiknev maapinnalähedase veetasemega korrastamisala. Kõrge maapinna temperatuur ning suhteliselt kõrge veetase soodustavad metaani teket. 2022.a. olid pika põua tingimustes Maima jääksoos metaani tekkeks äärmiselt soodsad tingimused. Kuigi veega kaetud korrastatud aladel oli suvel keskmiselt kõrgem metaani emissioon, oli ka nii pinnaspaisude kui täidetud kraavidega korrastatud alasid, kus metaani voog oli suur. Samas pinnaspaisudega ala 10 (D) ja võrdlusala olid endiselt väga madala metaani emissiooniga, aga eelneval 2021.a. suvel oli just ala 10 kõrge vooga kui seal veetase lühiajaliselt väga kiiresti muutus. 2023.a. jäi eelnevate aastatega võrreldes metaanivoog oluliselt väiksemaks oli korrastamismeetodist sõltumatult sarnane.
Joonis 27. Kuu keskmine süsiniku kadu metaanina lendumise kaudu Maima kontrollalal (2017-2023) ja korrastamisjärgselt nii kontrollalal kui korrastatud aladel.
Naerugaasi voog oli 2022.a. sarnaselt eelnevatele aastatele toitainevaestes tingimustes kõigis uuritavates jääksoodes ebaoluliselt väike (joonis 28). Suhteliselt pika kuiva perioodi ja hoovihmadest tingitud veetaseme kiirete kõikumiste tulemusel suurenes N2O voog korrastamise järgselt Maimal juba 2021.a. ning veelgi selgemalt 2022.a., aga ka need vood on väga väikesed. Ainsaks erandiks oli september kui pärast pikka põuaperioodi ja sügavale langenud veetaseme juures algasid intensiivsed sajuhood, mis kiirelt täitsid pinnaspoore ning soodustasid lühiajalist naerugaasi heidet. Sarnane põuajärgne järsk naerugaasi voo lendumine septembris leidis aset ka teistel uurimisaladel. 2023.a. kiire kevadine veetaseme alanemine ja püsimine stabiilsena kuni augustis ohtrate sademete tõttu veetase taas kiirelt tõusis ja stabiliseerus, jäi naerugaasi voog väga madalaks. Teistest aladest eristub juba teist põuast aastat järjest Maima võrdlusala 1, kus korrastamise järgselt on veetase muutunud kõikuvamaks (kevadel ja sügisel naaberalade tõttu veetase kraavides tõuseb) ning see on kaasa toonud võrdlusalal naerugaasi suurema emissiooni, mis absoluutväärtuselt on siiski ebaoluline.
Joonis 28. Naerugaasi emissioon Maima jääksoost perioodil 2017-2023. Märgalade gaasivood on ajaliselt ja ruumiliselt suure varieeruvusega, seetõttu on ennatlik paari korrastamisjärgse aasta ning ühe või kahe ala tulemuste põhjal teha järeldusi korrastamismeetmete tõhususe osas. Maima jääksoos on samblafragmentide abil taimestumise kiirendamine valdavalt ebaõnnestunud liiga kõrge ning kõikuva veetaseme tõttu, aga samas on kõikidel samblafragmentide laotamisega aladel vähemalt mingil määral hajusalt kasvama läinud samblaid ning lisandunud on teisi raba liike. Aladel kuhu samblafragmente ei laotatud ei ole ka sõltumata veetasemest või paiknemisest looduslikuma taimestikuga rabaosa suhtes turbasamblaid iseseisvalt alale ilmunud. Kaks põuast suve on Maimal taimestumist oluliselt kiirendanud, eriti aladel 1 (M), 2 (L), 5 (B) ning 7 (N). Turbavõtuaukude juures on korrastamise mõju vähemärgatav, ilmselt põuaste suvede tõttu, sest kevadel on turbaaukudes veetase tervikute tasapinnani, kuid suveks taandub oluliselt. Sama tähelepanek kehtib ka Ess-soo kohta, kus taimestiku taastumise aeg on olnud oluliselt lühem, aga Ess-soos on just turbavõtuaukude ja metsas eelkuivenduskraavide sulgemisega alal veetase püsinud hästi ka suvedel ning turbasammalde laienemine olnud kiire (sh. ekskavaatori tekitatud aukudes ja pinnaspaisude külgedel niisemates lohkudes). Seevastu Laiuse jääksoos on kõigil korrastatud väljakutel ilmunud vähemalt mõnes piirkonnas ka iseseisvalt turbasamblaid, kohati on turbasammalde areaali laienemine alates 2022.a. suve lõpust muutunud kiireks. Detailne ülevaade taimkatte muutustest seiratavate jääksoode püsiseireruutudes on esitatud aruande II osas „RMK taimestiku seire KOONDARUANNE.pdf“.
Tähelepanekuid ja soovitusi korrastamisalade põhjal Alade jagamine väiksemateks hüdroloogilisteks üksusteks on ennast õigustanud, vähendades nii veelgi ulatuslikumaid üleujutusi või ulatuslikumaid liiga kuivi alasid. Iga eraldusvalli sees tuleb väljaku madalama osa juures tekitada ülevool, ülevoolude puhul tuleks kasutada reguleeritava kõrgusega ülevoolu lahendusi (joonis 29). Need on lihtsad, kuid võimaldavad vähemalt esimestel taimestumise seisukohast kriitilisel aastatel ilmastiku, projekteerimis- või ehitusvigade tõttu tekkinud veetaseme probleeme leevendada.
Joonis 29. Pinnasvall jääksoode eraldajana (vasakul) ning lihtne kuid tõhus reguleeritav ülevool veetaseme reguleerimiseks. Selliseid ülevoole tuleks kasutada iga hüdroloogiliselt eraldatud jääksoo eraldusvalli juures. Pinnaspaisudega suletavate kraavide puhul tuleks rajada igale kraavile pinnaspais iga 30 cm kõrgusmuudu kohta, aga vähemalt 3 pinnaspaisu. See võimaldab hüdroloogiliselt eraldatud üksustes juhtida väljaku madalamas piirkonnas lumesulamise vee serpentiinina läbi ala nii, et pinnaspaisudega kraavid oleks üle ühe ühendatud erineval pool kesksest pinnaspaisude reast (joonis 30). Antud lahendus on väga hästi toiminud Ess-soo põhjapoolsel alal (Alad 5, 7, 11, kood J ja H, C), soodustades vee pikemat säilitamist korrastatud alal, kuid kahandades suuremat üleujutust.
Joonis 30. Serpentiinina ühendatud kraavid Ess-soos.
Laiuse jääksoo Laiuse jääksoo oli esimene mis sai uuringualadest korrastatud 2019.a oktoobriks ning seega on korrastamisjärgseid muutusi saanud jälgida peaaegu 4 aastat. Esimesel kahel aastal oli taimestiku kujunemine aeglane ja kohati mõjutas liiga kõrge veetase, aga samas soodustas see veelindude saabumist uurimisalale, kes levitavad ka taimede seemneid ja eoseid, aga ka väetavad ala. Väetamise efekt on tugev kevadel, mil laudteed on lausaliselt väljaheidetega kaetud ning paigal seismist nõudvate välitööde korral on kasuks vihmavarju või kapuutsiga kummimantli kasutamine sõltumata ilmast. Mõju avaldub selgelt ka madalaveelise veekogu kõrgendatud DN sisalduses, mis tõuseb ka sügisese rändeperioodi ajal, kuid sügisvihmade lahjendava toime tõttu pole sama tuntav kui kevadel. Alates 2022.a. algas kiire taimestiku areng, mis jätkus jõudsalt 2023.a. 2023.a. algas ka madalaveelise veekogu kallastel taimestiku laienemine veekogu suunas. Jääksoo lääneosas laiematele kraavidele rajatud lainerahusti (joonis 31) on oma eesmärki täitnud suurepäraselt ja soodustanud kiiret taimestiku laienemist kraavides. Laienenud on peamiselt pilliroog ja hundinui, aga ka tarnad ja kohati turbasamblad.
Joonis 31. Laiuse jääksoo korrastamisalad. Taimkatte arengut Laiuse jääksoos enne korrastamist, vahetult pärast korrastamist ja uuringu lõpuaastal 2023 septembrikuiste satelliidipiltide alusel iseloomustab joonis 32. Võrreldes algseisuga on oluliselt paremini taimestunud ala loodepoolne osa, madalaveelise veekogu kallastele on kujunenud kuni mõnekümne meetri laiune sootaimedega taimestunud kaldavöönd-õõtsik, vaatamata puurinde eemaldamisele on kirdepoolne osa on NDVI indeksi väärtuse järgi taimestunud juba paremini kui kontrollala, aga kui kontrollala indeksit mõjutab eeskätt puurinne ja villpea, siis loodepoolses osas on domineerivad sootaimed (tarnad, pilliroog, rabakarusammal, jõhvikas jmt). Jõhvika areaal laieneb aastas keskmiselt 40-50 cm võrra peenarde keskosast ääreala suunas ja moodustab kohati lausalise katte.
Joonis 32. Taimkatte muutused Laiuse jääksoos enne korrastamist (2018), korrastanmise ajal (2019) ja uuringuperioodi lõpus (2023) Sentinel-2 satellidipildi alusel NRG ja NDVI indeksitena väljendatuna. Kõima jääksoo Kõima jääksoo korrastati 2019.a. lõpuks. Ala oli juba eelnevalt peaaegu lausaliselt taimestunud ja vaid üksikutes kohtades turbavõtu aukude vahelistel tervikutel oli taimestumata laike. Korrastamistööde käigus eemaldati suuremad puud evapotranspiratsiooni kahandamiseks, pinnaspaisudega suleti kraavid ja väljavool turbavõtu aukudest. Tänu eelnevalt olemasolevale rabataimestikule taastus kogu alal korrastamise käigus paljandunud pinnas kiiresti. Madalamad alad kattusid nii nokkheina kui turbasammaldega (joonis 33), kõrgemad pinnaspaisud peamiselt kanarbiku, karusambla ja villpeaga (joonis 34).
Joonis 33. Pinnaspaisude rajamiseks turba võtmise auk (vasakul) ja endine kirdepoolne kogujakraav (Kõima-N väljavool) on turbasammalde,villpea ning nokkheinaga kattumas.
Joonis 34. Kõrgemad pinnaspaisud kattuvad villpea, kanarbiku, karusambla ja murakaga, madalamad servad ja turba võtmisel tekkinud lohud nokkheinaväljaga.
Joonis 35. Pinnaspaisude tõttu seisva veega kraavid ning turbavõtuaugud täituvad turbasammaldega, tervikutel märjemates piirkondades kanarbik hääbub.
Joonis 36. Edelapoolses osas kus veetase on kõrgem ja püsivalt maapinnale lähedal ka põuastel suvedel (maapinna kalle tagab vee pealevoolu) on ka suuremad pinnaspaisud peaaegu täielikult taimestunud.
Joonis 37. Edelaoas lausaliselt täidetud kogujakraav ning vee liikumist tõkestavad massiivsed pinnaspaisud hoiavad turbavõtuaukudes veetaset kõrgena ka kesksuvel ning tagavad soodsad tingimused kiireks taimestumiseks. Turbasammaldega kaetud areaal on nelja aastaga jõudsalt laienenud.
Joonis 38. Edelapoolne väljavool on aastaringselt kuiv, soost valguv vesi on leidnud endale tee metsa alla, kuhu valgub ühtlaselt laial alal. Maima jääksoo Maima jääksoo korrastamine toimus 2020.a. sügisel ja oli uurimisaladest esimene kus kasutati kõiki erinevaid korrastamisvõtteid (madalaveeline veekogu, pinnaspaisud kraavidel, kraavide lausaline täitmine, pinnaspaisud kraavidel ja turbasambla fragmentide laotamine, kraavide lausaline täitmine ja turbasambla fragmentide laotamine, turbavõtuaukude väljavoolude sulgemine). Kavandatud tegevused osaliselt ebaõnnestusid ebaõige veetaseme tõttu, kuid soovitust kõrgem veetase ei takista soostumist. Turbasammalde areng ja levik on alal liiga kõrge või muutliku veetaseme tõttu piiratud, pilliroo, tarnade, villpea ja nokkheina, üksikutes piirkondades ka jõhvika laienemine on viimase aastaga kiirenenud.
Taimestumist kiirendas kõige enam 2022 ja 2023.a. põuased suved, mis tagas taimede arenguks soodsama veetaseme. 2021.a. lõpus väga edukalt laienenud nokkheina areaali alal 2 (L) ja 10 (D) hävitas peaaegu täielikult sügisrände eel 100-200-pealine sookurgede parv, mis rebis taimed lausaliselt juurtega välja. Uurimisalal on kohatud merikotkast (sageli Ala 4 rabapoolsel küljel kõrgema männi ladvas), kuni 20 luigest koosnevat parve, koovitajaid, põtra, hunti ja pruunkaru. Uurimisala projekteerimisel/korrastamisel tehtud suurim eksimus oli liigse vee äravoolu planeerimine läbi olemasoleva osaliselt täidetud kogujakraavi. Kõrge veetaseme korral täidetud kraavis mudajas mass kerkib koos veetasemega ja takistab vee äravoolu, suvel alaneva veetaseme korral aga alaneb ka mudajas mass äravoolukraavis ja pigem soodustab kiiremat veetaseme alanemist turbaväljal. Maima eksimust võeti arvesse Ess-soos, kus kogujakraav täideti või sulgeti pinnaspaisudega ja liigvee äravooluks kujundati eraldi voolunõva serpentiinina läbi turbaväljade.
Joonis 39. Madalaveelise märgala (Ala 1, kood M) veetase jäi planeeritust madalamaks kuna soovitud veetaseme korral oleks kõik rabapoolsed väljakud veelgi sügavamalt üleujutatud olnud. Taimestumise seisukohast on veetase alal soodne ja pilliroo, hundinuia ning tarnade jõudne levik algas 2022.a. ja 2023.a. sügiseks on taimestumine peaaegu lausaline.
Joonis 40. Kraavide lausalise täitmisega ja samblafragmentide laotamisega alal (ala 3, K) on veetase liiga kõrge, valdava osa aastast on ala veega kaetud ja taimestunud on peamiselt täidetud kraavide kohad, kus maapind kerkib koos veetasemega. Siiski leidub hajusalt ka üksikuid elusaid turbasamblaid.
Joonis 41. Ebasoodsalt kõrge veetaseme puhul toimub taimestumine kiiremini just täidetud kraavide kohal kuna seal maapind kerkib koos veetasemega. Taimestikus domineerivad villpead, hundinui, tarnad, nokkhein ja mätaste vahel üksikuid turbasamblaid, mis on fragmentide laotamisest säilinud. Veega kaetud ala on luikede kasutuses.
Joonis 42. Pinnaspaisudega suletud kraavidega turbasambla fragmentide laotamise alal on taimestumine võimalik vaid soodsa niiskusrežiimiga vööndis. Liiga sügava veega alal toimub aeglane taimestumine pilliroo ning hundinuiaga. Sobivates tingimustes on turbasambla katvus hea ja sammal elujõuline.
Joonis 43. Pinnaspaisudega suletud kraavidega alal 5 (B) toimub looduslik taimestumine kõrge veetaseme tingimustes ja levivad madalsoole iseloomulikud liigid. Taimestumise kiirust toetab sel alal lahustunud lämmastikuga rikastunud põhjavee väljakiildumine.
Joonis 44. Kraavide lausalise täitmisega ja samblafragmentide laotamisega alal (ala 11, E) on veetase liiga kõrge, valdava osa aastast on ala veega kaetud. See ala vajas korrastamisel pinnase suuremamahulist tasandamist ja seetõttu pole pindmine turbakiht tihedalt alumiste kihtidega seotud ning liigub koos veetasemega kaasa. Taimestunud on peamiselt täidetud kraavide kohad, kus maapind kerkib kergemini koos veetasemega, aga taimestumine on ulatuslikum kui sarnaselt töödeldud alal 3 (K).
Joonis 44. Lausaliselt täidetud kraavidega alal (10, D) toimub iseeneslik taimestumine ebaühtlaselt. Kanarbik ja sinikas hääbuvad, villpea, nokkhein, tarnad ja jõhvikas laiendavad areaali. Vaatamata sobivatele niisketele laikudele ala sees ja külgnemisele samblafragmentide laotamise alaga, pole iseseisvalt turbasamblaid ilmunud.
Joonis 45. Pinnaspaisudega suletud kraavidega turbasambla laotamisega alal (9, P) oli esimesel aastal sammalde elulevus väga hea, kuid järgneval talvel kannatas kõrge lumesulavee uhtumise ja tugeva külmakohrutuse all. 2022/2023 kohrutuse kahju kordus. Siiski on kogu ala samblafragmentidega hajusalt
kaetud, püsivad koloniseerimistuumakesed tekkinud ning alal on esindatud paljud tüüpilised rabaliigid. Taimestumine on küll oodatust aeglasem, aga püsiv. Sel alal on niiskemal perioodil sambalaga paremini kattunud gaasirüngastes mõõdetud päevasel ajal ökosüsteemi hingamist ületavaid CO2 sidumise väärtusi.
Joonis 46. Võrdlusala on kõige kehvemini taimestunud. Alustaimestikus domineerivad üksikud hajusalt paiknevad villpeamättad, kraavi kallastel ka jõhvikas, samblikud. Kased ja männid kannatavad mineraliseerumise ning tuuleerosiooni tõttu paljanduvate juurte käes. Kuigi korrastamise käigus võrdlusala veetase tõusis, ei ole see veel oluliselt mõjutanud taimestumist.
Joonis 47. Pinnaspaisudega suletud kraavidega ja turbasambla fragmentide laotamisega ala (7, N), mis oli korrastamise eelselt tugevalt pilliroo ja noorte mändidega kaetud, on esimesest aastast saati olnud kõikuva veetasemega, aga juba esimesel sügisel risoomidest võrsunud varred takistasid lainetusel laotatud samblafragmente ja kattepõhku ära uhtuda ning pilliroo vahel esineb ohtralt turbasammalt, huulheina, kanarbikku. Esimeste aastate tulemus on paljulubav ja samblad elujõulised, kuid ebaselge on kas pikemas perspektiivis hakkab pilliroog turbasammalt varjutama või suudab sammal moodustada tugeva ühtlase katte.
Ess-soo Ess-soo ala korrastati 2021. a. sügisel ja selle käigus tehti võrreldes Maima alaga projektis mitmeid muudatusi. Eeldatavad veetasemed modelleriti iga ala lõikes, äravooluteed planeeriti serpentiinina väljakute keskosa kaudu, kohati säilitati üleujutuste vältimiseks avatud kogujakraavi lõike ning looduslikuma rabaosa ja freesturbavälja vahele rajati kogujakraavile veekogu. Korrastamistööde käigus tehti jooksvalt täiendusi vastavalt nivelleerimise tulemustele, lisati pinnaspaise kraavidele ning rajati madala pinnasvalliga eraldatud terrasseeritud väljak. Kuigi korrastamisest on seireperioodi lõpuks möödunud alla 2 aasta ja mõlemad korrastamisjärgsed suved on olnud äärmiselt põuased ning ebasoodsad turbasambla fragmentide siirdamisega korrastamiseks, on üldtulemused siiski lootustandvad ja metoodilises mõttes võib Ess-soo korrastamist edukaks näiteks pidada.
Joonis 48. Kuigi lausaliselt täidetud kraavidega turbasambla fragmentide laotamise ala nr. 2 (N) on arvestatava pikisuunalise nõlvakaldega ja külgnev avatud äravoolukraaviga, on see põuased suved kõige edukamalt üle elanud ja elujõulisi samblafragmente esineb lausaliselt. Sarnaselt samasugusele korrastamismeetodile Maima jääksoos, on ka siin kiiremini taimestunud just täidetud kraaviga osa. Samas domineerib täidetud kraavi osas villpea, turbasammalt leidub hajusalt kõikjal ning kuivemal osal on enam kanarbikku. Kohati esineb ka nokkheina laike.
Joonis 49. Pinnaspaisudega kraavid hoiavad ka arvestatava nõlvakalde korral edukalt veetaset üleval. Samblafragmentide laotamine sel alal (1, L) pole enamasti sama edukas kui kraavide lausalise täitmisega naaberalal (2, N), kuid sobiva niiskusega piirkonnas on kujunenud ulatuslik lausalise turbasambla katvusega ala. Taimestumine toimub edukamalt ka kraavide kallastel, aga traavidevahelistel väljakutel on taimestik hõre ning elusaid turbasamblaid vähem kui naaberalal. Kas samblafragmendid on säilitanud kahe põuase suve järel elujõu, selgub järgnevatel aastatel.
Joonis 50. Kõikidel aladel kuhu samblafragmente on laotatud, on hajusalt elujõulisi turbasablaga laigukesi ning seisva veega kraavilõikudes sageli ka vohavat turbasammalt (ala 9, E).
Joonis 51. Kuigi pealtnäha mõõdab eddy covariance mast nukralt tühja välja (ala 10, D) CO2 ja CH4 voogu, on siiski kogu alal hajusalt elujõulisi turbasambla laigukesi ja soodsamate aastate saabumisel võib sambla katvus kiiresti laieneda. Sarnaselt Maima jääksoos üleujutatud mudasele väljale (ala 11, E) on ka siin esimese 2 aasta taimestumine väga tagasihoidlik, aga tüüpiliselt toimub taimestumine alguses kiiremini just täidetud kraavide kohal. Pioneerliigiks villpea, kuivematel aladel kanarbik, hajusalt elus turbasamblaid, nokkheina, huulheina.
Joonis 52. Looduslikult kujunenud sootaimestiku säilitamine korrastamise ajal kiirendab veetaseme tõstmisel maapinna kattumist taimedega. Alal 11 (B) on pinnaspaisude abil veetaseme tõstmisel ja stabiliseerimisel jõhvikas laiendanud kaetavat areaali 60-70 cm võrra aastas, kraavides hõljuval mudal laiutavad villpead ning servas laiendavad kasvuala nokkheinad.
Joonis 53. Avatuks jäävatel kraavilõikudel haost või põhupallidest tõkete tekitamine/säilitamine on kasulik nii heljumi kahandamiseks kui kraavi kinnikasvaise kiirendamise seisukohast. Kogujakraav K-17 on tõketevahelisel lõigul täitumas turbasammalde, ubalehtede, soovõhkade, villpeade ja tarnadega.
Joonis 54. Kuigi pinnaspaisud kraavidel ja terrasseerimine madala eraldusvalliga hoiavad veetaset võrdväärselt teiste korrastatud väljakutega, on alal 4 (F) turbasammalde kasvama minek oluliselt kehvem. Üheks põhjuseks oli külmunud kängardes fragmentide laotamine pärast tugevat öökülma külmunud maapinnale, teiseks põhjuseks oli sel alal erandlikult esinev külmakohrutus 2022/2023 talvel ning kolmandaks põhjuseks 2022.a. 30. augustil esinenud erakordselt intensiivne sadu (Korelas mõõdeti 24 h jooksul 84 mm sademeid), mis tulvaveega uhtus ära Ess-soo uurimisala peamise ülevoolu mulde (P3) ning uhtus peenarde kõrgematesse osadesse nii samblafragmendid kui kattepõhu.
Joonis 55. Näide meandreeruvast paisudega suletud kraave ühendavast vooluteest (vasakul) ning liiga kõrget veetaset vältivast voolunõvast (paremal).
Joonis 56. Kogujakraavi võib sulgeda laiade turbaga täidetud lõikudega, kus sulgev lävend on kaetud taimede juurtega tihedalt läbikasvanud mätastega. Selline veekogu aitab hoida freesturbavälja otstele iseloomulikku kõrgemat serva niiskemana ja tagab kiirema taimestumise ning väiksema erosiooni, mis muidu kannaks turvast madalamal paiknevatele laotatud samblafragmentidele. KOKKUVÕTE Korrastamata jääksood olid olulised CO2 allikad. Enne korrastamist oli CO2 emissioon sõltuvalt aasta ilmastikust ja alast 4.7 (3.2 – 8.3) CO2-C t/ha*a. Metaani emissioon oli tagasihoidlik 0.09 t CH4-C t/ha*a. Toitainevaese rabaturbaga jääksoode naerugaasi emissioon oli samuti väike (0.0003 N2O-N t/ha*a) ja korrastamisejärgsel oluliselt ei muutunud. CO2 voog korrastamisjärgselt kahanes ja Laiuse jääksoos neli aastat pärast korrastamist jõudis aastabilansina süsinikuneutraalsuseni. Teistel korrastatud aladel oli aasta bilanss CO2 osas jätkuvalt emiteeriv 0.4-1.9 CO2-C t/ha*a. Kuigi gaasivood on suuremad suvekuudel (v.a. naerugaas, mil puudub selge aastaajaline käik), võivad külmumata pinnasega talvekuud oluliselt mõjutada gaasivoo aastast bilanssi. Süsihappegaasi sidumist mõjutab kõige enam fotosünteetiliselt aktiivne kiirgus (PAR), temperatuur (õhu ja pindmise 10 cm mullatemperatuur). Viimastel aastatel Eestis enam Keskkonnaagentuuri hallatavates ilmajaamades PAR ei mõõdeta ja ainsad teadaolevad pidevad PAR mõõtmised toimuvad hetkel RMK jääksoodes paiknevates mõõtekohtades. Ilma PAR pideva aegreata ei ole ökosüsteemi puhasgaasivahetuse (Net Ecosystem Exchange, NEE) usaldusväärne modelleerimine võimalik. Korrastamisjärgse seire periood vastavalt 4, 3, 2 ja 0 aastat on ebapiisav, et teha järeldusi meetodite tõhususe, taimestumise kiiruse või kasvuhoonegaaside voo kahanemise kohta. Esimestel aastatel mõjutab kasvuhoonegaase samblafragmentidega korrastataval alal põhu ja surnud fragmentide lagunemine. Äärmiselt suur määramatus on seotud ilmastikuga. Taimestumine kiirenes alates kolmandast korrastamisjärgsest aastast, kuid selgusetu on kui suurt rolli selle juures mängisid viimased kaks põuase suvega aastat. Kõikidel aladel kus rakendati turbasamblafragmentide laotamist, on vähemalt hajusalt elusaid turbasambla kogumeid ja vähestel aladel moodustavad ka väiksemaid lausalise katvusega alasid. Meetodi edukust kahandas projekteerimisviga veetaseme osas Maima jääksoos ning vahetult korrastamisele järgnenud 2 väga põuast suve Ess-soos. Aladel kus turbasambla fragmente ei laotatud, iseseisvalt turbasamblaid kasvama hakanud ei ole. Samuti on sambla fragmentide laotamisega aladel
rohkem rabale iseloomulikke liike. Esimeste aastate tulemused näitavad, et samblafragmentide laotamise teel korrastatavate jääksoode puhul taimestuvad nii üleujutatavate kui põuast mõjutatud aladel kiiremine lausaliselt täidetud kraavidega alad, aga pikemas ajaskaalas ei pruugi see kehtida. Ka pinnaspaisudega kraavide kallastel laieneb taimestik. Kriitiline on siiski sobilik veetaseme vahemik ja suvine niiskuse olemasolu, see sõltub aga nii võimalikust külgnevast tagamaast kui konkreetsete aastate ilmastikust. Drooniseire on väga tõhus abivahend korrastatava alaseisundi eelnevaks kaardistamiseks, seirealade optimaalseks valikuks, allikaliste alade tuvastamiseks, pinnaspaisude lekete avastamiseks ning ligikaudseks pinnase niiskuse määramiseks. Taimkatte kaardistamiseks on võimalik kasutada k-means klasterdamisel põhinevat lähenemist koos välitööde käigus klassidele sisu andmisega või suure õpetusandmestiku olemasolul masinõppe meetodil (random forest, bagging jmt). Pikaajalise homogeense aegrea saavutamine taimkatte dünaamika kaardistamiseks on väga kallis (tehniliselt ning tööjõukulult), aeganõudev ja keeruline, mida omakorda mõjutab tehnoloogia kiire areng ning sensorite muutus. Satelliidiseire on jääksoode korrastamise tulemuslikkuse jälgimiseks asjakohane, kuid kasutegur on suurem pika seireperioodi puhul. Lühikese perioodi puhul jääb muutuste suhtes väga tundliku jääksoo dünaamika oluliselt kiiremaks (nt. veetasemete muutus ja üleujutatavate alade ulatus) kui pilvevabade piltide saamine satelliitidelt. Samuti eeldab selline seire suuremate seireruutude rakendamist maapealses seires, et andmestik oleks võrreldav piksli suurusega. Paljude klassikaliste indeksite kasutamise muudab keeruliseks ka jääksoodele sagedane olukord, kus taimede vahelt paistab vesi, mitte maapind. See raskendab ka muidu pilvedest vähem mõjutaud radari andmestiku kasutamist. Dendrokronoloogia abil on võimalik näha turbaväljade rajamisega kaasnevaid mõjusid, raskustega luua kronoloogiaid jääksoos kasvavate puude osas (puud erivanuselised ja seega muutuva nooruskasvuga ning samaaegselt kiire keskkonnatingimuste muutusega), aga veetaseme tõstmise avaldumise tuvastamiseks ei ole 3-4 aastat piisav. Männid jätavad ebasoodsates tingimustes aastarõngaid vahele ja seega nii lühikesed perioodid ei allu kronoloogia loomisele. Lahustunud orgaanilise süsiniku ja lämmastiku kontsentratsioonid korrastamisjärgselt küll kuni kaheks aastaks tõusid, kuid selgusetu on seos ilmastiku (kuumad põuased suved) ja korrastamistööde osakaalu osas. Ärakanne on aga tagasihoidlik kuna vee äravoolu esineb uuritud aladel 2-4 kuud aastas ja needki madalama kontsentratsiooniga hilissügisel ja varakevadel. Vooluhulga ja kontsentratsiooni järgi hinnates on süsiniku ärakanne DOC kujul jääksoodest vahemikus 62-87 kg/ha*aastas. Lagunemiskatse esmased tulemused näitavad, et peamine massikadu toimub väga kiiresti esimese aasta jooksul ning selles mängib omakorda suurimat rolli esimeste kuude jooksul leostumiskadu. Veetase ja taimestik mõjutavad lagunemist oluliselt. Erinevate taimsete materjalide (varis, peenjuured, erinevad liigid) lagunemiskatsete tulemused selguvad kolme aasta pärast.
7. PROJEKTIGA HAAKUVAD TEADUSTEEMAD, GRANDID, DOKTORI- JA MAGISTRITÖÖD, JÄRELDOKTORITE UURIMISTEEMAD, LEPINGUD: Teavitustegevus: Lühiartikkel projekti eesmärkidest Eesti Loodus 8/2017, lk. 5. http://www.eestiloodus.ee/arhiiv/Eesti_Loodus08_2017.pdf ja Rahvusvahelise Märgalade Kaitse Grupi kuukirjas IMCG Bulletin, June, 2017 pp. 13-14 http://www.imcg.net/media/2017/imcg_bulletin_1706.pdf. 04.10.2017 TÜ geograafia osakonna seminar projekti eesmärgi ja hetkeseisu tutvustamiseks ning võimalike täiendavate huviliste (omafinantseeringu korras) kaasamine ülikooli teistest uurimisgruppidest. 18.-19.oktoober 2017 Toilas Keskkonnaministeeriumi turbaümarlaual projekti eesmärgi ja hetkeseisu tutvustamine. Suuline ettekanne: A. Kull & G. Veber, Abandoned peat extraction sites – will future be wetter and better? 10.-12.10.2018 Tartu, 18th Baltic Peat Producers Forum. Jääksoode korrastamisega seonduvat on laiema üldsuse teavitamiseks käsitletud populaarteaduslikus väljaandes "Samblasõber" nr 23, 2020, lk 10-15: https://sisu.ut.ee/sites/default/files/samblasober/files/samblasober_23_0.pdf Esinemine ERR Aktuaalne Kaamera, Osoon ja Vikerraadios intervjuudega.
Magistritööd ja doktoritööd Ott Toomsalu, 2019. Jääksoodes toimuvate muutuste analüüsimine LiDAR andmetel. Magistritöö. Kaitsutud Tartu Ülikooli geograafia osakonnas. https://dspace.ut.ee/handle/10062/65031 MarjanSadat Barekaty, 2021. Compare the performance of applying Machine Learning concepts to landcover classification models using very high-resolution UAV data. Magistritöö. Kaitsutud Tartu Ülikooli geograafia osakonnas. https://dspace.ut.ee/handle/10062/72820 Kärt Erikson, 2022. Veerežiimi häiringute ja ilmastiku mõju hariliku männi (Pinus sylvestris L.) radiaalsele juurdekasvule Lehtmetsa soo näitel. Magistritöö. Kaitsutud Tartu Ülikooli geograafia osakonnas. https://dspace.ut.ee/handle/10062/82873 Tauri Tampuu doktoritöö: Application of spaceborne SAR polarometry and interferometry for landscape ecological studies in bogs (Tartu Ülikool, kaitstud 2022.a. augustis). Artiklid Birgit Viru, Gert Veber, Jaak Jaagus, Ain Kull, Martin Maddison, Mart Muhel, Alar Teemusk, and Ülo Mander, 2017. Winter nitrous oxide and methane emissions from drained peatlands. Geophysical Research Abstracts, Vol. 21, EGU2019-15964. The abstract identification number EGU2019-15964. https://meetingorganizer.copernicus.org/EGU2019/EGU2019-15964.pdf?pdf Tampuu, Tauri; Praks, Jaan; Uiboupin, Rivo; Kull, Ain (2020). Long Term Interferometric Temporal Coherence and DInSAR Phase in Northern Peatlands. Remote Sensing, 12 (10), ARTN 1566. DOI: 10.3390/rs12101566 Tampuu, T.; Praks, J.; Kull, A. (2020). Insar Coherence for Monitoring Water Table Fluctuations in Northern Peatlands. International Geoscience and Remote Sensing Symposium (IGARSS). IGARSS, 4738−4741. DOI: 10.1109/IGARSS39084.2020.9323709 Burdun, Iuliia; Kull, Ain; Maddison, Martin; Veber, Gert; Karasov, Oleksandr; Sagris, Valentina; Mander, Ülo (2021). Remotely Sensed Land Surface Temperature Can Be Used to Estimate Ecosystem Respiration in Intact and Disturbed Northern Peatlands. Journal of Geophysical Research Biogeosciences, 126 (11), e2021JG006411. DOI: 10.1029/2021JG006411 T. Tampuu, J. Praks, A. Kull, R. Uiboupin, T. Tamm, K. Voormansik (2021).Detecting peat extraction related activity with multi-temporal Sentinel-1 InSAR coherence time series. International Journal of Applied Earth Observation and Geoinformation,Vol. 98,102309, https://doi.org/10.1016/j.jag.2021.102309 Tampuu, Tauri; De Zan, Francesco; Shau, Robert; Praks, Jaan; Kohv, Marko; Kull, Ain (2022). Can Bog Breathing be Measured by Synthetic Aperture Radar Interferometry. 2022-July, 16−19. DOI: 10.1109/IGARSS46834.2022.9883421. Kull, Anne; Kikas, Tambet; Penu, Priit; Kull, Ain (2023). Modeling Topsoil Phosphorus—From Observation-Based Statistical Approach to Land-Use and Soil-Based High-Resolution Mapping. Agronomy, 13 (5), 1183. DOI: 10.3390/agronomy13051183 Palviainen, M., Könönen, M., Peltomaa, E., Pumpanen, J., Ojala, A., Hasselquist, E., Laudon, H., Ostonen, I., Renou-Wilson, F., Kull, A., Veber, G., Mosquera, V., and Laurén, A.: Processes affecting lateral carbon fluxes from drained forested peatlands, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-6367, https://doi.org/10.5194/egusphere-egu23-6367, 2023. Tampuu, T.; Praks, J.; De Zan, F.; Kohv, M.; Kull, A. (2023). Relationship between ground levelling measurements and radar satellite interferometric estimates of bog breathing in ombrotrophic northern bogs. Mires and Peat, 29, 1−28. DOI: 10.19189/MaP.2022.OMB.Sc.1999815
8. Projekti juht (nimi): Ain Kull
Allkiri: allkirjastatud digitaalselt
Kuupäev: allkirjastatud digitaalselt
9. Taotleja allkirjaõigusliku esindaja kinnitus aruande õigsuse kohta (nimi, amet): Ain Kull, kaasprofessor
Allkiri: allkirjastatud digitaalselt
Kuupäev: allkirjastatud digitaalselt
NB! Aruanne esitada elektrooniliselt aadressil [email protected]
1. Introduction Peatlands cover only ∼3% of the global land area (J. Xu et al., 2018), but they store 21% of global terres- trial soil carbon (C) (Scharlemann et al., 2014), which is double the amount in the world's forests (Pan et al., 2011). Approximately 80% of peatland C stock is stored in peatlands north of 45°N (Yu et al., 2010). Historically, intact northern peatlands have acted as a vast C sink with an estimated average rate of C accu- mulation of 18.6 g/m2 per year (Yu, 2011).
Intact peatlands bind atmospheric CO2 as C within peat (Clymo et al., 1998; Salm et al., 2012). However, peatlands also lose C through CH4 emissions due to shallow (ground-) water table depths (WTDs) and anox- ic conditions in the peat layer (Waddington & Roulet, 2000). CH4 has a more significant radiative efficiency than CO2 but a much shorter lifetime in the atmosphere (Change, 2013). Therefore, over a millennial time
Abstract Remotely sensed land surface temperature (LST) enables global modeling and monitoring of CO2 fluxes from peatlands. We aimed to provide the first overview of the potential for using LST to monitor ecosystem respiration (Reco) in disturbed (drained and extracted) peatlands. We used chamber- measured data (2017–2020) from five disturbed and two intact northern peatlands and LST data from Landsat 7, 8, and MODIS missions. First, we studied the strength of the relationships between fluxes and their in situ drivers (i.e., thermal and moisture conditions). Second, we examined the association between LST and in situ temperatures. Third, we compared chamber-measured Reco with the modeled Reco driven by in situ measured water table depth and (a) in situ measured surface temperature and (b) remotely sensed MODIS LST data. In situ temperatures were a stronger driver of CO2 fluxes in disturbed sites (repeated measures correlation rmR = 0.8–0.9) than in intact ones (rmR = 0.5–0.8). LST had a higher association with in situ measured temperatures in disturbed sites (mean rmR = 0.79 for MODIS) and weaker in the intact (hummocks and hollows) peatlands (mean rmR = 0.38 for Landsat and 0.48 for MODIS). Reco models driven by MODIS LST and in situ surface temperature yielded similar accuracy: R2 was 0.27, 0.66, and 0.67 and 0.29, 0.70, and 0.66 for intact and for drained and extracted sites, respectively. Overall, these findings suggest the applicability of LST as a proxy of the thermal regime in Reco models, particularly for disturbed peatlands.
Plain Language Summary Organic carbon (C) in the peat layer of peatlands has been accumulating for thousands of years. Under anthropogenic impacts, such as drainage for forestry, agriculture, or peat extraction, peatlands start releasing the accumulated C back into the atmosphere as CO2 and CH4 much faster than historical rates of C accumulation. CO2 and CH4 are the potent greenhouse gases that lead to climate warming. The thermal regime is among the main factors controlling CO2 and CH4 fluxes in peatlands. In this study, we demonstrated the potential of satellite thermal data for monitoring CO2 fluxes from intact and disturbed peatlands. We used a long-term (2017–2020) data set of CO2 data measured in seven Estonian peatlands. The thermal regime explains CO2 fluxes. Also, satellite thermal data better represent both the thermal regime and CO2 fluxes in disturbed rather than in intact peatlands. Furthermore, we modeled CO2 fluxes from natural and disturbed peatlands: first, with thermal data measured in the field, and second, with satellite thermal data. Both models yielded similar prediction accuracy, which suggests that satellite thermal data have the potential to be used for modeling CO2 fluxes from peatlands with a varying level of disturbance.
BURDUN ET AL.
© 2021. American Geophysical Union. All Rights Reserved.
Remotely Sensed Land Surface Temperature Can Be Used to Estimate Ecosystem Respiration in Intact and Disturbed Northern Peatlands Iuliia Burdun1 , Ain Kull1 , Martin Maddison1 , Gert Veber1, Oleksandr Karasov1 , Valentina Sagris1 , and Ülo Mander1
1Department of Geography, Institute of Ecology & Earth Sciences, University of Tartu, Tartu, Estonia
Key Points: • Temperature is a stronger driver of
CO2 fluxes in disturbed peatlands compared with intact ones
• Remotely sensed land surface temperature (LST) is a strong predictor of in situ thermal conditions in disturbed peatlands
• Remotely sensed LST has a great potential for modeling ecosystem respiration in disturbed peatlands
Supporting Information: Supporting Information may be found in the online version of this article.
Correspondence to: I. Burdun, [email protected]
Citation: Burdun, I., Kull, A., Maddison, M., Veber, G., Karasov, O., Sagris, V., & Mander, Ü. (2021). Remotely sensed land surface temperature can be used to estimate ecosystem respiration in intact and disturbed northern peatlands. Journal of Geophysical Research: Biogeosciences, 126, e2021JG006411. https://doi.org/10.1029/2021JG006411
Received 21 APR 2021 Accepted 14 OCT 2021
Author Contributions: Conceptualization: Iuliia Burdun, Ain Kull, Martin Maddison, Valentina Sagris, Ülo Mander Data curation: Iuliia Burdun, Ain Kull, Martin Maddison, Gert Veber Formal analysis: Iuliia Burdun, Martin Maddison Funding acquisition: Ülo Mander Investigation: Iuliia Burdun, Ain Kull, Martin Maddison, Gert Veber, Oleksandr Karasov, Valentina Sagris, Ülo Mander Methodology: Iuliia Burdun, Ain Kull, Martin Maddison, Gert Veber, Oleksandr Karasov, Ülo Mander Project Administration: Ain Kull, Ülo Mander Resources: Ain Kull, Valentina Sagris, Ülo Mander Software: Iuliia Burdun
10.1029/2021JG006411 RESEARCH ARTICLE
1 of 19
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
2 of 19
scale, intact peatlands have a cooling effect on the Earth's climate, despite being a source of CH4 (Günther et al., 2020).
Over the past three centuries, human activity and a warming climate have lowered WTD in peatlands, leading to the oxidation of the peat layer (Kotta et al., 2018; Leifeld et al., 2019; Regan et al., 2019; Swindles et al., 2019). In drained peatlands, the groundwater level has fallen and such peatlands consequently have a thicker aerobic layer. In extracted peatlands, in addition to increased drainage, the vegetation layer has also been removed. Under warmer oxic conditions, the peat layer decomposes and releases accumulated C as CO2 (Hanson et al., 2020; Rinne et al., 2020; Salm et al., 2009; Waddington et al., 2001). The rate of C loss can be 4.5–18 times faster than historical rates of C accumulation (Hanson et al., 2020). Currently, disturbed peatlands account for up to 10% of the global anthropogenic CO2 emissions annually (Leifeld & Menichetti, 2018). Hence, disturbed peatlands are a significant source of CO2 and have a long-term impact on climate warming (Leifeld et al., 2019; Ojanen et al., 2013). Notably, because of the CO2 emissions from disturbed peatlands, the global peatland biome is expected to shift from sink to source in this century (Leif- eld et al., 2019; Loisel et al., 2021).
CO2 exchange, particularly ecosystem respiration (Reco), strongly depends on the climatic conditions of dis- turbed peatlands, including soil and air temperatures (Lloyd & Taylor, 1994; Maljanen et al., 2010; Veber et al., 2018). For example, a temperature-dependent function is widely used to model spatial and temporal Reco from intact and disturbed peatlands (Alm et al., 2007; Bubier et al., 2003; Järveoja et al., 2020; Lafleur et al., 2001). In previous studies, C fluxes were shown to have positive exponential relationships with peat temperatures at different depths, including −20 cm (Helbig et al., 2019), −10 cm (Davidson et al., 2019), and −5 cm (Acosta et al., 2017), as well as with surface temperature (X. Huang et al., 2021). However, owing to a limited spatial coverage of in situ temperature measurements, the modeling of C fluxes is only possible at the plot scale. To overcome this limitation, remotely sensed parameters, including land surface temperature (LST), have been applied to allow a global modeling of Reco in peatlands (Lees et al., 2018).
Rahman et al. (2005) were among the first researchers to apply remotely sensed data for Reco modeling. They found that MODIS LST had an exponential relationship with Reco over the wide range of North American land covers, and furthermore, this relationship varied between land covers. Later, Kimball et al. (2009) de- veloped a terrestrial C flux model driven by remotely sensed inputs for boreal biomes; however, none of the validation sites were located in peatland. Subsequently, remotely sensed data were actively used to model C fluxes mainly for forest land covers (Crabbe et al., 2019; N. Huang et al., 2014, 2015; Jägermeyr et al., 2014; Olofsson et al., 2008; Tang et al., 2011; Wu et al., 2014; Xiao et al., 2010). The major part of these studies uti- lized MODIS data with a coarse spatial resolution (1 km for LST and 250 and 500 m for vegetation indices) with C data measured at eddy covariance towers and by chambers. So far, we know of only two studies that have utilized remotely sensed data of a higher spatial resolution, 30 m (Landsat), for CO2 fluxes estimation. The first study was conducted over beech forest (Crabbe et al., 2019) and the second, over forested peatland (C. Xu et al., 2020).
The relationships between Reco and remotely sensed LST in peatlands have received much less research attention than other ecosystems. Schubert et al. (2010) reported strong relationships between MODIS LST and Reco in different types of peatlands, including bog (precipitation fed) and fen (additionally fed with groundwater and sometimes surface runoff). Later, Y. Gao et al. (2015) and Ai et al. (2018) developed mod- els for Reco simulation driven by MODIS LST and enhanced vegetation index (EVI). Those models were vali- dated over large areas and diverse land covers, including marshes and wetlands, and both studies supported the existence of a strong relationship between Reco and LST. More recently, Park et al. (2020) and Junttila et al. (2021) applied MODIS LST to estimate Reco in tropical and northern peatlands, respectively. The case study on northern peatlands only included five peatlands (four fens and one bog), yet it demonstrated that the performance of LST varies between peatland types (Junttila et al., 2021).
Despite the progress made in estimating Reco with remotely sensed data, much uncertainty remains regard- ing the strength of the relationships between Reco and LST in disturbed (drained and extracted) peatlands. To our knowledge, no study has specifically addressed the applicability of LST for modeling CO2 fluxes in such peatlands. We present the first attempt to fill this knowledge gap to tap into the potential of remotely sensed LST, which is especially urgent, given the need to manage substantial CO2 emissions from disturbed
Supervision: Valentina Sagris, Ülo Mander Validation: Iuliia Burdun, Ain Kull, Gert Veber, Oleksandr Karasov Visualization: Iuliia Burdun Writing – original draft: Iuliia Burdun Writing – review & editing: Iuliia Burdun, Ain Kull, Martin Maddison, Gert Veber, Oleksandr Karasov, Valentina Sagris, Ülo Mander
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
3 of 19
peatlands. This article aims to quantitatively assess relationships between Reco and remotely sensed LST in disturbed and intact northern peatlands. We evaluated the applicability of LST for Reco modeling in compar- ison with in situ measured surface temperature. Overall, we used Reco data from seven Estonian peatlands. Five of these peatlands experienced peat extraction activity and water drainage in the past, while the other two peatlands are natural bog sites. Flux data were measured using closed chambers during the snow-free period in March–November (2017–2020). We studied relationships between in situ measured temperatures and LST data from MODIS Terra, Landsat 7, and Landsat 8 satellites. Finally, we examined the applicability of MODIS LST for Reco modeling and compared the performance of this model with the model using in situ measured surface temperature.
2. Materials and Methods 2.1. Study Area
We collected Reco data in seven boreal peatlands (Figure 1) with various drainage conditions (Table 1) in Es- tonia. In addition to CO2 data, we measured CH4 fluxes (Burdun et al., 2021). The study area has a temperate climate with the long-term (1991–2020) mean annual temperature and precipitation of 7°C and 662 mm, respectively (Estonian Weather Service, 2021). Figure 1 shows the location of study peatlands (upper panel) and zoomed-in orthophotos for each peatland (bottom panels).
Ess-soo bog in southwest Estonia is of limnogenic origin. Its peat layer varies from 4 to 6 m, but in an aban- doned (in 1994) milled peat extraction site, it varies from 2 to 4 m. Vegetation cover in the abandoned milled
Figure 1. The study area includes seven boreal peatlands located in Estonia. The main panel shows a true-colored cloudless mosaic of Landsat 8 obtained in the summer of 2018. The lower small panels show the locations of sites where ecosystem respiration was measured. Orthophotos for the summers of 2019 and 2020 are presented in the lower small panels (Estonian Land Board, 2020).
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
4 of 19
peat extraction area is sparse and is dominated by Eriophorum vaginatum, Calluna vulgaris, Empetrum nigrum, Vaccinium uliginosum, Polytrichum strictum, Betula pubescens, and Pinus sylvestris.
The Kildemaa study site in the northern part of Kõrsa bog comprises an abandoned milled peat production site (remaining peat layer depth 0.8–2 m). A densely drained part of the bog was prepared for peat extraction but abandoned before extraction began (peat deposit up to 3 m). The extracted site is sparsely vegetated with
Sampling site Peatland condition
Number of chambers (ch.) and microtopographic units Dominant species Lat. Long.
Ess-soo
Ess-soo 0 Extracted 4 ch. Eriophorum vaginatum, Calluna vulgaris, Vaccinium uliginosum, Polytrichum strictum, Betula pubescens, and Pinus sylvestris
57.914 26.697
Ess-soo 1 Extracted 3 ch. Eriophorum vaginatum, Calluna vulgaris, Vaccinium uliginosum, Polytrichum strictum, Betula pubescens, and Pinus sylvestris
57.914 26.697
Ess-soo 2 Extracted 3 ch. Oxycoccus palustris, Empetrum nigrum, Vaccinium uliginosum, Polytrichum strictum, Eriophorum vaginatum, and Calluna vulgaris
57.913 26.687
Kildemaa
Kildemaa 1 Extracted 3 ch. Eriophorum vaginatum, Calluna vulgaris, Rhynchospora alba, Betula pubescens, and Pinus sylvestris
58.427 24.786
Kildemaa 2 Drained 3 ch. Calluna vulgaris, Ledum palustre, Polytrichum strictum, Andromeda polifolia, and Pinus sylvestris
58.424 24.784
Kõima
Kõima 1 Drained 3 ch. Various Sphagnum species, Calluna vulgaris, Ledum palustre, Rubus chamaemorus, Andromeda polifolia, and Pinus sylvestris
58.617 24.233
Kõima 2 Natural 3 ch. at lawn Various Sphagnum species, Calluna vulgaris, Andromeda polifolia, and Pinus sylvestris
58.614 24.239
Laiuse
Laiuse 0 Extracted 4 ch. Polytrichum strictum, Eriophorum vaginatum, Calluna vulgaris, Betula pubescens, and Pinus sylvestris
58.790 26.528
Laiuse 1 Extracted 3 ch. Polytrichum strictum, Eriophorum vaginatum, Calluna vulgaris, and Pinus sylvestris
58.790 26.528
Laiuse water Extracted 1 floating ch. 58.789 26.529
Linnussaare
Linnussaare Natural 3 ch. at hollows, 3 ch. at hummocks, 1 floating ch. in
pool
Various Sphagnum species, Ledum palustre, Vaccinium uliginosum, Calluna vulgaris, and Pinus sylvestris
58.878 26.219
Maima
Maima 1 Extracted 3 ch. Eriophorum vaginatum, Calluna vulgaris, Oxycoccus palustris, Vaccinium uliginosum, Betula pubescens, and Pinus sylvestris
58.599 24.379
Maima 2 Extracted 3 ch. Eriophorum vaginatum, Rhynchospora alba, Calluna vulgaris, and Pinus sylvestris
58.596 24.370
Männikjärve
Männikjärve 1 Natural 2 ch. at hollows, 2 ch. at hummocks Various Sphagnum species, Calluna vulgaris, Chamaedaphne calyculata, Rhynchospora alba, Ledum palustre, Oxycoccus microcarpus, and Pinus sylvestris
58.874 26.254
Männikjärve 2 Natural 2 ch. at hollows, 2 ch. at hummocks Various Sphagnum species, Calluna vulgaris, Oxycoccus microcarpus, Carex, and Pinus sylvestris
58.876 26.249
Männikjärve 3 Natural 2 floating ch.in pool, 2 ch. at hummocks
Various Sphagnum species, Calluna vulgaris, Oxycoccus microcarpus, and Pinus sylvestris
58.876 26.247
Table 1 Overview of Peatland Sites
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
5 of 19
E. vaginatum, C. vulgaris, Rhynchospora alba, B. pubescens, and P. sylvestris, while the drained part is dense- ly covered with dwarf pines (P. sylvestris), C. vulgaris, Ledum palustre, lichens, and mosses.
Kõima and Maima peatlands belong to the Lavassaare bog complex, where the peat deposit depth reaches 7.5 m. The Kõima study site covers a former peat extraction site and an adjacent nearly pristine reference site in the northwest part of Kõima bog. Peat was extracted by cutting the peat into blocks with a machine or by hand. The extraction site was abandoned in the 1980s and left for natural recovery. Ditches and de- pressions are mainly recovered with Sphagnum species and drained unexcavated parts are covered with C. vulgaris, L. palustre, Rubus chamaemorus, Andromeda polifolia, and P. sylvestris. In Maima, milled peat extraction took place until the 1990s, when it was abandoned. Currently, the site is only sparsely vegetated with E. vaginatum, C. vulgaris, Oxycoccus palustris, V. uliginosum, B. pubescens, and P. sylvestris.
Laiuse bog is of limnogenic origin and is situated between drumlins. Mining activity ceased there in 1996, and the peatland was left for natural regeneration. The northern part is partly covered with P. strictum, E. vaginatum, C. vulgaris, B. pubescens, and P. sylvestris, while the southern part has been flooded due to beaver activity since 2013.
Linnussaare and Männikjärve bogs belong to the Endla Nature Reserve and are included in the Ramsar List of Wetlands of International Importance (no. 907). These peatlands are of limnogenic origin; their peat layer varies from 4 to 7 m and consists of residuals of Sphagnum, Bryales, Carex, and Pinus (Sillasoo et al., 2007). Vegetation includes dwarf pines (P. sylvestris), grasses and dwarf shrubs (C. vulgaris, E. vagina- tum, Chamaedaphne calyculata, A. polifolia, R. alba, L. palustre, Oxycoccus microcarpus, and O. palustris), and a wide variety of Sphagnum mosses (Sphagnum fuscum, Sphagnum balticum, Sphagnum magellani- cum, and Sphagnum rubellum) (Burdun, Bechtold, Sagris, Komisarenko, et al., 2020).
2.2. Field Measurements of CO2, CH4, WTD, and Soil Temperature
We measured Reco (CO2) with CH4 fluxes with the closed-chamber method (Hutchinson & Livingston, 1993) during the snow-free period (March–November) in 2017–2020. Each chamber was measured multiple times over the growing season, although this varied between sites (see Figure 2 for further details). Chambers (40- cm height, 50-cm diameter, and 65-L volume) were made of white polyvinyl chloride (PVC) to minimize their heating. The chambers were sealed with water-filled PVC collars (20 cm depth) on the peat surface. Each sampling site had replicates (Table 1) and was instrumented with piezometers (perforated pipes with 5-cm diameter and up to 1.5-m length). We sampled gas using pre-evacuated (0.3 mbar) glass vials (50-mL volume) every 20 min during a 1-hr session. Later, the gas concentration in vials was measured using a Shimadzu GC-2014 gas chromatography system equipped with an electron capture detector and a flame ionization detector. WTD was measured in piezometers on the same days that gas samples were collected. Negative numbers for WTD data indicate a water table position below the peat surface, while positive num- bers indicate flooding above the peat surface. In addition to WTD, we measured soil temperature at depths −10 cm (T10), −20 cm (T20), −30 cm (T30), and −40 cm (T40) and at the surface (T0).
2.3. Flux Calculation
Fluxes of CO2 and CH4 were calculated from the linear change in gas concentration in a chamber at 20-min intervals. We adjusted gas concentration by the surface area enclosed by collar and chamber volume. After- ward, we filtered out samples with a determination coefficient (R2) of the linear fit <0.95 (p-value < 0.01) except for the samples with fluxes changes below the gas-chromatographer accuracy (20 ppm for CO2 and 20 ppb for CH4). Additionally, we filtered out CH4 values higher than 30,000 μg C m−2 h−1 interpreted as ebullition fluxes. For the final analyses, we calculated the average CO2 and CH4 fluxes across replicates in each sampling position (Table 1). The flux data were grouped by peatlands' conditions and microtop- ographic characteristics, creating five groups: flooded sites (data from floating chambers in Männikjärve 3, Linnussaare, and Laiuse water), hollows (Männikjärve 1, Männikjärve 2, and Linnussaare), hummocks (Männikjärve 1 to Männikjärve 3, Linnussaare, and Kõima 2), drained sites (Kõima 1 and Kildemaa 2), and extracted sites (Ess-soo 0 to Ess-soo 2, Kildemaa 1, Laiuse 0, Laiuse 1, Maima 1, and Maima 2). Figure 2 shows the time series of CO2 and CH4 fluxes in 2017–2020 for those five groups.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
6 of 19
2.4. Landsat and MODIS LST
We calculated LST from Landsat 7 and Landsat 8 data in the Google Earth Engine (GEE) online platform using open-source code by Ermida et al. (2020). This LST retrieval algorithm utilizes Landsat thermal infra- red and optical data (to derive the normalized difference vegetation index, NDVI), total column water vapor values from NCEP/NCAR reanalysis data, and the ASTER GEDv3 data set to estimate surface emissivity. All these data sets are freely available in GEE (Gorelick et al., 2017).
The field sampling was carried out on the days when Landsat 7 or Landsat 8 passed over the study area. Be- cause of cloudy weather conditions, we had to mask out many LST pixels around the sampling sites. Thus, we decided to calculate the median Landsat LST value over each peatland for each time scene (Figure 3). This decision not only increased data availability for analyses, but it also introduced uncertainty since Land- sat LST values can vary up to 6°C within one peatland (Figure S1 in Supporting Information S1).
MODIS aboard Terra provides MOD11A1 daily LST at a 1-km spatial resolution (Wan et al., 2015). We masked pixels covered with clouds and shadows using the quality control band, which is included in the
Figure 2. Time series of CO2 and CH4 fluxes: one time measured values (marks) and averaged for each day (bars) for flooded sites (a and f), hollows (b and g), hummocks (c and h), drained (d and i), and extracted sites (e and j).
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
7 of 19
MOD11A1 data set in GEE. As with Landsat LST data, we calculated the MODIS LST value as a median across all pixels that cover peatland for each time scene. MODIS LST values accorded well with Landsat LST values (Figure S2 in Supporting Information S1). Nevertheless, the slope of relationships between MODIS LST and Landsat LST varies from 0.778 to 0.887 for different peatlands. This means that under warmer conditions (>15°C), Landsat LST values are higher in comparison with MODIS LST values, and vice versa under cooler conditions: lower Landsat LST values in comparison with MODIS LST values.
In this study, we used remotely sensed MODIS data with a coarse spatial resolution and aggregated Landsat data. The use of these data leads to uncertainty regarding the spatial variation of LST data within the study peatlands. This uncertainty, in turn, raises the problem of representativeness of the LST values for the loca- tion where Reco was measured. Mainly, this problem affects the sites with several types of land management (Kõima and Kildemaa) and sites with a high spatial variation of Landsat LST values (Ess-soo and Maima, Figure S1 in Supporting Information S1). Therefore, we acknowledge the bias arising from our data sources, while noting that this is offset by the long time-series availability and a high temporal resolution that are more critical within the scope of our study.
Figure 3. Time series of MODIS land surface temperature (LST) median (yellow circles), Landsat LST median values (blue circles), and Landsat LST standard deviation within peatland area (blue error bars) in Ess-soo (a), Kildemaa (b), Kõima (c), Laiuse (d), Linnussaare (e), Maima (f) and Männimjärve (g) peatlands.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
8 of 19
The final numbers of Landsat and MODIS images were, respectively, 167 and 420 for Ess-soo, 88 and 387 for Kildemaa, 131 and 387 for Kõima, 78 and 302 for Laiuse, 111 and 441 for Linnussaare, 95 and 379 for Maima, and 98 and 372 for Männikjärve (Figure 3).
2.5. Reco Modeling
We modeled Reco following the approach developed by Lloyd and Taylor (1994) and modified by Tuittila et al. (2004). We utilized a model adjusted by Gaussian curve functions of a second term that account for additional WTD and phenological phase effects (Equation 1) as in previous reports (Järveoja et al., 2016; Riutta et al., 2007):
2 2Pp Pp WTD WTDopt opt1 1 0.5 0.50 Pp WTDtol tolref min min eco ref ,
E T T T TR R e e e
(1)
where Rref (mg CO2 m −2 h−1) is the respiration rate at 10°C, E0 (K) denotes temperature sensitivity, Tref (°C) is
a reference temperature set at 10°C, Tmin (°C) is temperature minimum at which respiration reaches zero set at −46.021°C, T is field-measured surface temperature, Pp (day) denotes the days in a phenological phase that starts in spring when the daily average air temperature is above 5°C (Jaagus & Ahas, 2000), Ppopt (day) denotes the optimal day for maximum Reco from the beginning of vegetation period, Pptol (day) is a vegeta- tion period tolerance for maximum Reco, and WTDopt (cm) is an optimal soil water level for respiration and WTDtol (cm) denotes the soil water-level tolerance (deviation from the optimum at which Reco is 61% of its maximum).
Parameters utilized in Equation 1 were fitted with a Microsoft Excel Solver tool for calculation of the eco- system respiration CO2-response curve (Lobo et al., 2013). First, we derived the parameters utilizing surface temperature (T0) as T in Equation 1 for each sampling site separately and summarized them by groups (intact, drained, and extracted sites) as in a previous study (Turetsky et al., 2014). In doing so, we aimed to explore the intergroup variability of the fitted parameters. Afterward, we derived one set of fitted param- eters using T0 for each of the three groups. In this way, we attempted to test the applicability of our model to estimate the Reco for the specific peatland group, regardless of the spatial variability within each group (Olofsson et al., 2008; Wu et al., 2014; Xiao et al., 2010). After we estimated Reco for specific peatland groups using T0 data, we modeled Reco with MODIS LST data utilizing the same T0-based fitted parameters. In this way, we assessed the interchangeability of in situ temperature and LST for Reco modeling.
2.6. Statistical Analysis
Statistical analysis was performed in R software (R Core Team, 2020). We averaged the collar flux data for replicates at each site (Table 1) for further statistical analysis to avoid pseudoreplication. Furthermore, we applied principal component analysis (PCA) to derive information about the relationships among all variables measured in situ and cluster data, depending on the relevance of different variables for four stud- ied groups, namely hummocks, hollows, and drained and extracted sites. Before PCA, the variables were standardized to zero mean. We did not include flooded sites in PCA since we did not measure the temporal variation of WTD or water column depth in ponds. Temporal changes in the water column depth affect CH4 and CO2 fluxes from ponds (Duchemin et al., 1995; McEnroe et al., 2009); therefore, PCA without this parameter would not be representative.
To estimate the common linear associations in paired repeated measures data, we calculated repeated meas- ures correlation, rmR (Bakdash & Marusich, 2017), between CO2 and CH4 fluxes and in situ measured parameters, and between in situ temperatures, MODIS LST, and Landsat LST. The results of rmR were used to draw conclusions about the variability between data from different sampling sites but included in one group. Similar to the correlation coefficient, rmR varies from −1 to 1. However, rmR does not violate the assumption of independence of observations. Therefore, this method can be used to reveal the associations shared among individual observations in the aggregated data set. We calculated rmR using the rmcorr pack- age (Bakdash & Marusich, 2017) in R software (R Core Team, 2020). Before rmR calculation, distributions of CO2 and CH4 data were normalized with Tukey's Ladder of Powers and log transformations, respectively.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
9 of 19
The normality of transformed data was estimated visually with quantile-quantile plots. We calculated Pear- son correlation (R) for the data originating from one sampling site. Specifically, we provide R values between in situ temperatures and LST in the flooded site (Section 3.2) since only data from Männikjarve 3 sampling site were present. Both rmR and R were estimated with a p-value < 0.05 indicating a statistical significance.
Owing to the higher temporal resolution of the MODIS LST product compared with Landsat LST, we esti- mated rmR between Landsat LST and T0–T40 only for intact sites. For the same reason, we did not perform Reco modeling with Landsat LST data. Instead, MODIS LST was incorporated in the correlation analysis and Reco modeling for all sites. The goodness of Reco model performance was evaluated with R-squared (R2) and root-mean-square error (RMSE) statistics.
3. Results 3.1. Environmental Controls on CO2 and CH4
In Figure 4, the PCA of in situ data shows the separation between different peatland groups. The in situ data projected onto the first two principal components (PCs), which explain 78.4% of the variance in data. PC1 is correlated with temperatures and CO2 fluxes, whereas PC2 is correlated with CH4 fluxes and WTD. The distributions of intact (hummocks and hollows) and disturbed (drained and extracted) sites are well separated by high CH4 fluxes and WTD. At the same time, the distributions of all four groups show a minor separation along PC1.
To compare the relations between CO2 and CH4 fluxes and in situ measured parameters, we performed repeated measures correlation analysis. Figure 5 shows the correlation matrices for the peatland groups. The flooded sites stand out from others because their CO2 fluxes do not have any statistically significant relations with in situ parameters. However, CH4 fluxes are positively associated with water temperature. In
Figure 4. Principal component analysis for in situ measured data for hollows (blue), hummocks (green), drained (yellow), and extracted (red) sites. PC1 and PC2 correspond to the first two principal components (PCs).
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
10 of 19
other groups, CO2 fluxes have weak to strong rmR with temperatures and WTD. In hollows and hummocks, CO2 fluxes have higher rmR values with surface and soil temperatures than with WTD. Both hollows and hummocks show higher rmR with CO2 fluxes for upper soil layers. It is further noteworthy that rmR values between CO2 fluxes and T0–T40 in drained and extracted sites are higher than those in intact sites. The high- est rmR values for CO2 fluxes are observed with T10–T30 in drained sites.
Furthermore, Figure 6 shows the relations between T0 and CO2 and CH4 fluxes for five groups. As previous- ly shown in Figure 5, CO2 fluxes are positively associated with temperature increases. Therefore, the maxi- mum values of median CO2 fluxes are observed in the summer months. In contrast, the lowest median val- ues of CO2 fluxes occur at the beginning of spring (March and April) and the end of autumn (October and November). Also, the weak negative association between CO2 fluxes and WTD is noticeable in Figures 6b– 6e. The positive association between CH4 fluxes and T0 can be seen for hollows and flooded, drained, and extracted sites (Figures 6f, 6g, 6i, and 6j). Similar to CO2, the highest median CH4 fluxes occur in summer.
3.2. LST Versus In Situ Temperatures
The profiles of temperature at different depths together with remotely sensed Landsat and MODIS LST val- ues are shown in Figure 7. We found that median peat temperatures decreased with depth; the highest tem- perature differential occurred between T0 and T10. Drained and extracted sites have high variability in peat temperature with bimodal distribution (Figures 7d and 7e). In contrast, hummocks, hollows, and flooded
Figure 5. Repeated measures correlation (rmR) between CO2 and CH4 fluxes (normalized values), water table depth (WTD), and surface (T0) and soil (T10–T40) temperatures in flooded (a), hollows (b), hummocks (c), drained (d) and extracted (e) sites. Intense red and blue colors indicate strong positive and negative rmR values, respectively. Crossed-out cells correspond to rmR values with a p-value > 0.05.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
11 of 19
sites have lower temperature variability and are close to the normal temperature distribution at almost all depths (Figures 7a–7c). Uneven measurement campaigns over time caused the difference in temperature distributions between drained and intact sites. As it is shown in Figure 2, the field data were collected once per month from March to November in one intact site (Kõima 2), all drained and extracted sites. However, the major amount of data from intact sites was collected in Linnussaare and Männikjärve peatlands once per several weeks from May to September. Therefore, we observe fewer low temperature values for early spring and late autumn in intact sites.
In Figure 7, the LST Landsat values are, on average, higher than MODIS LST values in intact peatlands. Moreover, MODIS LST had consistently higher rmR with in situ temperatures than Landsat LST (Fig- ures 7a–7c). We expected the different performance of MODIS LST and Landsat LST since the slope of their relationships varies from 1.16 to 3.43 (Figure S2 in Supporting Information S1). However, despite the lower spatial resolution, MODIS LST demonstrated superiority over Landsat LST. The mean rmR values between in situ temperatures were 0.38 for Landsat LST and 0.47 for MODIS LST in hummocks and hollows. We further estimated rmR between LST and in situ measured temperatures. For all sites except flooded, both MODIS (Figures 7b–7e) and Landsat LST (Figures 7b and 7c) had the highest rmR with T0. It is noteworthy that rmR between LST and in situ temperatures was higher for disturbed sites than for intact ones.
Figure 6. Scatterplots of surface temperature (T0), CO2, and CH4 fluxes (circles) in flooded (a and f), hollows (b and g), hummocks (c and h), drained (d and i) and extracted (e and j) sites. Monthly fluxes and T0 averages (square shapes with month numbers) are also given with monthly standard deviations (error bars). Colors indicate the water table depth (WTD) except for flooded sites, where no WTD data are available. Inset graphs in panels (h–j) present zoomed-in areas in red rectangles. Negative numbers for WTD data indicate a water table position below the peat surface, while positive numbers indicate flooding above the peat surface.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
12 of 19
3.3. Modeling Reco With In Situ Measured T0 and Remotely Sensed MODIS LST
To estimate the potential of LST to be used instead of in situ measured temperatures in Reco modeling, we compared the performance of Reco models driven by T0 and by MODIS LST data (Figure S3 in Support- ing Information S1). At first, we fitted the parameters from Equation 1 utilizing T0 for each sampling site separately and summarized them according to three groups: intact, drained, and extracted sites (Table 2). By doing this, we explored the intergroup variability of the fitted parameters as well as groups' specificities.
Table 2 highlights that Reco in intact sites is characterized by the highest temperature sensitivity and the lowest flux rate at 10°C. Additionally, we observe the widest vegetation period tolerance for maximum Reco (mean 91.70 days) and the shallowest optimal WTD for intact sites. Reco in disturbed sites has much lower temperature sensitivity and deeper optimal WTD with wider WTD tolerance. It should be noted that the drained sites have the highest respiration rate at 10°C, which agrees with the data shown in Figure 6d.
After we estimated the site-specific parameters, we derived one set of fitted parameters using T0 for each of the three groups (Table 3). It is noticeable that the group-specific parameters differ from the mean values of the parameters in Table 2. First, for the intact sites, E0 value was higher in a group-specific model (Table 3) than the mean one in the site-specific models. Additionally, in the site-specific models, E0 was lower for
Figure 7. Profiles of temperature variation (boxplot) and distribution (shaded area) sensed by Landsat and MODIS, measured at the surface level (T0) and 10–40 cm depths in the peat (T10–T40) for five groups. The median values (black diamonds) for the mentioned temperatures are connected with a dashed line. Blue and orange dots represent Pearson correlation (R) and repeated measures correlation (rmR) between Landsat land surface temperature (LST) and MODIS LST, respectively, and in situ measured temperatures.
Model parameter Intact (hummock, hollow) Drained Extracted
E0 (K) 201.78 ± 23.61 109.05 ± 22.65 132.00 ± 24.38
Rref (mg CO2 m −2 h−1) 56.18 ± 6.95 121.35 ± 13.85 69.10 ± 10.73
Ppopt (day) 104.80 ± 7.11 119.25 ± 3.25 100.00 ± 3.85
Pptol (day) 91.70 ± 13.11 63.45 ± 2.25 71.70 ± 3.60
WTDopt (cm) −18.95 ± 5.18 −30.10 ± 9.30 −34.00 ± 5.82
WTDtol (cm) 30.16 ± 4.35 32.60 ± 9.00 38.70 ± 6.50
Note. Negative values of WTDopt indicate the water table position below the peat surface. WTD, water table depth.
Table 2 Mean and Standard Errors of Estimated Parameters for Ecosystem Respiration (Reco) Models Fitted for Each Sampling Site and Grouped by Peatland Conditions: Intact, Drained, and Extracted
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
13 of 19
hollows (average 169.48 K) than for hummocks (average 234.08 K) and varied from 120.3 K (hollows at Mannikjarve 1) to 300 K (hummocks at Mannikjarve 1).
Second, the group-specific WTDopt values vary from the mean WTDopt shown in Table 2. Nevertheless, they laid within the ranges of the mini- mum and maximum values of the site-specific WTDopt. In intact sites, the site-specific WTDopt values varied from −51.8 cm (Mannikjarve 2 hum- mocks) to −6 cm (Mannikjarve 1 hummocks). A similar range of site-spe- cific WTDopt variation was observed for the drained sites: from −50.9 cm (Ess-soo 1) to −11.2 cm (Maima 2).
Finally, in group-specific models for intact and drained sites, WTDtol re- sulted in higher values than those in Table 2. This difference in WTDtol originated from merging WTD data measured at different microtopo- graphical units. The difference in surface altitude caused by hummock–
hollow microtopography resulted in a high spatial variation of WTD. As a result, the models for intact and drained groups resulted in a wide water-level tolerance. In comparison to intact sites, drained sites have a lower spatial variation of WTD caused by land subsidence.
To show the applicability of LST data for Reco modeling in intact, drained, and extracted peatlands, we com- pared the Reco modeled using (a) T0 data and (b) MODIS LST data. For modeling Reco with T0 and MODIS LST, we utilized the parameters presented in Table 3. We found that Reco values were modeled with higher accuracy for disturbed peatlands (Table 4). As shown in Figure 5, T0 has a strong relationship with CO2 flux- es in disturbed peatlands. Thus, R2 values for the model that utilized T0 were 0.75 for the whole data set and 0.70 for the days when MODIS LST data were available in drained sites. In extracted sites, those values were 0.70 and 0.66, correspondingly. Across the intact sites, R2 values were notably lower at 0.36 and 0.29. When we used MODIS LST instead of T0 in the model, we found a similar pattern: R2 was higher for the disturbed sites (0.67 in extracted sites and 0.66 in drained sites) than for the intact sites (0.27). It is worth noting that relatively high RMSE values were present in all models.
A comparison between measured and modeled CO2 fluxes reveals that we generally fail to catch the var- iability of CO2 in intact sites (Figure 8a). In particular, we observe that the modeling approach cannot be used with either T0 or MODIS LST to model CO2 fluxes higher than 100 mg C m−2 h−1 in the intact sites. Meanwhile, modeled CO2 fluxes better agreed with measured ones in disturbed sites (Figures 8b and 8c). However, some obvious outliers are noticeable for the highest CO2 fluxes for which CO2 fluxes were mod- eled with lower values. We found that those outliers were present in the model output produced with T0 as well as with MODIS LST.
Model parameter Intact (hummock, hollow) Drained Extracted
E0 147.5 114.2 154.7
Rref 50.9 113.8 64.4
Ppopt 99.6 117.6 99
Pptol 105.5 62.5 71
WTDopt −28.7 −25.5 −20.7
WTDtol 99 65.1 43.6
Note. WTD, water table depth.
Table 3 Parameters for Ecosystem Respiration (Reco) Model in Intact (Hummocks and Hollows Merged), Drained and Extracted Peatlands
Model input Model statistics Intact (hummocks, hollows) Drained Extracted
T0 a R2 0.36 0.75 0.70
RMSE (mg CO2 m −2 h−1) 27.38 30.77 21.92
T0 b R2 0.29 0.70 0.66
RMSE (mg CO2 m −2 h−1) 29.23 37.26 24.06
MODIS LST R2 0.27 0.66 0.67
RMSE (mg CO2 m −2 h−1) 29.59 39.27 23.71
Note. RMSE, root-mean-square error. aFor the whole data set. bFor the days when MODIS LST data were available.
Table 4 Performance of Ecosystem Respiration (Reco) Models Driven by Surface Temperature (T0) and MODIS Land Surface Temperature (LST) in Intact (Hummocks and Hollows Merged), Drained, and Extracted Peatlands
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
14 of 19
4. Discussion Prior studies have noted the importance of LST for Reco estimations in different ecosystems. So far, none of the studies have addressed the potential of LST as a proxy for in situ measured temperatures for modeling Reco in disturbed peatlands. Here, we enriched the current knowledge and provided evidence for the future application of LST for that purpose. Even though we utilized daytime MODIS LST data of a 1-km spatial resolution, we still managed to detect the temporal dynamics in temperatures measured in situ at a plot scale (Figure 7). This is particularly important for disturbed sites, where Reco was mainly driven by thermal conditions (Figure 5).
Using the model parameterized for T0, we utilized MODIS LST instead of T0 and obtained R2 equal to 0.27 for modeled Reco in intact sites and 0.66 and 0.67 in drained and extracted sites, respectively. For compar- ison, in a previous study by Junttila et al. (2021) that jointly used remotely sensed LST and EVI data, the average R2 was 0.56 among five peatlands. The lowest R2 was obtained for the bog site (0.23), while R2 was dramatically higher for fen sites and varied from 0.51 to 0.85. We did not have a fen site in our data set; how- ever, the modeling results for bogs are in line with those published by Junttila et al. (2021). Notably, the use of additional remotely sensed data (e.g., vegetation indices) could improve the Reco model performance for intact peatlands. For instance, Schubert et al. (2010) obtained high R2 for both Swedish bog (R2 = 0.89) and fen (R2 = 0.83) by using LST, NDVI, and EVI data from MODIS. Ai et al. (2018) modeled Reco utilizing LST and EVI for a big data set with nine wetland biomes and obtained R2 = 0.59.
Generally, we observed a weak rmR between MODIS LST and Landsat LST and in situ temperatures and between in situ temperatures and CO2 and CH4 fluxes in intact sites. As previously shown for bogs (Burdun et al., 2019), LST has a weak to moderate association with soil temperatures, and the strength of this associ- ation decreases with soil depth. LST dynamics are highly dictated by incident solar radiation, while deeper soil temperatures slowly react with fewer fluctuations (R. Huang et al., 2020). Additionally, we assume that weak rmR between LST and T10–T40 could be partially caused by a higher heat capacity of saturated peat in natural sites with shallow WTD (Zhao & Si, 2019). In previous work, Burdun et al. (2019) demonstrated that MODIS LST had higher rmR with T10–T40 during summers with abnormally high temperatures and correspondingly deeper WTD.
Interestingly, the MODIS LST had consistently higher rmR with in situ temperatures than Landsat LST. The superiority of MODIS LST might arise from a more accurate emissivity estimation in the MODIS prod- uct (Ermida et al., 2020). Nevertheless, both MODIS LST and Landsat LST reveal weaker rmR with T0 in intact sites. We believe this was primarily caused by vegetation cover properties. The studied intact bogs are covered with dense vegetation, primarily Sphagnum mosses, which demonstrate high water loss by
Figure 8. CO2 fluxes measured in situ and modeled with surface temperature—T0 (gray circle)—and remotely sensed MODIS land surface temperature (LST) (orange circle) for intact (hummocks and hollows together), drained, and extracted sites. The dashed line shows a 1:1 relationship.
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
15 of 19
evapotranspiration, which approaches the potential rate of open water evaporation (Kim & Verma, 1996). Through evapotranspiration, mosses cool the surface and perform as a thermal insulation layer (Blok et al., 2011). For these reasons, the disturbed sites with deeper WTD, covered with sporadic vegetation and open peat surface, had higher rmR between LST and T0–T40.
In situ temperatures had strong rmR with CO2 fluxes in disturbed sites but not in intact ones (Figure 5). This weaker rmR in intact sites could be explained by the significant effect of soil moisture on CO2 production (Waddington et al., 2001). Both peat temperature and moisture regulate the biological processes underlying CO2 production (Alm et al., 2007). However, soil moisture has a much larger effect on CO2 production in natural peatlands than in disturbed ones (Waddington et al., 2001). Waddington et al. (2001) observed a peak in CO2 production rate at approximately 92% saturation in the upper peat layer. This means that the disturbed peatlands with deeper WTD may rarely reach this value of saturation. Most of the time, their surface moisture remains far from CO2 production optimum. The highest rmR between CO2 and peat tem- peratures in the intact sites was obtained for T0–T10. In our study, T0–T10 correspond to the uppermost peat soil layer, which has previously been shown to have the largest CO2 production rates (Lafleur et al., 2005).
In the current study, we performed site- and group-specific modeling of Reco. Based on the modeling re- sults, we found a high inner-group variation of the site-specific parameters. This high variation resulted in discrepancies between the mean site-specific parameters shown in Table 2 and group-specific parameters in Table 3. Overall, the modeled group-specific parameters laid within the ranges of the minimum and maximum values of these site-specific parameters. The intact sites had the highest mean E0 in Table 2; how- ever, in the group-specific model, their E0 value decreased and became lower than the E0 modeled for the extracted sites. We hypothesize that differences in E0 and WTDtol values presented in Tables 2 and 3 can be explained by the variation in surface altitude and surface heterogeneity between the sites within one group.
In contrast, we did not observe the high discrepancies for Rref, Ppopt, and Pptol parameters in site- and group-specific models. Rref parameter was found to be the highest for the drained sites in Tables 2 and 3. This high respiration rate can be explained by strong relationships between temperature, heterotrophic, and autotrophic respiration (Pries et al., 2015). The drained sites have a higher heterotrophic respiration than the intact ones due to the deeper WTD (Figure 6) (Jaatinen et al., 2008). Additionally, the drained sites have a higher autotrophic respiration than the extracted ones due to the denser vegetation cover (Järveoja et al., 2016). The combination of both these types of respiration could have resulted in the observed high Rref value for the drained sites. Drained sites had later Ppopt and the narrower Pptol than the intact sites, which can be explained by the difference in vegetation cover. Drained sites were covered with vascular plants that have late start of the photosynthesis period (Korrensalo et al., 2017). In contrast, intact sites were dominant- ly covered with Sphagnum mosses that start their photosynthesis activity early in the spring (Korrensalo et al., 2017).
In this work, Reco models were driven by in situ measurements, among which were WTD time series. How- ever, only a small number of peatlands have in situ historical observations, which limit the future applica- bility of the provided model. Therefore, remotely sensed proxies of WTD must be used, such as radar data (Asmuß et al., 2019; Tampuu et al., 2020) and Optical Trapezoid Model (Burdun, Bechtold, Sagris, Lohila, et al., 2020). Furthermore, given the well-established respiration dependency on LST in disturbed sites, future work could focus on the benefits of combining various remotely sensed data. For example, LAI, NDVI, and EVI were shown to increase the Reco model accuracy over various biomes, including peatlands (Ai et al., 2018; Y. Gao et al., 2015; Junttila et al., 2021). Additionally, vegetation indices have the potential to be utilized as proxies of Ppopt and Pptol parameters, since indices can indicate the ecosystem productivity (Dronova et al., 2021). Moreover, the parameterization of models separately for each peatland could in- crease the model performance (Junttila et al., 2021).
In accordance with previous works (Evans et al., 2021; Feng et al., 2020), in all the sites, rmR between CH4 fluxes and in situ measured parameters was weak and moderate (from −0.5 to 0.5) and periodically not sta- tistically significant (p-value > 0.05). The highest correlation (rmR = 0.53) was observed between CH4 flux- es and T10 in drained sites. Additionally, we observed a positive association (p-value > 0.05) between CH4 fluxes and water temperature in flooded sites. In Figure 2, it is noticeable that CH4 fluxes follow seasonal dynamics in flooded sites (panel f). Summer 2018 was warmer than summer 2019, so CH4 fluxes increased
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
16 of 19
dramatically during 2018 and were the greatest in midsummer. Similar results were discussed in a study by McEnroe et al. (2009), in which a weak positive (p-value < 0.001) R was found between air temperature and CH4 fluxes.
Our findings may be somewhat limited by the small number of sites and methodological constraints. First, we tested LST applicability in only seven sites where Reco data were measured with the closed-chamber technique. A well-known possibility is that chamber measurements of Reco might not accurately represent the fluxes at the landscape scale (Schrier-Uijl et al., 2010). Second, we applied MODIS LST data of a 1-km spatial resolution. The MODIS pixels' footprint covered neighboring territories around the peatlands, which could cause bias in the association between in situ measured Reco and LST. We did not utilize Landsat LST for Reco modeling because of the very limited number of cloud-free images for the disturbed sites. This lack of data occurred even though we calculated one median Landsat LST value over one site for each time scene to increase the amount of Landsat LST data. Unfortunately, high latitudes—where 80% of peatland C stock is located (Tanneberger et al., 2017)—are frequently covered by clouds. In this regard, modeling Reco with high-resolution Landsat data is challenging in northern peatlands. A good alternative to the original Land- sat LST data could be modeled Landsat LST data derived with temporal adaptive reflectance fusion model, such as STARFM (F. Gao et al., 2006). The fusion algorithms for Landsat and MODIS imagery have already shown promising results (Moreno-Martinez et al., 2020). Additionally, machine learning techniques could be used to fill the gaps in Landsat LST images (Buo et al., 2021).
Altogether, our results highlight that remotely sensed LST is a powerful tool for modeling Reco, particularly in disturbed peatlands. LST has the potential to be used in drained and extracted sites with deep WTD and those covered with sparse sedges or bare peat surface. However, more studies are needed to identify how our findings are generalizable across disturbed peatlands in the Northern Hemisphere.
5. Conclusions The purpose of this study was to estimate the strength of the relationships between Reco and LST in dis- turbed (drained and extracted) and intact peatlands. In particular, we aimed to examine the applicability of MODIS LST for Reco modeling and compare the performance of the MODIS LST-driven model with the model driven by the in situ measured surface temperature. This study indicates that LST has a great poten- tial to be utilized in Reco models as a proxy of thermal conditions in northern peatlands. The highest rmR (mean 0.78) was observed between MODIS LST and the in situ measured T0–T40 for drained and extracted sites. However, in intact sites, the relationships between LST and T0–T40 were dramatically weaker: mean rmR over hummocks and hollows was 0.38 for Landsat and 0.49 for MODIS. The Reco model driven by MODIS LST yielded similar accuracy to the model driven by in situ T0: R
2 was 0.29, 0.70, and 0.66, respec- tively, for intact (hummocks and hollows), drained, and extracted sites with the T0-driven model and 0.27, 0.66, and 0.67, respectively, with the MODIS LST-driven model.
The present study represents one of the first attempts to thoroughly examine the potential of remotely sensed LST for monitoring C fluxes of drained and extracted peatlands. Although our study was limited to only seven peatlands with an intermittent Reco time series based on the manual closed-chamber technique, we showed that LST data could be used as a tool to monitor CO2 fluxes with relatively high accuracy. Future research should be carried out to identify how generalizable our findings are across disturbed peatlands in the Northern Hemisphere.
Data Availability Statement Field measured data, reported in this study, are available at Zenodo: https://doi.org/10.5281/zenodo.5118730.
References Acosta, M., Juszczak, R., Chojnicki, B., Pavelka, M., Havránková, K., Lesny, J., et al. (2017). CO2 fluxes from different vegetation commu-
nities on a peatland ecosystem. Wetlands, 37(3), 423–435. https://doi.org/10.1007/s13157-017-0878-4 Ai, J., Jia, G., Epstein, H. E., Wang, H., Zhang, A., & Hu, Y. (2018). MODIS-based estimates of global terrestrial ecosystem respiration.
Journal of Geophysical Research: Biogeosciences, 123(2), 326–352. https://doi.org/10.1002/2017JG004107
Acknowledgments The authors are grateful to Dr Alar Teemusk for gas sample analyses at the laboratory of the Department of Geog- raphy, Institute of Ecology and Earth Sciences, University of Tartu, Estonia. The authors acknowledge the kind sup- port of three anonymous reviewers and the editor; their input in increasing the quality and clarity of the manuscript is invaluable. This work was financially supported by the Estonian Research Council (research grants PRG-352 and MOBERC20), the European Commis- sion through the European Regional Development Fund (the Center of Excellence EcolChange), and the European Commission and ETAG for funding ERA-NET Cofund project Wa- terJPI-JC-2018_13: ReformWater and the Estonian State Forest Management Centre (project LLTOM17250 “Water level restoration in cut-away peatlands: development of integrated monitoring methods and monitoring,” 2017–2023).
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
17 of 19
Alm, J., Shurpali, N. J., Minkkinen, K., Aro, L., Hytönen, J., Laurila, T., et al. (2007). Emission factors and their uncertainty for the exchange of CO2, CH4 and N2O in Finnish managed peatlands. Boreal Environment Research, 12, 191–209.
Asmuß, T., Bechtold, M., & Tiemeyer, B. (2019). On the potential of sentinel-1 for high resolution monitoring of water table dynamics in grasslands on organic soils. Remote Sensing, 11(14), 1659. https://doi.org/10.3390/rs11141659
Bakdash, J. Z., & Marusich, L. R. (2017). Repeated measures correlation. Frontiers in Psychology, 8. https://doi.org/10.3389/fpsyg.2017.00456 Blok, D., Heijmans, M., Schaepman-Strub, G., van Ruijven, J., Parmentier, F. J. W., Maximov, T. C., & Berendse, F. (2011). The cooling ca-
pacity of mosses: Controls on water and energy fluxes in a Siberian Tundra site. Ecosystems, 14(7), 1055–1065. https://doi.org/10.1007/ s10021-011-9463-5
Bubier, J. L., Bhatia, G., Moore, T. R., Roulet, N. T., & Lafleur, P. M. (2003). Spatial and temporal variability in growing-season net ecosystem carbon dioxide exchange at a large peatland in Ontario, Canada. https://doi.org/10.1007/s10021-003-0125-0
Buo, I., Sagris, V., & Jaagus, J. (2021). Gap-filling satellite land surface temperature over heatwave periods with machine learning (pp. 1–5). IEEE Geoscience and Remote Sensing Letters. https://doi.org/10.1109/LGRS.2021.3068069
Burdun, I., Bechtold, M., Sagris, V., Komisarenko, V., De Lannoy, G., & Mander, Ü. (2020). A comparison of three trapezoid models using optical and thermal satellite imagery for water table depth monitoring in Estonian bogs. Remote Sensing, 12(12), 1–24. https://doi. org/10.3390/rs12121980
Burdun, I., Bechtold, M., Sagris, V., Lohila, A., Humphreys, E., Desai, A., et al. (2020). Localizing pixels with SWIR-based moisture index representative of overall peatland water table dynamics. Remote Sensing.
Burdun, I., Kull, A., Maddison, M., Veber, G., Karasov, O., Sagris, V., & Mander, Ü. (2021). CO2 and CH4 gas fluxes in disturbed and intact northern peatlands (Data set). https://doi.org/10.5281/zenodo.5118730
Burdun, I., Sagris, V., & Mander, Ü. (2019). Relationships between field-measured hydrometeorological variables and satellite-based land surface temperature in a hemiboreal raised bog. International Journal of Applied Earth Observation and Geoinformation, 74, 295–301. https://doi.org/10.1016/j.jag.2018.09.019
Change, I. P. on C. (2013). Anthropogenic and natural radiative forcing. In Climate change 2013 the physical science basis: Working group I contribution to the fifth assessment report of the intergovernmental panel on climate change (pp. 659–740). Cambridge University Press. https://doi.org/10.1017/CBO9781107415324.018
Clymo, R. S., Turunen, J., & Tolonen, K. (1998). Carbon accumulation in peatland. Oikos, 81(2), 368. https://doi.org/10.2307/3547057 Crabbe, R. A., Janouš, D., Dařenová, E., & Pavelka, M. (2019). Exploring the potential of LANDSAT-8 for estimation of forest soil CO2
efflux. International Journal of Applied Earth Observation and Geoinformation, 77, 42–52. https://doi.org/10.1016/j.jag.2018.12.007 Davidson, S. J., Strack, M., Bourbonniere, R. A., & Waddington, J. M. (2019). Controls on soil carbon dioxide and methane fluxes from a
peat swamp vary by hydrogeomorphic setting. Ecohydrology, 12(8), e2162. https://doi.org/10.1002/eco.2162 Dronova, I., Taddeo, S., Hemes, K. S., Knox, S. H., Valach, A., Oikawa, P. Y., et al. (2021). Remotely sensed phenological heterogeneity of re-
stored wetlands: Linking vegetation structure and function. Agricultural and Forest Meteorology, 296, 108215. https://doi.org/10.1016/J. AGRFORMET.2020.108215
Duchemin, E., Lucotte, M., Canuel, R., & Chamberland, A. (1995). Production of the greenhouse gases CH4 and CO2 by hydroelectric reservoirs of the boreal region. Global Biogeochemical Cycles, 9(4), 529–540. https://doi.org/10.1029/95GB02202
Ermida, S. L., Soares, P., Mantas, V., Göttsche, F.-M., & Trigo, I. F. (2020). Google Earth Engine open-source code for land surface temper- ature estimation from the Landsat series. Remote Sensing, 12(9), 1471. https://doi.org/10.3390/rs12091471
Estonian Land Board. (2020). Orthophotos. Retrieved from https://geoportaal.maaamet.ee/index.php?page_id=309&lang_id=2 Estonian Weather Service. (2021). Estonian weather service. Retrieved from https://www.ilmateenistus.ee/?lang=en Evans, C. D., Peacock, M., Baird, A. J., Artz, R. R. E., Burden, A., Callaghan, N., et al. (2021). Overriding water table control on managed
peatland greenhouse gas emissions. Nature, 593, 548–552. https://doi.org/10.1038/s41586-021-03523-1 Feng, X., Deventer, M. J., Lonchar, R., Ng, G. H. C., Sebestyen, S. D., Roman, D. T., et al. (2020). Climate sensitivity of peatland meth-
ane emissions mediated by seasonal hydrologic dynamics. Geophysical Research Letters, 47(17), e2020GL088875. https://doi. org/10.1029/2020GL088875
Gao, F., Masek, J., Schwaller, M., & Hall, F. (2006). On the blending of the landsat and MODIS surface reflectance: Predicting daily landsat surface reflectance. IEEE Transactions on Geoscience and Remote Sensing, 44(8), 2207–2218. https://doi.org/10.1109/TGRS.2006.872081
Gao, Y., Yu, G., Li, S., Yan, H., Zhu, X., Wang, Q., et al. (2015). A remote sensing model to estimate ecosystem respiration in Northern China and the Tibetan Plateau. Ecological Modelling, 304, 34–43. https://doi.org/10.1016/j.ecolmodel.2015.03.001
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 202, 18–27. https://doi.org/10.1016/j.rse.2017.06.031
Günther, A., Barthelmes, A., Huth, V., Joosten, H., Jurasinski, G., Koebsch, F., & Couwenberg, J. (2020). Prompt rewetting of drained peatlands reduces climate warming despite methane emissions. Nature Communications, 11(1), 1–5. https://doi.org/10.1038/ s41467-020-15499-z
Hanson, P. J., Griffiths, N. A., Iversen, C. M., Norby, R. J., Sebestyen, S. D., Phillips, J. R., et al. (2020). Rapid net carbon loss from a whole-ecosystem warmed peatland. AGU Advances, 1(3), e2020AV000163. https://doi.org/10.1029/2020AV000163
Helbig, M., Humphreys, E. R., & Todd, A. (2019). Contrasting temperature sensitivity of CO2 exchange in peatlands of the Hudson Bay Lowlands, Canada. Journal of Geophysical Research: Biogeosciences, 124(7), 2126–2143. https://doi.org/10.1029/2019JG005090
Huang, N., Gu, L., Black, T. A., Wang, L., & Niu, Z. (2015). Remote sensing-based estimation of annual soil respiration at two contrasting forest sites. Journal of Geophysical Research: Biogeosciences, 120(11), 2306–2325. https://doi.org/10.1002/2015JG003060
Huang, N., Gu, L., & Niu, Z. (2014). Estimating soil respiration using spatial data products: A case study in a deciduous broadleaf forest in the Midwest USA. Journal of Geophysical Research: Atmospheres, 119(11), 6393–6408. https://doi.org/10.1002/2013JD020515
Huang, R., Huang, J. X., Zhang, C., Ma, H. Y., Zhuo, W., Chen, Y. Y., et al. (2020). Soil temperature estimation at different depths, using remotely-sensed data. Journal of Integrative Agriculture, 19(1), 277–290. https://doi.org/10.1016/S2095-3119(19)62657-2
Huang, X., Silvennoinen, H., Kløve, B., Regina, K., Kandel, T. P., Piayda, A., et al. (2021). Modelling CO2 and CH4 emissions from drained peatlands with grass cultivation by the BASGRA-BGC model. Science of the Total Environment, 765, 144385. https://doi.org/10.1016/j. scitotenv.2020.144385
Hutchinson, G. L., & Livingston, G. P. (1993). Use of chamber systems to measure trace gas fluxes (pp. 63–78). John Wiley & Sons, Ltd. https://doi.org/10.2134/asaspecpub55.c4
Jaagus, J., & Ahas, R. (2000). Space-time variations of climatic seasons and their correlation with the phenological development of nature in Estonia. Climate Research, 15(3), 207–219. https://doi.org/10.3354/cr015207
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
18 of 19
Jaatinen, K., Laiho, R., Vuorenmaa, A., Castillo, U. D., Minkkinen, K., Pennanen, T., et al. (2008). Responses of aerobic microbial commu- nities and soil respiration to water-level drawdown in a northern boreal fen. Environmental Microbiology, 10(2), 339–353. https://doi. org/10.1111/J.1462-2920.2007.01455.X
Jägermeyr, J., Gerten, D., Lucht, W., Hostert, P., Migliavacca, M., & Nemani, R. (2014). A high-resolution approach to estimating ecosys- tem respiration at continental scales using operational satellite data. Global Change Biology, 20(4), 1191–1210. https://doi.org/10.1111/ gcb.12443
Järveoja, J., Nilsson, M. B., Crill, P. M., & Peichl, M. (2020). Bimodal diel pattern in peatland ecosystem respiration rebuts uniform temper- ature response. Nature Communications, 11(1), 1–9. https://doi.org/10.1038/s41467-020-18027-1
Järveoja, J., Peichl, M., Maddison, M., Soosaar, K., Vellak, K., Karofeld, E., et al. (2016). Impact of water table level on annual carbon and greenhouse gas balances of a restored peat extraction area. Biogeosciences, 13(9), 2637–2651. https://doi.org/10.5194/bg-13-2637-2016
Junttila, S., Kelly, J., Kljun, N., Aurela, M., Klemedtsson, L., Lohila, A., et al. (2021). Upscaling northern peatland CO2 fluxes using satellite remote sensing data. Remote Sensing, 13(4), 818. https://doi.org/10.3390/rs13040818
Kim, J., & Verma, S. B. (1996). Surface exchange of water vapour between an open sphagnum fen and the atmosphere. Boundary-Layer Meteorology, 79(3), 243–264. https://doi.org/10.1007/bf00119440
Kimball, J. S., Jones, L. A., Zhang, K., Heinsch, F. A., McDonald, K. C., & Oechel, W. C. (2009). A satellite approach to estimate land-atmos- phere CO2 exchange for boreal and Arctic biomes using MODIS and AMSR-E. IEEE Transactions on Geoscience and Remote Sensing, 47(2), 569–587. https://doi.org/10.1109/TGRS.2008.2003248
Korrensalo, A., Alekseychik, P., Hájek, T., Rinne, J., Vesala, T., Mehtätalo, L., et al. (2017). Species-specific temporal variation in photosyn- thesis as a moderator of peatland carbon sequestration. Biogeosciences, 14(2), 257–269. https://doi.org/10.5194/BG-14-257-2017
Kotta, J., Herkül, K., Jaagus, J., Kaasik, A., Raudsepp, U., Alari, V., et al. (2018). Linking atmospheric, terrestrial and aquatic environments: Regime shifts in the Estonian climate over the past 50 years. PLoS One, 13(12), e0209568. https://doi.org/10.1371/journal.pone.0209568
Lafleur, P. M., Moore, T. R., Roulet, N. T., & Frolking, S. (2005). Ecosystem respiration in a cool temperate bog depends on peat temperature but not water table. Ecosystems, 8(6), 619–629. https://doi.org/10.1007/s10021-003-0131-2
Lafleur, P. M., Roulet, N. T., & Admiral, S. W. (2001). Annual cycle of CO2 exchange at a bog peatland. Journal of Geophysical Research, 106(D3), 3071–3081. https://doi.org/10.1029/2000JD900588
Lees, K. J., Quaife, T., Artz, R. R. E., Khomik, M., & Clark, J. M. (2018). Potential for using remote sensing to estimate carbon fluxes across northern peatlands—A review. Science of the Total Environment, 615, 857–874. https://doi.org/10.1016/j.scitotenv.2017.09.103
Leifeld, J., & Menichetti, L. (2018). The underappreciated potential of peatlands in global climate change mitigation strategies. Nature Communications, 9(1), 1–7. https://doi.org/10.1038/s41467-018-03406-6
Leifeld, J., Wüst-Galley, C., & Page, S. (2019). Intact and managed peatland soils as a source and sink of GHGs from 1850 to 2100. Nature Climate Change, 9(12), 945–947. https://doi.org/10.1038/s41558-019-0615-5
Lloyd, J., & Taylor, J. A. (1994). On the temperature dependence of soil respiration. Functional Ecology, 8(3), 315. https://doi. org/10.2307/2389824
Lobo, F. de A., de Barros, M. P., Dalmagro, H. J., Dalmolin, Â. C., Pereira, W. E., de Souza, É. C., et al. (2013). Fitting net photosynthetic light-response curves with Microsoft Excel—A critical look at the models. Photosynthetica, 51(3), 445–456. https://doi.org/10.1007/ s11099-013-0045-y
Loisel, J., Gallego-Sala, A. V., Amesbury, M. J., Magnan, G., Anshari, G., Beilman, D. W., et al. (2021). Expert assessment of future vulnera- bility of the global peatland carbon sink. Nature Climate Change, 11(1), 70–77. https://doi.org/10.1038/s41558-020-00944-0
Maljanen, M., Sigurdsson, B. D., Guömundsson, J., Öskarsson, H., Huttunen, J. T., & Martikainen, P. J. (2010). Greenhouse gas balances of managed peatlands in the Nordic countries present knowledge and gaps. Biogeosciences, 7(9), 2711–2738. https://doi.org/10.5194/ bg-7-2711-2010
McEnroe, N. A., Roulet, N. T., Moore, T. R., & Garneau, M. (2009). Do pool surface area and depth control CO2 and CH4 fluxes from an ombrotrophic raised bog, James Bay, Canada? Journal of Geophysical Research, 114(G1), G01001. https://doi.org/10.1029/2007JG000639
Moreno-Martinez, A., Izquierdo-Verdiguier, E., Camps-Valls, G., Moneta, M., Munoz-Mari, J., Robinson, N., et al. (2020). Down-scaling MODIS vegetation products with Landsat GAP filled surface reflectance in Google Earth Engine (pp. 2320–2323). International Geoscience and Remote Sensing Symposium (IGARSS). https://doi.org/10.1109/IGARSS39084.2020.9324007
Ojanen, P., Minkkinen, K., & Penttilä, T. (2013). The current greenhouse gas impact of forestry-drained boreal peatlands. Forest Ecology and Management, 289, 201–208. https://doi.org/10.1016/j.foreco.2012.10.008
Olofsson, P., Lagergren, F., Lindroth, A., Lindström, J., Klemedtsson, L., Kutsch, W., & Eklundh, L. (2008). Towards operational remote sensing of forest carbon balance across Northern Europe. Biogeosciences, 5(3), 817–832. https://doi.org/10.5194/bg-5-817-2008
Pan, Y., Birdsey, R. A., Fang, J., Houghton, R., Kauppi, P. E., Kurz, W. A., et al. (2011). A large and persistent carbon sink in the world’s forests. Science, 333(6045), 988–993. https://doi.org/10.1126/science.1201609
Park, H., Takeuchi, W., & Ichii, K. (2020). Satellite-based estimation of carbon dioxide budget in tropical peatland ecosystems. Remote Sensing, 12(2), 250. https://doi.org/10.3390/rs12020250
Pries, C. E. H., van Logtestijn, R. S. P., Schuur, E. A. G., Natali, S. M., Cornelissen, J. H. C., Aerts, R., & Dorrepaal, E. (2015). Decadal warm- ing causes a consistent and persistent shift from heterotrophic to autotrophic respiration in contrasting permafrost ecosystems. Global Change Biology, 21(12), 4508–4519. https://doi.org/10.1111/GCB.13032
Rahman, A. F., Sims, D. A., Cordova, V. D., & El-Masri, B. Z. (2005). Potential of MODIS EVI and surface temperature for directly estimat- ing per-pixel ecosystem C fluxes. Geophysical Research Letters, 32(19). https://doi.org/10.1029/2005GL024127
R Core Team. (2020). R: A language and environment for statistical computing. Retrieved from https://www.r-project.org/ Regan, S., Flynn, R., Gill, L., Naughton, O., & Johnston, P. (2019). Impacts of groundwater drainage on peatland subsidence and its
ecological implications on an Atlantic raised bog. Water Resources Research, 55(7), 6153–6168. https://doi.org/10.1029/2019WR024937 Rinne, J., Tuovinen, J. P., Klemedtsson, L., Aurela, M., Holst, J., Lohila, A., et al. (2020). Effect of the 2018 European drought on methane
and carbon dioxide exchange of northern mire ecosystems: 2018 drought on northern mires. Philosophical Transactions of the Royal Society B: Biological Sciences, 375(1810), 20190517. https://doi.org/10.1098/rstb.2019.0517
Riutta, T., Laine, J., & Tuittila, E. S. (2007). Sensitivity of CO2 exchange of fen ecosystem components to water level variation. Ecosystems, 10(5), 718–733. https://doi.org/10.1007/s10021-007-9046-7
Salm, J.-O., Kimmel, K., Uri, V., & Mander, Ü. (2009). Global warming potential of drained and undrained Peatlands in Estonia: A synthe- sis. Wetlands, 29(4), 1081–1092. https://doi.org/10.1672/08-206.1
Salm, J.-O., Maddison, M., Tammik, S., Soosaar, K., Truu, J., & Mander, Ü. (2012). Emissions of CO2, CH4 and N2O from undisturbed, drained and mined peatlands in Estonia. Hydrobiologia, 692(1), 41–55. https://doi.org/10.1007/s10750-011-0934-7
Journal of Geophysical Research: Biogeosciences
BURDUN ET AL.
10.1029/2021JG006411
19 of 19
Scharlemann, J. P. W., Tanner, E. V. J., Hiederer, R., & Kapos, V. (2014). Global soil carbon: Understanding and managing the largest terres- trial carbon pool. Carbon Management, 5(1), 81–91. https://doi.org/10.4155/cmt.13.77
Schrier-Uijl, A. P., Kroon, P. S., Hensen, A., Leffelaar, P. A., Berendse, F., & Veenendaal, E. M. (2010). Comparison of chamber and eddy covariance-based CO2 and CH4 emission estimates in a heterogeneous grass ecosystem on peat. Agricultural and Forest Meteorology, 150(6), 825–831. https://doi.org/10.1016/j.agrformet.2009.11.007
Schubert, P., Eklundh, L., Lund, M., & Nilsson, M. (2010). Estimating northern peatland CO2 exchange from MODIS time series data. Remote Sensing of Environment, 114(6), 1178–1189. https://doi.org/10.1016/j.rse.2010.01.005
Sillasoo, U., Mauquoy, D., Blundell, A., Charman, D., Blaauw, M., Daniell, J. R. G., et al. (2007). Peat multi-proxy data from Männikjärve bog as indicators of late Holocene climate changes in Estonia. Boreas, 36(1), 20–37. https://doi.org/10.1111/j.1502-3885.2007.tb01177.x
Swindles, G. T., Morris, P. J., Mullan, D. J., Payne, R. J., Roland, T. P., Amesbury, M. J., et al. (2019). Widespread drying of European peat- lands in recent centuries. Nature Geoscience, 12(11), 922–928. https://doi.org/10.1038/s41561-019-0462-z
Tampuu, T., Praks, J., Uiboupin, R., & Kull, A. (2020). Long term interferometric temporal coherence and DInSAR phase in northern peatlands. Remote Sensing, 12(10), 1566. https://doi.org/10.3390/rs12101566
Tang, X., Liu, D., Song, K., Munger, J. W., Zhang, B., & Wang, Z. (2011). A new model of net ecosystem carbon exchange for the de- ciduous-dominated forest by integrating MODIS and flux data. Ecological Engineering, 37(10), 1567–1571. https://doi.org/10.1016/j. ecoleng.2011.03.030
Tanneberger, F., Tegetmeyer, C., Busse, S., Barthelmes, A., Shumka, S., Mariné, A. M., et al. (2017). The peatland map of Europe. Mires & Peat, 19. https://doi.org/10.19189/MaP.2016.OMB.264
Tuittila, E.-S., Vasander, H., & Laine, J. (2004). Sensitivity of C sequestration in reintroduced sphagnum to water-level variation in a cuta- way peatland. Restoration Ecology, 12(4), 483–493. https://doi.org/10.1111/j.1061-2971.2004.00280.x
Turetsky, M. R., Kotowska, A., Bubier, J., Dise, N. B., Crill, P., Hornibrook, E. R. C., et al. (2014). A synthesis of methane emissions from 71 northern, temperate, and subtropical wetlands. Global Change Biology, 20(7), 2183–2197. https://doi.org/10.1111/gcb.12580
Veber, G., Kull, A., Villa, J. A., Maddison, M., Paal, J., Oja, T., et al. (2018). Greenhouse gas emissions in natural and managed peatlands of America: Case studies along a latitudinal gradient. Ecological Engineering, 114, 34–45. https://doi.org/10.1016/j.ecoleng.2017.06.068
Waddington, J. M., Rotenberg, P. A., & Warren, F. J. (2001). Peat CO2 production in a natural and cutover peatland: Implications for resto- ration. Biogeochemistry, 54(2), 115–130. https://doi.org/10.1023/A:1010617207537
Waddington, J. M., & Roulet, N. T. (2000). Carbon balance of a boreal patterned peatland. Global Change Biology, 6(1), 87–97. https://doi. org/10.1046/j.1365-2486.2000.00283.x
Wan, Z., Hook, S., & Hulley, G. (2015). MOD11A1 MODIS/Terra land surface temperature/emissivity daily L3 global 1km SIN grid V006. NASA EOSDIS Land Processes DAAC. https://doi.org/10.5067/MODIS/MOD11A1.006
Wu, C., Gaumont-Guay, D., Andrew Black, T., Jassal, R. S., Xu, S., Chen, J. M., & Gonsamo, A. (2014). Soil respiration mapped by exclusive- ly use of MODIS data for forest landscapes of Saskatchewan, Canada. ISPRS Journal of Photogrammetry and Remote Sensing, 94, 80–90. https://doi.org/10.1016/j.isprsjprs.2014.04.018
Xiao, J., Zhuang, Q., Law, B. E., Chen, J., Baldocchi, D. D., Cook, D. R., et al. (2010). A continuous measure of gross primary production for the conterminous United States derived from MODIS and AmeriFlux data. Remote Sensing of Environment, 114(3), 576–591. https:// doi.org/10.1016/j.rse.2009.10.013
Xu, C., Qu, J. J., Hao, X., Zhu, Z., & Gutenberg, L. (2020). Monitoring soil carbon flux with in-situ measurements and satellite observations in a forested region. Geoderma, 378, 114617. https://doi.org/10.1016/j.geoderma.2020.114617
Xu, J., Morris, P., Liu, J., & Holden, J. (2018). PEATMAP: Refining estimates of global peatland distribution based on a meta-analysis. Catena, 160, 134–140. https://doi.org/10.1016/j.catena.2017.09.010
Yu, Z. (2011). Holocene carbon flux histories of the world’s peatlands. The Holocene, 21(5), 761–774. https://doi.org/10.1177/0959683610 386982
Yu, Z., Loisel, J., Brosseau, D. P., Beilman, D. W., & Hunt, S. J. (2010). Global peatland dynamics since the Last Glacial Maximum. Geophys- ical Research Letters, 37(13). https://doi.org/10.1029/2010GL043584
Zhao, Y., & Si, B. (2019). Thermal properties of sandy and peat soils under unfrozen and frozen conditions. Soil and Tillage Research, 189, 64–72. https://doi.org/10.1016/j.still.2018.12.026
Agronomy 2023, 13, 1183. https://doi.org/10.3390/agronomy13051183 www.mdpi.com/journal/agronomy
Article
Modeling Topsoil Phosphorus—From Observation-Based
Statistical Approach to Land-Use and Soil-Based
High-Resolution Mapping
Anne Kull 1, Tambet Kikas 2, Priit Penu 2 and Ain Kull 3,*
1 Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Kreutzwaldi 1,
51006 Tartu, Estonia 2 Centre of Estonian Rural Research and Knowledge, Teaduse 4, 75501 Saku, Estonia 3 Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise 46, 51003 Tartu, Estonia
* Correspondence: [email protected]
Abstract: Phosphorus (P) is a macronutrient that often limits the productivity and growth of terres-
trial ecosystems, but it is also one of the main causes of eutrophication in aquatic systems at both
local and global levels. P content in soils can vary largely, but usually, only a small fraction is plant-
available or in an organic form for biological utilization because it is bound in incompletely weath-
ered mineral particles or adsorbed on mineral surfaces. Furthermore, in agricultural ecosystems,
plant-available P content in topsoil is mainly controlled by fertilization and land management. To
understand, model, and predict P dynamics at the landscape level, the availability of detailed ob-
servation-based P data is extremely valuable. We used more than 388,000 topsoil plant-available P
samples from the period 2005 to 2021 to study spatial and temporal variability and land-use effect
on soil P. We developed a mapping approach based on existing databases of soil, land-use, and
fragmentary soil P measurements by land-use classes to provide spatially explicit high-resolution
estimates of topsoil P at the national level. The modeled spatially detailed (1:10,000 scale) GIS da-
taset of topsoil P is useful for precision farming to optimize nutrient application and to increase
productivity; it can also be used as input for biogeochemical models and to assess P load in inland
waters and sea.
Keywords: agricultural land; geographical information system; interpolation; land use; machine
learning bagging model; soil phosphorus mapping
1. Introduction
Phosphorus (P) is an important macronutrient for plant growth, and the primary role
of P in plants is to store energy needed for plant growth and reproduction produced by
photosynthesis. Most phosphorus is relatively immobile in soil and even P from phosphate
fertilizers will readily react with soil minerals, making it less available for plants.
Soil phosphorus consists of two forms: organic (non-plant available) and inorganic
(plant available) P. Approximately 29 to 65 percent of total soil phosphorus is in organic
forms, which is not plant available, while the remaining 35 to 71 percent is in inorganic
forms [1] available to plants mainly in the form of phosphate as labile or occluded forms
of P [2]. Organic forms of phosphorus include dead plant/animal residues and soil micro-
organisms. Soil microorganisms play a key role in processing and transforming these or-
ganic forms of phosphorus into plant-available forms, especially in natural ecosystems.
The inorganic form (plant-available form of P) is highly reactive and can be tied up to
chemical compounds (e.g., iron, aluminum, and calcium) in soils as absorbed phosphorus
[3]. The inorganic P has to be dissolved into a solution (P-solution) for plants to be able to
Citation: Kull, A.; Kikas, T.; Penu, P.;
Kull, A. Modeling Topsoil
Phosphorus—From
Observation-Based Statistical
Approach to Land-Use and
Soil-Based High-Resolution
Mapping. Agronomy 2023, 13, 1183.
https://doi.org/10.3390/agron-
omy13051183
Academic Editor: Alberto San
Bautista
Received: 13 February 2023
Revised: 26 March 2023
Accepted: 12 April 2023
Published: 22 April 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license
(https://creativecommons.org/license
s/by/4.0/).
Agronomy 2023, 13, 1183 2 of 24
uptake it. P absorption (incorporation of plant-available P in soil solution into soil clay min-
erals, such as Fe, Al, and Ca oxides) is a fast process, and absorbed P can be released into
soil solution (plant-available P) via desorption. P absorption (retention) is higher in soils that
contain more clay and iron and aluminum oxides [4]. A lower soil pH favors P absorption
by aluminum and iron (pH below 5.5), and a higher pH (over 7.5) enhances plant-available
P fixation by calcium [4,5].
The solution P pool (plant-available P) is very small and ranges from 0.001 mg/L to 1
mg/L [6]. In general, roots absorb phosphorus in the form of orthophosphate but can also
absorb certain forms of organic phosphorus (e.g., glycerophosphate, lecithin, and phytin)
[7]. The active (labile) pool is P in the solid phase that is more readily released into soil
solution, which is the water surrounding soil particles. As plants take up phosphate, the
concentration of phosphate in the solution decreases and some phosphate from the active
P pool is released. An active P pool consists of mainly P absorbed into soil clay minerals
(Fe, Al, and Ca oxides) and easily mineralizing organic P. Labile inorganic P in soils is
predominantly present as specifically adsorbed orthophosphate. This P is in equilibrium
with the soil solution and acts to buffer the soil solution against concentration change [8,9].
Thus, P is desorbed into solution in response to plants’ P uptake or adsorbed from solution
when the P concentration is raised by mineralization or by the addition of a fertilizer ma-
terial.
Soil age is also an important factor influencing P availability, with P becoming in-
creasingly limiting in ancient soils because it gradually disappears through leaching and
erosion [10,11]. In arable lands, P is also partly removed with yield, and additional P fer-
tilizer application is needed to maintain soil fertility. Contrarily to agricultural lands, in
natural ecosystems, P uptaken by plants will mainly remain in place after plants’ death,
and after decomposition and mineralization, it will be available in topsoil for other plants.
Climatic conditions affect P availability as a higher temperature increases the activity of
soil microorganisms and contributes to faster organic matter decomposition and P minerali-
zation. On the other hand, a higher temperature also increases P sorption. In well-aerated soils,
P is released faster than in saturated wet soils. A neutral soil pH (6–7.5) is the best for P avail-
ability, while pH values below 5.5 and between 7.5 and 8.5 limit P availability to plants due to
fixation by aluminum, iron (in case of lower pH), or calcium (in case of higher pH) [5]. Wet
soil conditions decrease soil pH, which increases P sorption to Fe and Al oxides, but flooding
the soil reduces P sorption by increasing the solubility of phosphates that are bound to alumi-
num and iron oxides and amorphous minerals [12]. This aspect has to be considered when
mapping soil P in high latitudes rich in histosols across all land uses and soil types.
Soil phosphorus maps are used to estimate P availability to calculate the need for
fertilization and to model P loss. As only a small amount of P fertilizers used may be read-
ily available for plant uptake due to P sorption and P loss via erosion and runoff, the use
of fertilizers should be based on deeper knowledge about soil properties. Improper P fer-
tilizer use contributes to higher P loss and eutrophication rather than better plant growth.
Detailed and spatially high-resolution soil P measurements covering equally all land
uses and soil types are not usually available for larger areas to produce accurate data-
driven P maps, which are based only on soil sampling data. If high-spatial-resolution P
sampling data are available (e.g., for small areas), geostatistical methods, such as interpo-
lation techniques (e.g., ordinary kriging, co-kriging, or regression kriging) are usually
used to produce soil P maps [13,14]. In the case that soil P sampling grid is not dense
enough, other available environmental variables are also used to predict P content by us-
ing machine learning algorithms, co-kriging, or some other multi-criteria algorithms, in-
cluding multiple regression models. Hybrid geostatistical methods, which incorporate
spatially distributed soil observations and readily available ancillary environmental data
(e.g., topographic variables and satellite data) are recommended instead of univariate
methods, such as ordinary kriging, in cases where natural soil-forming processes are com-
plex or landscapes have high anthropogenic influence [15,16].
Agronomy 2023, 13, 1183 3 of 24
Gaussian process regression (GPR) works principally like co-kriging with a number
of covariates. Ballabio et al. [17] applied GPR to create maps of different LUCAS topsoil
chemical properties (pH, P, N, CaCO3, K, etc.), including indexes calculated from remote
sensing data, meteorological parameters, XY coordinates, and topographic, land-use, and
geological variables.
Recently, machine learning algorithms are increasingly used in predictive mapping
of soil parameters over larger areas as the number of available variables and computing
capacity are increased. Matos-Moreira et al. [18] used a machine learning tool (Cubist) to
develop rule-based predictive models from a calibration dataset. Covariates included pe-
dological, geological, agricultural, terrain and geophysical-related attributes. Rossel and
Bui [19] used a machine learning algorithm (Cubist) to predict total phosphorus in six
different soil layers and ordinary kriging to predict the residuals at each of the standard
depths in Australian soils. To derive the final estimates of total P, the predictions from
Cubist and the kriging estimates of the residuals were summed. The spatial modeling was
performed on 50 bootstraps, and 90 m grid size was used for modeling. Hosseini et al. [20]
used different statistical and machine learning algorithms, such as genetic algorithm (GA),
artificial neural network (ANN), fuzzy inference system (FIS), adaptive neuro-fuzzy in-
ference system (ANFIS), partial least squares (PLS), principal component regression
(PCR), ordinary least squares (OLS), and multiple regression (MR), to determine the best
model to predict soil P in Iran. Four predictive variables (clay, sand, soil organic matter
(SOM), and pH) were used to predict soil P, and the best results were obtained by PLS
(among the regression models) and GA and ANN (among the intelligent models).
Esfandiarpour-Boroujeni et al. [21] used different methods (decision tree (DT), ran-
dom forest (RF), artificial neural network (ANN), and support vector machine (SVM)) for
predicting soil classes (different WRB classification levels) based on environmental varia-
bles (planar and profile curvature, aspect, elevation, slope, catchment area, topographic
wetness index (TWI), LS factor, NDVI, etc.) and expert knowledge from soil scientists
(presence or absence of soil horizons and other soil properties). The SVM algorithm had
the highest overall accuracy for prediction of all qualitative soil properties. The ANN al-
gorithm showed good performance in predicting some quantitative variables (e.g., pure
clay percentage), and the DT algorithm had the lowest uncertainty value.
However, most of these studies that have been conducted to predict soil P with good
results have used detailed measured data for smaller areas [16,18,20–22] or small-scale
maps over larger areas [17,19,23].
The aim of this study was to create a high-resolution (1:10,000) topsoil plant-available
P map for the entire Estonia (area 45,339 km2) covering all land uses and soil types based
on datasets with highly unbalanced data availability and spatial resolution across land-
use categories. The available soil sampling data are usually collected for different pur-
poses and, thus, have different spatial and temporal resolutions, usually covering agricul-
tural areas mainly on mineral soils well while being extremely scarce over natural ecosys-
tems (e.g., forests and wetlands) on histosols and other less fertile soils. Based on the typ-
ical large bias of sampling data along soil types and land-use categories, we hypothesized
that topsoil P content can be sufficiently accurately mapped with geostatistical methods
(e.g., kriging interpolation) in arable lands, but the use of machine learning algorithms
increases prediction accuracy, especially in less intensively managed land-use categories
(e.g., short-term and permanent grasslands), while these established relationships are not
applicable for natural ecosystems (e.g., forests and wetlands), where more robust models
such as the two parameter ordination method should be applied.
Our study objectives were to establish the relationships between observed soil plant-
available P and environmental variables by soil and land-use classes, to determine the
most effective predictive factors related to soil P, establish a cost-effective modeling ap-
proach, assess the accuracy of different modeling methods, and create a high-resolution
topsoil P map at the national level.
Agronomy 2023, 13, 1183 4 of 24
2. Materials and Methods
2.1. Soil Phosphorus Data
Soil phosphorus data available for agricultural soils were obtained from the PANDA
database, which contains regular soil monitoring and voluntary soil sampling data by
farmers. The database contains soil sampling data collected from agricultural lands for
the period 2005–2021 and is managed by the Centre of Estonian Rural Research and
Knowledge. For this period in the PANDA database, there are data about 387,904 compo-
site soil samples all over Estonia, including topsoil phosphorus content (mg/kg). Compo-
site soil samples in the PANDA database were collected by licensed personnel following
the same prescribed sampling strategy (prescribed sampling route and sampling density,
depth, and volume) and analyzed in the same laboratory using the Mehlich-3 method. All
soil samples from arable lands were collected in late summer or autumn at minimum 2–3
months after last fertilization and, in case of winter crops in crop rotation, before the sow-
ing of crops. In the case of application of organic fertilizers, the soil samples were collected
not earlier than 6 months after fertilization. The same field was resampled in each 4–5
years. Up to 2012, the Mehlich-3 colorimetric and, since 2013, the Mehlich-3 inductively
coupled plasma (ICP) analysis method were used to determine soil available phosphorus.
The Mehlich-3 P extraction method is the main method used for estimating plant-available
P in Estonia since 2004. This method is robust, provides the advantage of multielement
analysis [24,25], and, thus, is well-suited for long-term monitoring of Estonian agricultural
soils, where pH is in the range between 5 and 7 for 75.9% of the samples. The PANDA
dataset from 2021 (29,261 soil samples) was not included in model building but used as
an additional independent test dataset to evaluate model performance.
There are no similar comprehensive topsoil phosphorus content databases for other
land-use categories available in Estonia. Comparable values of topsoil phosphorus content
for other land-use types (forest, wetland, peat extraction areas, and quarries) on different
soil types were searched through a literature review of scientific papers [26–44], reports
[45,46], and the Estonian Environmental Monitoring System and supplemented by origi-
nal unpublished datasets of the authors. Therefore, the datasets for these land-use catego-
ries vary by sample size, sampling, and analysis methods. For all samples from agricul-
tural soils, the Mehlich-3 method was applied, while the topsoil P concentration in the
forest and peatland soil samples collated from multiple studies and literature sources
were determined with various methods (Aqua Regia, Olsen, and Kjeldahl), and, thus, con-
version coefficients based on Kulhánek et al. [47] and Wolf and Baker [48] were used prior
to statistical analysis to convert phosphorus content to the same level with the Mehlich-3
method.
2.2. Land Use and Land Cover
Land-use data were combined from the ARIB (Agricultural Registers and Information
Board) 2020 database, where main crop types for agricultural lands are registered. Data on
natural grasslands were taken from the EELIS (Estonian Nature Information System) database
for semi-natural grassland layer, and all other land-use types (wetlands, forests, mining areas,
settlement, waterbodies, etc.) were taken from the ETD (1:10,000, Estonian Topographic Data-
base) (Estonian Land Board, 2020). For topographic information (ground elevation), a 10 m
resolution digital elevation model generated from LiDAR data (Estonian Land Board, 2021)
was used.
Crop data by year were obtained from the ARIB database. In the ARIB database, all
crops that are grown in agricultural massive (complex of neighboring fields) are listed in
alphabetical order. Agricultural massive is an agricultural unit that is in the agricultural
registry and can be applied for EU agriculture support (area-related aid). The ARIB crops
were classified into 8 categories: natural grassland, fallow, cultural grassland, permanent
cultures (e.g., orchards and berries), crop (e.g., rye, wheat, and barley), legumes (e.g., pea,
Agronomy 2023, 13, 1183 5 of 24
bean, and lentil), technical cultures (rapeseed, flax, and hemp), and vegetables (e.g., po-
tato, carrot, and cabbage). The main crop type was selected for each agricultural massive
from the ARIB crop list by prioritizing the crops by their probability of having the largest
share of this agricultural massive (e.g., the area for crop growing is probably larger than
the area for vegetables). From the EELIS database, data on semi-natural grasslands were
obtained and added to the natural grassland category. These areas not covered by the
ARIB data, and EELIS semi-natural grasslands were obtained from the ETD database and
additional categories of wetland, peat mining areas, forest, settlements, and waterbodies
were added. Settlements and waterbodies were omitted from the further analysis as we
did not estimate phosphorus for these areas.
We distinguished 11 land-use classes, with 8 of which being agricultural land catego-
ries (crop types) for which topsoil P values had been obtained from the PANDA database
(permanent grassland, fallow, cultivated grassland, permanent cultures (e.g., orchards and
berries), cereals (e.g., rye, wheat, and barley), legumes (e.g., pea, bean, and lentil), technical
cultures (rapeseed, flax, and hemp), vegetables (e.g., potato, carrot, and cabbage)). Three
non-agricultural land-use types (forest, wetland, and peat extraction area) were included in
the models to allow P value prediction for the entire Estonia (excluding settlement areas,
waterbodies, and mining areas).
2.3. Soil Data
The Estonian Soil Map (1:10,000) in digital form based on extensive field surveys in-
cludes information about soil type, fertility, texture, coarse fragments, and rock content
by layers. Kmoch et al. [49] used the Estonian Soil Map to develop EstSoil-EH: A high-
resolution eco-hydrological modelling parameters dataset for Estonia where soil map in-
formation was made machine-readable and used to predict soil properties (such as clay,
silt and sand content, organic carbon content, and bulk density) by pedotransfer functions
(PTF). In our study, we used the Estonian Soil map data obtained from Kmoch et al. [49].
Kmoch et al. [49] used the SAGA GIS functions (Conrad et al., 2015) to calculate the mean,
median, and standard deviation of several topographic factors (slope, USLE slope length
and steepness factor LS, topographic wetness index (TWI), and topographic ruggedness
index (TRI)) and environmental variables (share of drained area and different land-use
types within the soil polygon) as predictor variables in a random forest model to predict
soil organic carbon content (SOC). The 5 m resolution LiDAR-based DEM provided by the
Estonian Land Board was used for deriving topographic factors. In our study, we modeled
only topsoil phosphorus content; thus, we used only top-layer soil variables (bulk density
(bd1); clay (clay1), silt (silt1), and sand (sand1) contents; soil organic carbon content (soc1);
hydrologic conductivity (k1); and available water content (awc1)) from Kmoch et al. [49].
Quaternary sediment deposit map in a scale of 1:400,000 (Estonian Land Board, 2020)
was used as an indicator of soil parent material. This information was classified into 9
categories (alluvial, glaciolacustrine, glaciofluvial, till, aeolian, wetland sediment, shallow
quaternary sediment, lacustrine, and marine deposits).
2.4. Data Processing
2.4.1. Soil-Type Aggregation
In the Estonian Soil map (1:10,000), there are 119 soil types distinguished according
to the Estonian national soil classification system. The Estonian soil classification and soil
texture types based on the Kachinsky texture system [50] cannot be directly converted to
WRB or FAO classification as the soil types are distinguished based on different principles
[49]. To avoid potential conversion-related issues, we made all our calculations based on
the original texture and soil-type classes.
The original Estonian soil types were in some cases grouped into larger categories
based on genesis and expected similarities in soil properties related to phosphorus content
(soil wetness, gleyic processes, erosion, anthropogenic influences, etc.) to increase soil P
Agronomy 2023, 13, 1183 6 of 24
sample number per soil group. Median phosphorus content by soil types or groups was
calculated (Soil_MedP variable in Supplementary Table S1), and differences between soil
groups were tested by using the Kruskal–Wallis and Dunn tests with Bonferroni multiple
comparison correction in R. There were no statistically significant differences between dif-
ferent Gleysols (Gh, GI, Gk, Go, and Gor). Due to the limited number of samples and the
mixed origin of soils in the anthropogenic soil category (mainly quarries and recultivated
quarries), these soils also had no statistically significant difference from many other soil
categories. Altogether, 26 soil categories were distinguished.
Soil texture was classified into 12 categories based on Estonian soil texture types [50]
as some of these categories could not be directly converted to the USDA texture system. An
approximate reference was used as listed (Estonian texture codes given in brackets): gravel
(kr), gravel limestone (r), sand (s), sandy loam (ls1), loamy sand (ls2, ls3, and sl), fine sand
(pl), clay (s), peat (t), fine sandy loam (tls), and fine loamy sand (tsl). Differences in median
phosphorus content between soil texture categories were calculated (Texture_MedP variable
in Table S1), and differences between categories were tested by using the Kruskal–Wallis
and Dunn tests in R.
2.4.2. Data Analysis and Modeling
Three different approaches were used to model phosphorus content: (1) spatial inter-
polation (kriging), (2) statistical-empirical soil and land-use ordination scalar, and (3) ma-
chine learning bagging (bootstrap aggregation) model.
Spatial interpolation techniques (kriging in our case) create continuous smooth raster
surfaces based on the predictive variable values. Soil and land-use ordination and bagging
methods need some data units (polygons) for which all predictive variables are available.
Therefore, for model building, only variables that were available for all study area (entire
Estonia) could be used. For topsoil P content prediction, we used the unique units created
by intersecting soil and land-use polygons.
The ArcGIS Pro (2.9) tools were used to prepare units (so-called soil–land use poly-
gons) for P prediction. Soil map layer enriched with the predicted soil parameters (EstSoil-
EH) was intersected with land-use/land-cover layer (composed from the ARIB, EELIS, and
ETD data as described above) to create a layer that contains information about both soil
and land-use parameters. Settlements and waterbodies were excluded from the prediction
dataset. The output layer contained over 3 million polygons. Sediment (parent material)
information was added to the soil–land use polygons by spatial join function with the
largest share option. The soil–land use polygons central point X and Y coordinates were
calculated, and the central point elevation was extracted from a 10 m resolution LiDAR
DEM to represent longitude, latitude, and elevation variables in the bagging model.
Kriging
The most widespread approach to create a continuous surface from spatially well-
distributed point measurements is interpolation [51–53]. Kriging is among the most pop-
ular spatial interpolation techniques because it can give the best linear unbiased predic-
tion if suitable parameters are selected. This method is widely used in regional hydrolog-
ical and spatial nutrient runoff models, but also for mapping spatial distribution of nutri-
ents in detailed field-level test areas.
In our study, topsoil phosphorus content for agricultural areas all over the Estonia
was interpolated based on the measured phosphorus values of the PANDA database by
using simple kriging and stable variogram method in the ArcGIS Pro Geostatistical Ana-
lyst extension. Normal score transformation was used to transform the measured phos-
phorus data to follow a univariate standard normal distribution. Kriging is using Gauss-
ian process to estimate the mean value of a dependent variable and thus normal distribu-
tion of data are very important. Kriging assumes normal distribution and stationarity of
data (close points should have quite similar values, with low variance nearby). Spatial
Agronomy 2023, 13, 1183 7 of 24
trend was investigated based on the scatter plots of P values against the X and Y coordi-
nates, but, as no trend was detected, trend removal was not applied in the kriging models.
Two kriging models were built in order to compare the kriging results with other models:
the Kriging1 method, which used all available sample points to build a model to get the
best results, and the Kriging2 method, where the original phosphorus dataset was divided
into training (75%) and test (25%) datasets by using the ArcGIS Geostatistical Analyst ex-
tension tool Subset Features, which created a random subset from the dataset. The
Kriging2 model served only a model validation purpose.
Soil–Land Use Ordination
As the PANDA database covers only agricultural lands, kriging interpolation predic-
tions outside agricultural areas are not valid as the topsoil phosphorus content is largely
depending on anthropogenic factors, such as fertilizer application, pollution, and crop
harvesting, which are different in agricultural lands and cannot be interpolated or extrap-
olated to other land-use categories without taking into consideration other factors.
Ordination can be considered a synonym for multivariate gradient analysis. Ordina-
tion methods are mainly used in community ecology, where multiple spatial axes are used
to arrange or order multiple variables (of species and/or sample units) along gradients.
The most famous ordination methods are principal component analysis (PCA), factor
analysis (FA), detrended correspondence analysis (DCA), and other similar multidimen-
sional statistical analysis methods [54,55].
According to the exploratory data analysis, soil and land-use types were the main
independent factors influencing phosphorus content in topsoil. The median values of P
content of different soil types, land-use types, and their combinations were calculated to
create a land-use/soil-type matrix. Soil and land-use types were ordered ascendingly
based on the median topsoil phosphorus values. Some soil-type/land-use combinations
had no or limited number of measured soil P values. To interpolate topsoil phosphorus
values for these soil–land use combination without sufficient sampling data, the so-called
soil–land use ordination surface was created. Soil type was used as the X axis and land-
use type was used as the Y axis to generate Cartesian coordinates. The X and Y coordinates
representing soil and land-use types and the corresponding class’s median topsoil phos-
phorus content for groups with sufficient measured topsoil P data were used to interpo-
late trend surface in Surfer (ver. 24) by using radial basis multiquadratic (R2 parameter
0.2) gridding. The created ordination model surface represents the P value for different
soil and land-use combinations. This model was used to assign median phosphorus values
for all soil–land use polygons in the entire Estonia according to particular combination of
soil type and land-use class. The soil–land use ordination model results were validated by
calculating the linear relationship between the model results predicted for the soil–land
use polygons all over Estonia and all phosphorus sampling points in the PANDA data-
base. Additionally, the soil–land use polygons’ median P values calculated from the meas-
ured P data from the PANDA database were used for validation. As the P values in the
PANDA database are not normally distributed, a nonparametric correlation analysis was
implemented, and Kendall tau correlation metrics between the measured and predicted
data were calculated.
Bagging
To predict the topsoil phosphorus content for soil–land use/land cover polygons all
over Estonia based on multiple variables, the prediction model was built by using bag-
ging-based machine learning method (bootstrap aggregation and classification and re-
gression trees (CART) method) [56]. Bagging works best with high-variance unstable
models, and it is usually applied to decision tree methods, where a forest of many trees is
created to average the model and predict the best outcome, to add stability and accuracy,
and to reduce overfitting and variance. Bagging decreases the variance in prediction in
high-variance models without increasing the bias. It creates new bootstrap training sets
Agronomy 2023, 13, 1183 8 of 24
by sampling with replacement and then fits a model to each new training set. These mod-
els are combined by averaging the predictions for the regression case. Other machine
learning methods, such as random forest and extreme gradient boosting methods, were
tried, but they gave very similar results; thus, bagging was chosen because the bagging
prediction was faster in case of large dataset (over 3 million soil–land use polygons). Mi-
crosoft R and R Studio (v4.0.2) were used to build machine learning models and for ex-
ploratory data analysis. For bagging, the tidymodels and baguette libraries were used. As
the dependent variable (topsoil phosphorus content (mg/kg)) was a continuous variable,
a regression mode and rpart (recursive partitioning and regression tree) model [57] was
used for bagging. The dataset was randomly divided into training (75%, 268,882 samples)
and test data (25%, 89,625 samples), and high and low phosphorus values were selected
proportionally. A total of 25 bootstrap samples were taken from the training dataset to fit
the regression model. Additionally, 50 and 100 bags (bootstrap samples) were also tried,
but the model improvement was very limited compared to the increase in computing time.
For bagging, individual trees are grown deep and not pruned. These trees have high var-
iance and low bias. Overfitting and high variance are solved by averaging the model over
the number of trees. In this study, bagging fitted 25 regression models and averaged the
results to reduce the variance. A total of 32 variables (Supplementary Table S1) were used
to build the bagging model to predict topsoil P content. The categorical variables (soil
type, land-use/land-cover type, sediments (parent material), soil texture, and soil hydro-
logical group) were replaced with the median phosphorus content value of each category
(Supplementary Table S1).
2.4.3. Model Evaluation
To compare the results from different models, similar evaluation methods and metrics
were used when applicable. The main metrics we used to evaluate the model performance
were coefficient of determination (R2), root mean squared error (RMSE), and mean absolute
error (MAE). R2 is the metric used to describe the performance of any models (including
cross-validation and test set validation). R2 explains to what extent the variance of the pre-
dicted P explains the variance of the measured P (or polygon median P). In our case, R2 also
evaluates the linear regression between the measured and predicted values. RMSE is the
standard deviation of the residuals (prediction errors) in original units. As the effect of each
error (deviation from regression line) is proportional to the size of the squared error, RMSE
is sensitive to outliers. MAE is the average of the absolute values of the errors in original
units.
In the kriging method, prediction errors are evaluated using cross-validation by leav-
ing one sample point out and using all other data points to predict the value for this omit-
ted point, but cross-validation method uses all data points to estimate the trend and auto-
correlation. As the cross-validation differs from validation applied for evaluation of other
models, our original phosphorus dataset was divided into training (75%) and test (25%)
datasets by using the ArcGIS Geostatistical Analyst extension tool Subset Features, which
created a random subset from the dataset. Both training data cross-validation and test data
validation against the measured P values were used to evaluate the kriging model's per-
formance. As the ArcGIS Geostatistical extension kriging evaluation statistics does not in-
volve any other metrics, the cross-validation dataset was exported to Excel, and the eval-
uation metrics (R2, RMSE, and MAE) were calculated separately.
The bagging model was built based on 75% of data (268,882 samples) being used as
the training set. The other 25% of data (89,625 samples) were used for model validation.
To test the bagging model performance, R2, RMSE, and MAE were calculated.
As bagging uses random subsampling with replacement, each bagged predictor was
trained on approximately 63% of training data, and the remaining 37% of data were called
out-of-bag (OOB) observations. OOB error estimates require less computation than cross-
validation and can be used to evaluate model performance in the training process. An-
other benefit of OOB is that it does not require a separate test dataset. Cross-validation
Agronomy 2023, 13, 1183 9 of 24
(CV) and 10-fold random sampling were used to evaluate the model performance on the
training data, and the evaluation metrics R2, RMSE, and MAE, were calculated.
Besides using traditional (validation on test data) and model-specific (OOB error and
cross-validation) validation methods, all model results were also compared with the meas-
ured composite sample P values and the soil–land use polygon median P values. The
model prediction results were extracted at the measured P sample locations (358,947 sam-
ples), and evaluation metrics (MAE, RMSE, and R2) to compare the measured vs. predicted
values were calculated. The topsoil P samples collected in 2021 (29,261 samples) served as
an additional independent validation dataset not used in any other analysis.
The topsoil P samples were also compared to the soil–land use polygon median P
values as the ordination and bagging method predicted values to soil–land use polygons.
Likewise, the topsoil P samples collected in 2021 were also compared to the soil–land use
polygon median P values, which were calculated based on the topsoil P samples taken in
the period 2005–2020.
3. Results
3.1. Long-Term Changes of Topsoil Phosphorus in Agricultural Soils
Plant-available P is relatively stable in natural ecosystems but strongly influenced by
management in agricultural areas. To assess trends in topsoil P content, we used long-
term monitoring data from the PANDA database for agricultural land in the period 2005–
2021.
The lowest number of P samples was collected in 2012 (5243 samples), while 24,857
P samples were collected in 2015. The median P value was lowest in 2005 (47 mg/kg) and
highest in 2016 (82 mg/kg). The P sample locations, sample size, and median P content by
years are presented in the supplementary materials (Figure S1). As the exploratory analy-
sis detected no specific temporal trend for P values collected in different years (Figure 1),
the temporal component was left out from further topsoil P modeling. The P values vary
interannually mostly because different locations were sampled in different years and
resampling usually occurred in four years. Moreover, the topsoil P value is affected by
fertilizer application and crop rotation. Extremely high P values might be measured in
case of recent fertilizer application nearby and were treated as outliers.
Figure 1. Measured topsoil P values (mg/kg) for agricultural lands in 2005–2021 and corresponding
median P values for main land-use classes.
Agronomy 2023, 13, 1183 10 of 24
3.2. Dependence of Topsoil P on Soil Parameters in Agricultural Lands
To understand the dynamics and potential loss of P at the landscape level, it is im-
portant to understand how soil parameters are related to P content. We used exploratory
statistical analysis to check the measured topsoil P content relations with other variables
related to soil, land use, and topography available for the entire Estonia (Figure 2). The
correlation analysis shows that in agricultural ecosystems, plant-available P content in
topsoil is mainly controlled by fertilization and land management. The best explanatory
variables in the single-parameter linear regression to predict P values are soil type (R2 =
0.15) and other soil-related categorical variables, including texture (R2 = 0.08) and soil hy-
dro group (R2 = 0.09). Land-use category contributes less to P prediction in the single-
parameter model (R2 = 0.02). According to the linear regression model built by using soil
and land-use categories, both soil and land-use categories are statistically significant fac-
tors describing the measured P values (R2 = 0.16). The exploratory data analysis shows
that higher median P values are associated with automorphic soils (original soil classes
shown, approximate WRB reference is provided in [49], mainly with K (Kh 116 mg/kg, KI
93 mg/kg, Ko 81 mg/kg, and Kr 75 mg/kg) and L (Lk 126 mg/kg, LP 101 mg/kg, and LkG
77 mg/kg) soils and also D soils (69 mg/kg). Lower P values are common for hydromorphic
soils (AM 16, M 18, T 18, AG 24, G1 27, and Gk 29 mg/kg) (see also Figure 3).
Figure 2. Kendall correlation between topsoil P values (mg/kg) for agricultural lands in 2005–2020
and variables used in the bagging model.
Agronomy 2023, 13, 1183 11 of 24
Quaternary sediments as an indicator of parent material are directly related to soil
properties, and different sediments contribute differently to topsoil P values. The highest
median P values are associated with glaciofluvial deposits (MedP = 90 mg/kg). Shallow
quaternary deposits in the case of Rendzic soils (MedP = 76), aeolian deposits (MedP = 74),
and till (MedP = 74) have medium median P values. The lowest median P values belong
to wetland sediments (peat) where MedP = 35, lacustrine deposits (MedP = 44), glaciola-
custrine deposits (MedP = 51), and marine and alluvial deposits (MedP = 59).
3.3. Topsoil Phosphorous Modeling
3.3.1. Soil–Land Use Ordination Model
Soil–land use/land cover ordination is principally a categorical model that assigns the
same P values for a particular soil–land use category combination. It can distinguish values
for types in a matrix predefined by soil categories (29) multiplied by land-use categories
(11), which leads to a total of 298 soil–land use combinations. Each unique soil–land use
ordination combination corresponds to single predicted P value. The advantage of this
method is that if the soil and land use ordination is known, then further interpolation can
also provide topsoil P values for categories where observed values are missing or limited in
number, or estimate continuous values for transition from one soil/land use category to an-
other. An ordination model was created to predict the median topsoil phosphorus values
for pre-defined soil–land use combinations. The model predicts topsoil P values for pre-
defined soil–land use combinations based on the trend surface interpolated by using all
available P values (including data from the literature), and the soil and land-use categories
are ordinated in an ascending order (Figure 3).
Figure 3. Data gap filling by radial basis multiquadratic spatial interpolation along soil and crop
ordination axes to model topsoil P values for any soil-type and land-use combinations.
The results show that soil and land-use types have influence on topsoil phosphorus
content, but these factors do not describe the majority of variance, which depends on
many other factors. This kind of simple two-parameter ordination and classification
model does not cover enough variance in topsoil P content to use the predicted P values
for, e.g., P loss modeling. To improve the ordination model, the ordination values were
replaced with the soil–land use polygon median phosphorus values (Figure 4) calculated
Agronomy 2023, 13, 1183 12 of 24
based on the PANDA database (2005–2020) in those polygons where phosphorus sam-
pling points existed (there were 108,128 polygons out of 3,063,558 where at least one sam-
pling point was located). In this way, the model quality was improved in agricultural ar-
eas. Model validation against the measured P sample values (2005–2020) gave the follow-
ing results: R2 = 0.74, MAE = 18.8, and RMSE = 36.4, showing the relationship between
soil–land use polygon median topsoil phosphorus content and measured phosphorus
sampling points. The results suggest that there is quite high variance in topsoil phospho-
rus content even within the same agricultural parcel with an identical soil type.
Figure 4. Modeled topsoil median plant-available P content (mg/kg) based on soil–land use ordina-
tion + polygon median P model.
3.3.2. Kriging Model
Simple kriging was used to interpolate topsoil P content in Estonia (Figure 5) based
on the P sampling points (PANDA database) collected in 2005–2020. Two kriging models
were built in order to validate the kriging results by different methods. The first kriging
model (Kriging1) was built traditionally by using all sample points and validated by cross-
validation (by leaving-one-out method). The Kriging1 model cross-validation evaluation
gave the following results: R2 = 0.63, MAE = 27.1, and RMSE = 43.1.
Agronomy 2023, 13, 1183 13 of 24
Figure 5. Modeled topsoil median plant-available P content (mg/kg) based on kriging model. Non-
agricultural land uses are excluded from kriging to avoid extrapolation.
To test the kriging model performance on the test data, Kriging2 model was built by
using only 75% of randomly selected P sample data. The remaining 25% of P samples
(89,661 samples) were used for model validation, which gave the following results: R2 =
0.62, MAE = 27.7, and RMSE = 44.4.
The kriging models represent the topsoil P values well in fields which are uniform in
soil cover and are excessively sampled, but they fail in areas where soil cover is mosaic,
crop rotation has been different over the years, or the distance between existing sampling
points is greater than the average soil mapping unit diameter.
3.3.3. Bagging Model
The bagging model was introduced in order to overcome the disadvantages of the
kriging and ordination models, namely a low share of P variation explained by the two-
parameter soil–land use ordination model and the unreliability of the kriging model re-
sults outside of agricultural areas and far from the P sampling points.
Ensemble machine learning models can give much more accurate predictions than
simple models, but the result is like a Black Box Model which underlying functional form
is so complex that it cannot be written down as a simple formula. We used 32 variables
(Supplementary Table S1) to build the bagging model. Ensemble models, such as bagging
models, that average the results of multiple models are not easy to interpret as each vari-
able may appear in different trees in different positions, if at all. However, bagging gives
the aggregated variable importance scores (Figure 6), which can be used to understand
which variables are more important in predicting the topsoil P values. According to the
bagging model that we used to predict topsoil P values (Figure 7), the two most important
variables were the Y coordinate (latitude) and X coordinate (longitude). The next im-
portant variables were soil-type median phosphorus content (Soil_MedP) and the soil-
type median phosphorus content if the land use was natural grassland (NatSoil_MedP).
The absolute elevation was in the fifth rank of importance. The most important topo-
graphic variables calculated from the digital elevation model were the standard deviation
Agronomy 2023, 13, 1183 14 of 24
of slope in soil polygon (slp_stdev), mean terrain ruggedness index (tri_mean), and stand-
ard deviation of LS factor (ls_stdev). Soil organic carbon (soc1) was also important as ex-
pected, and so was soil polygon size (unit_area).
Figure 6. Standardized variable importance score of factors in the bagging model.
In many cases, the importance of X and Y coordinates is complicated to interpret, but in
our model, they represent a known set of natural phenomena following these directions. The
north–south direction represents the gradient in bedrock (from limestone to sandstone) and
the presence of naturally occurring phosphorite brought up by glacial activity in the northern-
most part of Estonia [58]. The west–east direction coincides with the gradients of maritime to
continental climate, lower to higher topography, and soil formation age after the retreat of the
last continental ice sheet.
Figure 7. Modeled topsoil median plant-available P content (mg/kg) based on the bagging model.
Agronomy 2023, 13, 1183 15 of 24
3.3.4. Hybrid Map of Topsoil Phosphorus
All models created based on different methods have limitations inherited from data
availability or bias by land-use or soil-type classes. Agricultural land is spatially well cov-
ered with sampling data, but only limited soil types and few sampling points are available
in natural ecosystems (e.g., forests and wetlands) or human-altered environment (e.g., quar-
ries), thus severely hindering use of data-driven methods such as machine learning or
kriging. Moreover, the relationship between the explanatory variables and predicted varia-
ble established in a managed environment, such as agricultural land, could not be directly
applied in natural ecosystems. On the other hand, simple land-use/soil-type relationship
with topsoil P in a relatively uniform natural environment does not describe the spatial and
temporal variability of managed ecosystems where the same soil polygon might be divided
by different land uses or experience various fertilization practices or crop rotation affecting
the soil phosphorus status.
To overcome the problems related to different methods, including (a) interpolation
performs well only in densely sampled agricultural lands, (b) the bagging model covers
with sufficient accuracy most land-use categories with meaningful sampling data by land-
use and soil classes, but topsoil P predictions overshoot in natural ecosystems with limited
data availability, and (c) the ordination model’s substantive problem that single variables,
such as soil and land use/land cover, explain only a minor share of variance related to
topsoil P content in managed and fertilized land-use classes, we created a hybrid map
(Figure 8).
Figure 8. Hybrid model of topsoil median plant-available P content (mg/kg) based on the bagging
model and soil–land use polygon median P values if available for agricultural land, and the ordina-
tion model for natural vegetation-covered areas.
The hybrid map was created by combining arable land polygons with the median
value of existing sampling data, the bagging-based predicted values for other agricultural
lands, and the two-parameter soil–land use ordination matrix-based model values for for-
ests, wetlands, and peat extraction areas. The topsoil P values for each soil–land use unit
of eight agricultural land-use categories (permanent grassland, fallow, cultivated grass-
land, permanent cultures, cereals, legumes, technical cultures, and vegetables) were based
Agronomy 2023, 13, 1183 16 of 24
on the observed polygon median topsoil P content value if a particular soil–land use unit
had sufficient number of sampling data to calculate the median value. The bagging model
predicted topsoil P values were used in agricultural land where sampling data were lim-
ited or missing. The remaining three non-agricultural land-use types (forest, wetland, and
peat extraction area) were based on the ordination model results.
The created high-resolution topsoil phosphorus map reveals clear a latitudinal–longitu-
dinal and elevation-related spatial pattern. Higher predicted P values are concentrated in the
northern and western coastal areas. In the northern coast region (Maardu, Toolse, and Rak-
vere), there are phosphorite deposits due to glacial deposits that naturally contribute to the
higher P values in the topsoil of this region. In West Estonian islands and in uplands, the P
content in the topsoil is higher than average, partly due to long-term management and fertili-
zation. In Estonian northern and western coastal areas, thin and calcareous juvenile soils con-
tribute to the higher P content as soil age is one of the factors contributing to the P content.
Low P values are associated with large wetland areas and older soils with lighter texture,
which contain less P due to leaching and erosion.
3.3.5. Validation of Topsoil P Prediction Models
All models were validated against the measured topsoil P values, the median of the
measured P values of the soil–land use unit area, as well against each other.
• Ordination validation
Due to the data-driven approach of the ordination model, classical model validation
methods (test dataset and cross-validation) were not available. All topsoil P sample points
were used to evaluate the ordination model performance (R2 = 0.15, MAE = 42.9, and RMSE
= 66.8) (Table 1, Ordination vs. P).
Table 1. Ordination model validation results.
Model MAE RMSE R2 Explanation
Ordination vs. P 42.9 66.8 0.15 Soil–land use ordination model results validated against
measured P samples (all P samples used for validation).
Ordination vs.
poly MedP 36.7 56.5 0.2
Soil–land use ordination model results validated against
polygon median P values (all P samples used for validation).
Ordination vs. P
2021 46.2 73.4 0.17
Soil–land use ordination model results validated against
P data collected in 2021 and not used for model building.
Ordination +
polygon MedP
vs. P 2021
34.5 57.4 0.48
Soil–land use ordination model values replaced with
soil–land use polygon median P values in polygons
where P sample points measured in 2005–2020 exist, and
evaluated against P samples collected in 2021.
The model results were validated with an independent dataset (P samples collected in
2021), and the results were as follows: R2 = 0.17, MAE = 46.2, and RMSE = 73.4 (in Table 1
Ordination vs. P 2021). The ordination model results were also evaluated against the soil–land
use polygon median P values as the ordination model predicts values based on the soil–land
use types. The ordination model evaluation against the polygon median P values gave slightly
better results (R2 = 0.20, MAE = 36.7, and RMSE = 56.5) (Table 1, Ordination vs. poly MedP)
compared to the evaluation against the sampled P values. The ordination + polygon median
P model results were validated with an independent dataset (P samples collected in 2021), and
the results were as follows: R2 = 0.48, MAE = 34.5, and RMSE = 57.4 (Table 1: Ordination +
polygon MedP vs. P 2021).
• Kriging validation
The Kriging1 model cross-validation evaluation gave the following results: R2 = 0.63,
MAE = 27.1, and RMSE = 43.1 (Table 2, Kriging1 CV vs. P). The R2 between the Kriging1 pre-
dicted values and all measured phosphorus values is 0.82, MAE = 19.8, and RMSE = 31.3 (Table
2, Kriging1 vs. P), but this comparison is not valid as the validation considering the kriging1
Agronomy 2023, 13, 1183 17 of 24
model was built by using the same P sample points, and the kriging interpolator is the exact
interpolator; thus, the result rather suggests that variation in the topsoil P values is quite high
even in a very short distance.
Table 2. Validation of kriging model results.
Model MAE RMSE R2 Explanation
Kriging1 CV vs.
P 27.1 43.1 0.63
Kriging1 (all sample points used for interpolation) cross-
validation (by leaving-one-out method).
Kriging1 vs. P 19.8 31.3 0.82
Kriging1 (all sample points used for interpolation) valida-
tion against all measured P samples (same points that were
used for interpolation).
Kriging1 vs.
poly MedP 15.6 25.1 0.84
Kriging1 (all sample points used for interpolation) validation
against polygon median P values (if there was only one P sam-
ple in the polygon, then the same values were used for inter-
polation).
Kriging1 vs. P
2021 30.7 49.6 0.59
Kriging1 (all sample points used for interpolation) validation
against P samples (collected in 2021) not used for model build-
ing.
Kriging2_test
vs. P 27.7 44.4 0.62
Kriging2 model (surface was interpolated by using only
75% of P sample data) validated against test data (25% of
data). Only 25% of data were used to calculate these met-
rics.
Kriging2_test
vs. poly MedP 19.7 31.6 0.75
Kriging2 model (surface was interpolated by using only
75% of P sample data) validated against test data (25% of
data) polygon median P values. Only 25% of data were
used to calculate these metrics.
The Kriging1 model results were also compared with the soil–land use polygon median
P values (kriging model value in the measured P sample location was compared with the
median P value of soil–land use polygon; Table 2, Kriging1 vs. poly MedP). These results
show very good model performance; however, as the model was built by using all P samples
and it was evaluated using the same data, the good agreement was expected. The Kriging1
model performance based on the data that were not included in model building (P samples
taken in 2021) gave the following results: R2 = 0.59, MAE = 30.7, and RMSE = 49.6 (Table 2,
Kriging1 vs. P 2021), indicating significant interannual and spatial topsoil P variation due to
fertilization and crop rotation.
The Kriging2 model results were evaluated by comparing the test dataset that was
not used in the kriging surface prediction to the soil–land use polygon median P values
(R2 = 0.75, MAE = 19.7, and RMSE = 31.6; Table 2, Kriging2_test vs. poly MedP). With rela-
tively good accordance of this comparison in proportion to the abovementioned compar-
ison between the Kriging1 model and 2021 sampling data, it reflects that interannual var-
iation in arable-land topsoil P content might be a more important factor than spatial vari-
ation.
• Bagging validation
Bagging model performance was validated on the test dataset (bagging model was
built on 75% of randomly selected bootstrap sample and 25% of data were left for model
validation), and the results were as follows: R2 = 0.54, MAE = 30.8, and RMSE = 48.6 (Table
3, Bagging_test vs. P). The evaluation of the bagging model results against all measured P
samples (both training and test data) show moderate prediction capability (R2 = 0.56, MAE
= 30.4, and RMSE = 47.3 (Table 3, Bagging vs. P)).
Agronomy 2023, 13, 1183 18 of 24
Table 3. Validation results of the bagging model.
Model MAE RMSE R2 Explanation
Bagging vs. P 30.4 47.3 0.56 Bagging prediction results validated against measured P val-
ues (all P samples used for validation).
Bagging vs.
poly MedP 20.9 33.6 0.72
Bagging prediction results validated against polygon median
P values (all P samples used for validation).
Bagging vs. P
2021 36.9 58.3 0.43
Bagging prediction results validated against P samples col-
lected in 2021 and not used for model building.
Bagging + pol-
ygon MedP
vs. P 2021
34.3 56.5 0.49
Bagging model values replaced with soil–land use polygon
median P values in polygons where P sample points measured
in 2005–2020 exist, and evaluated against P samples collected
in 2021.
Bagging_test
vs. P 30.8 48.6 0.54
Bagging model results validated on test dataset (25% of data
not used for model building).
Bagging 10-
fold CV 31.3 49.1 0.52
Bagging model performance evaluation by using a random 10-
fold cross-validation dataset (75% of data used for model build-
ing).
Bagging OOB
error 50.3
Bagging model performance evaluation by using out-of-bag
(OOB) samples. From the training dataset (75% of data) se-
lected randomly for each bagged predictor, approximately
63% of data were used for training, and the remaining 37%
were used as OOB samples.
Poly MedP vs.
P 18.8 36.4 0.74
Polygon median P values compared to P values measured in
2005–2020 (polygon median P value was calculated based on
the same sample).
Poly MedP vs.
P2021 32.6 54.6 0.53
Polygon median P values (calculated based on P sample val-
ues in 2005–2020) compared to P values measured in 2021.
The bagging predicted P values against the polygon median P values show good
agreement: R2 = 0.72, MAE = 20.9, and RMSE = 33.6 (Table 3, Bagging vs. poly MedP). The
PANDA database P samples collected in 2021 were used as an additional validation da-
taset to evaluate the bagging model performance. The result shows a weaker prediction
capability like in the case of kriging1 with similar validation: R2 = 0.43, MAE = 36.9, and
RMSE = 58.3 (Table 3, Bagging vs. P 2021). When the bagging prediction results were re-
placed with the soil–land use polygon median P values (in polygons where P samples
were taken in 2005–2020) and the results were compared to the P samples taken in 2021,
the results were only slightly better: R2 = 0.49, MAE = 34.3, and RMSE = 56.5 (Table 3,
Bagging + polygon MedP vs. P 2021).
The bagging model performance was also evaluated by calculating the out-of-bag
estimate, which used the training data that were left out from the random bootstrap sam-
pling with replacement (OOB sample) as a test dataset (RMSE = 50.3; Table 3, Bagging
OOB error). The cross-validation (CV) and 10-fold random sampling gave the following
results: RMSE = 49.1, R2 = 0.52, and MAE = 31.3 (Table 3, Bagging 10-fold CV).
4. Discussion
High-resolution modeling of topsoil plant-available P over large areas is a compli-
cated task as P content is influenced by both natural and human-driven processes; some
of these phenomena are continuous in time and space (e.g., soil development, parent ma-
terial, erosion, climate), while management of agricultural lands can affect P content in
topsoil sharply at the border of land parcels, even in cases of the same soil and crop types,
or over the years, bearing in mind legacy P as well.
Previous studies have shown that soil P content is heterogeneous even at the field
level [59], although land use is also an important factor determining P content. Our results
show that the topsoil P values inside the same soil–land use polygon can have quite high
variance (mean P range inside soil–land use polygon 58 mg/kg, median range 38 mg/kg,
Agronomy 2023, 13, 1183 19 of 24
maximum range of 198 mg/kg). Very high P values are usually related to croplands but
are sometimes observed in pasture areas as well, which can be explained by the presence
of livestock and continuous addition of manure. This phenomenon has been observed in
extensively grazed natural grasslands and wooded meadows in Estonia, and it is also rec-
orded by Roger et al. [13]. Roger et al. [13] showed that due to continuous addition of
manure (on average, 737 mg/kg organic P), permanent grasslands and mountain pasture
areas, compared to croplands, have the lowest quantity of total P, but the highest quantity
of available P.
There is an inherent problem of unbalanced monitoring data by land-use categories
but also by soil types, which affects spatial analysis of soil properties [60,61]. The relative
lack of information from non-agricultural areas results in bias of data by spatial extent and
land-use types. The relationship established between the explanatory variables and pre-
dicted variable in a managed environment could not be directly applied in natural ecosys-
tems. While topsoil P is relatively uniform by soil type in natural environments, the spatial
and temporal variability of managed ecosystems is very high, being affected by land uses,
fertilization practices, crop rotation, as well as the legacy P. Most authors, who have pre-
dicted P values for larger areas, have found that prediction uncertainty is high [23], and
higher uncertainty has been associated with sparse measurements and higher measured P
values [19].
Most commonly used ancillary environmental variables for P prediction models in-
clude soil properties, e.g., clay, sand, soil organic matter, pH, and soil type [13,17,20,62],
different topographic variables, such as slope, elevation, curvature, LS factor and topo-
graphic wetness index [13,17,20], vegetation and remote sensing data mostly in the form of
vegetation indices, e.g., NDVI [16,22], land use [13,17], and climatic variables (e.g., temper-
ature and precipitation indices) [17]. Similar to our results, higher P values are associated
with anthropogenic land use and fertilizer application [19]. According to Hosseini et al. [20]
attributes related to agricultural practices contributed more to plant-available P models,
whereas soil and geological attributes contributed more to TP (total P) models. Roger et al.
[13] found that environmental variables (elevation, slope, wetness index, and plan curvature
derived from a digital elevation model) explained about 20–25% of spatial variance of dif-
ferent P forms. In our study, soil type explained about 15% of P variance, land-use category
explained 13%, and other parameters contributed individually less, although most of them
were statistically significant in the bagging model.
The kriging model shows the best results in model evaluation, but its use is limited
to land-use categories with extensive sampling data available as interpolation methods
are primarily based on spatial autocorrelation. The bagging and ordination models are
not related to certain locations, although the bagging model includes also spatial variables
(latitude and longitude), and, therefore, the values that these models predict are not di-
rectly connected to the sample point values in particular locations.
5. Conclusions
Three methods were used for data-driven spatial modeling to produce a spatially
detailed map of plant-available topsoil P content in Estonia. Simple kriging performed the
best (R2 = 0.82, MAE = 19.8, and RMSE = 31.3) for agricultural lands where sampling den-
sity was high and spatially well distributed; however, due to the method’s reliance on
spatial autocorrelation, it is unable to predict values in areas with high spatial variance in
soil or land-use types, and in areas with missing sampling points or with low sampling
density where the active lag distance is exceeded.
The results of the bagging model were slightly inferior (R2 = 0.56, MAE = 30.4, and
RMSE = 47.3) to the kriging model. However, bagging models are not directly related to
certain sampling locations nor sampling density but incorporate supplementary spatial
variables (e.g., latitude, longitude, elevation land-use classes, and soil properties); there-
fore, these models are more capable of predicting available P values in agricultural lands
distant to any sampling point. Nevertheless, neither bagging nor kriging models, which
Agronomy 2023, 13, 1183 20 of 24
heavily rely on an extensive database of measured topsoil P content from agricultural
soils, could be used to predict topsoil P values outside of agricultural lands, where the
processes of P mineralization and immobilization, often biological in nature, are generally
different than in agricultural ecosystems where P is added as a fertilizer and crop rotation
affects the topsoil P status.
In large non-agricultural areas (wetlands and forests, covering 61% of Estonian land
area), topsoil plant-available P content was infrequently sampled with poor spatial cover-
age. To overcome the spatial soil P data scarcity, a two-parameter soil–land use ordination
matrix-based model was developed. The median topsoil P value of this ordination ap-
proach did not explain the spatial variance (R2 = 0.20, MAE = 36.7, and RMSE = 56.5) in
data-abundant agricultural areas; however, soil type alone explained about 15% of P var-
iance and land-use class contributed less (13%), thus making it sufficient for gap filling in
these spatially more homogeneous areas, especially for peat soils.
The final hybrid topsoil plant-available P map (R2 = 0.74, MAE = 18.8, and RMSE = 36.4)
was produced by combining arable-land polygons with the median value of the sampling
data, the bagging-based predicted values for agricultural lands with missing sampling data,
and the two-parameter soil–land use ordination matrix-based model values for forests and
wetlands.
The predicted topsoil plant-available P map can be used to assess soil P fertilization
need, estimate legacy P, and optimize P fertilizer application, taking into consideration
soil properties to reduce P loss and eutrophication of waterbodies.
Supplementary Materials: The following supporting information can be downloaded at
https://www.mdpi.com/article/10.3390/agronomy13051183/s1. Figure S1: Topsoil plant-available P
sample points taken from agricultural lands in 2005–2020. The count (n) and median P value (mg/kg)
are presented for every year; Table S1: Variables used in bagging model. P is the predicted variable,
and 32 variables were used to predict the P values. Index 1 after sand, silt, clay, rock, soc, bd, k, and
awc indicates the upper soil layer (topsoil). Different statistics (mean, median, and standard devia-
tion) of the topographic variables (TRI, TWI, LS, and SLP) were included in the bagging model.
References [63–68] are cited in Supplementary Materials.
Author Contributions: Conceptualization, A.K. (Ain Kull), P.P., and A.K. (Anne Kull); methodology,
A.K. (Ain Kull) and A.K. (Anne Kull); formal analysis, A.K. (Anne Kull), A.K. (Ain Kull), and T.K.; re-
sources, A.K. (Ain Kull) and P.P.; data curation, A.K. (Anne Kull), A.K. (Ain Kull), and T.K.; writing—
original draft preparation, A.K. (Anne Kull) and A.K. (Ain Kull); writing—review and editing, A.K. (Ain
Kull), A.K. (Anne Kull), T.K., and P.P.; visualization, A.K. (Anne Kull) and A.K. (Ain Kull); funding ac-
quisition, A.K. (Ain Kull) and P.P. All authors have read and agreed to the published version of the man-
uscript.
Funding: This research was supported by the Estonian Research Council research grants PRG352,
PRG1167, and SLTOM19384 (WaterJPI-JC-2018_13); the Estonian Environmental Investment Centre
grants SLOOM12006 and SLOOM14103; the Estonian State Forest Management Centre grant
LLTOM17250; and the Ministry of Rural Affairs.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The high-resolution plant-available topsoil phosphorous hybrid map
generated during the study is available at the following site: https://datadoi.ee/handle/33/405?lo-
cale-attribute=en (accessed on 13 January 2023).
Acknowledgments: The authors would like to thank Elsa Putku for her valuable comments and
discussions during manuscript preparation process.
Conflicts of Interest: The authors declare no conflicts of interest. The funders had no role in the
design of the study; in the collection, analyses, or interpretation of data; in the writing of the manu-
script, or in the decision to publish the results.
Agronomy 2023, 13, 1183 21 of 24
References
1. Harrison, A.F. Soil Organic Phosphorous: A Review of World Literature; CAB international: Singapore, 1987.
2. Bardgett, R. The Biology of Soil: A Community and Ecosystem Approach; Oxford University Press: New York, NY, USA, 2005.
3. van der Wal, A.; de Boer, W.; Lubbers, I.M.; van Veen, J.A. Concentration and vertical distribution of total soil phosphorus in
relation to time of abandonment of arable fields. Nutr. Cycl. Agroecosystems 2007, 79, 73–79. https://doi.org/10.1007/s10705-007-
9097-3. Available online: https://agris.fao.org/agris-search/search.do?recordID=US201300792922 (accessed on 13 January 2023).
4. Batjes, N.H. Global distribution of soil phosphorus retention potential. ISRIC World Soil Inf. 2011, 6.
5. Penn, C.; Camberato, J. A Critical Review on Soil Chemical Processes that Control How Soil pH Affects Phosphorus Availability
to Plants. Agriculture 2019, 9, 120. https://doi.org/10.3390/agriculture9060120. Available online: https://www.mdpi.com/2077-
0472/9/6/120 (accessed on 13 January 2023).
6. Brady, N.C.; Weil, R.R. Soil organic matter. The nature and properties of soils. Prentice Hall: Upper Saddle River, NJ, USA, 1999.
7. Tarafdar, J.C.; Claassen, N. Organic phosphorus compounds as a phosphorus source for higher plants through the activity of
phosphatases produced by plant roots and microorganisms. Biol Fertil. Soils 1988, 5, 308–312. https://doi.org/10.1007/BF00262137.
Available online: https://www.researchgate.net/profile/J-Tarafdar/publication/226638701_Organic_Phosphorus_Com-
pounds_as_a_Phosphorus_Source_for_Higher_Plants_through_the_Activity_of_Phosphatases_Pro-
duced_by_Plant_Roots_and_Microorganisms/links/5816002208ae90acb240e93f/Organic-Phosphorus-Compounds-as-a-Phos-
phorus-Source-for-Higher-Plants-through-the-Activity-of-Phosphatases-Produced-by-Plant-Roots-and-Microorganisms.pdf
(accessed on 13 January 2023).
8. Beckett, P.H.T.; White, R.E. Studies ON THE PHOSPHATE POTENTIALS OF SOILS: PART III: THE POOL OF LABILE INOR-
GANIC PHOSPHATE. Plant Soil 1964, 21, 253–282. https://doi.org/10.1007/BF01377744. Available online:
https://www.jstor.org/stable/42932073 (accessed on 13 January 2023).
9. Hinsinger, P. Bioavailability of soil inorganic P in the rhizosphere as affected by root-induced chemical changes: A review. Plant
Soil 2001, 237, 173–195. https://doi.org/10.1023/A:1013351617532. Available online: https://www.researchgate.net/pro-
file/Philippe-Hinsinger/publication/225852665_Bioavailability_of_soil_inorganic_P_in_the_rhizosphere_as_affected_by_root-
induced_chemical_changes_A_review/links/53d751c60cf228d363eaebbd/Bioavailability-of-soil-inorganic-P-in-the-rhizo-
sphere-as-affected-by-root-induced-chemical-changes-A-review.pdf (accessed on 13 January 2023).
10. Lambers, H.; Raven, J.A.; Shaver, G.R.; Smith, S.E. Plant nutrient-acquisition strategies change with soil age. Trends Ecol. Evol.
2008, 23, 95–103. https://doi.org/10.1016/j.tree.2007.10.008. Available online: https://www.sciencedirect.com/science/arti-
cle/pii/S0169534707003576 (accessed on 13 January 2023).
11. Nelson, L.; Cade-Menun, B.J.; Walker, I.J.; Sanborn, P. Soil Phosphorus Dynamics Across a Holocene Chronosequence of Aeo-
lian Sand Dunes in a Hypermaritime Environment on Calvert Island, BC, Canada. Front. For. Glob. Chang. 2020, 3, 83.
https://doi.org/10.3389/ffgc.2020.00083. Available online: https://doaj.org/article/2fbe4a8913d14de3a470466da86e84c5 (accessed
on 13 January 2023).
12. Tian, J.; Dong, G.; Karthikeyan, R.; Li, L.; Harmel, R. Phosphorus Dynamics in Long-Term Flooded, Drained, and Reflooded
Soils. Water 2017, 9, 531. https://doi.org/10.3390/w9070531.
13. Roger, A.; Libohova, Z.; Rossier, N.; Joost, S.; Maltas, A.; Frossard, E.; Sinaj, S. Spatial variability of soil phosphorus in the
Fribourg canton, Switzerland. Geoderma 2014, 217, 26–36. https://doi.org/10.1016/j.geoderma.2013.11.001. Available online:
https://dx.doi.org/10.1016/j.geoderma.2013.11.001 (accessed on 13 January 2023).
14. Mondal, B.P.; Sekhon, B.S.; Sadhukhan, R.; Singh, R.K.; Hasanain, M.; Mridha, N.; Das, B.; Dhyay, A.C.; Banerjee, K. Spatial
variability assessment of soil available phosphorus using geostatistical approach. Indian J. Agric. Sci. 2020, 90, 1170–1175.
https://doi.org/10.56093/ijas.v90i6.104795. Available online: https://www.researchgate.net/publication/344326265_Spatial_vari-
ability_assessment_of_soil_available_phosphorus_using_geostatistical_approach (accessed on 13 January 2023).
15. Scull, P.; Franklin, J.; Chadwick, O.A.; McArthur, D. Predictive soil mapping: A review. Prog. Phys. Geogr. 2003, 27, 171–197.
https://doi.org/10.1191/0309133303pp366ra. Available online: https://jour-
nals.sagepub.com/doi/full/10.1191/0309133303pp366ra (accessed on 13 January 2023).
16. Rivero, R.G.; Grunwald, S.; Bruland, G.L. Incorporation of spectral data into multivariate geostatistical models to map soil
phosphorus variability in a Florida wetland. Geoderma 2007, 140, 428–443. https://doi.org/10.1016/j.geoderma.2007.04.026. Avail-
able online: https://www.sciencedirect.com/science/article/pii/S001670610700122X (accessed on 13 January 2023).
17. Ballabio, C.; Lugato, E.; Fernández-Ugalde, O.; Orgiazzi, A.; Jones, A.; Borrelli, P.; Montanarella, L.; Panagos, P. Mapping LU-
CAS topsoil chemical properties at European scale using Gaussian process regression. Geoderma 2019, 355, 113912.
https://doi.org/10.1016/j.geoderma.2019.113912. Available online: https://www.sciencedirect.com/science/arti-
cle/pii/S0016706119304768 (accessed on 13 January 2023).
18. Matos-Moreira, M.; Lemercier, B.; Dupas, R.; Michot, D.; Viaud, V.; Akkal-Corfini, N.; Louis, B.; Gascuel-Odoux, C. High-reso-
lution mapping of soil phosphorus concentration in agricultural landscapes with readily available or detailed survey data. Eur.
J. Soil Sci. 2017, 68, 281–294. https://doi.org/10.1111/ejss.12420. Available online: https://onlineli-
brary.wiley.com/doi/abs/10.1111/ejss.12420 (accessed on 13 January 2023).
19. Viscarra Rossel, R.A.; Bui, E.N. A new detailed map of total phosphorus stocks in Australian soil. Sci. Total Environ. 2016, 542,
1040–1049. https://doi.org/10.1016/j.scitotenv.2015.09.119. Available online: https://dx.doi.org/10.1016/j.scitotenv.2015.09.119
(accessed on 13 January 2023).
Agronomy 2023, 13, 1183 22 of 24
20. Hosseini, M.; Rajabi Agereh, S.; Khaledian, Y.; Jafarzadeh Zoghalchali, H.; Brevik, E.C.; Movahedi Naeini, S.A.R. Comparison
of multiple statistical techniques to predict soil phosphorus. Appl. Soil Ecol. 2017, 114, 123–131. https://doi.org/10.1016/j.ap-
soil.2017.02.011. Available online: https://www.sciencedirect.com/science/article/pii/S0929139316304085 (accessed on 13 Janu-
ary 2023).
21. Esfandiarpour-Boroujeni, I.; Shahini-Shamsabadi, M.; Shirani, H.; Mosleh, Z.; Bagheri-Bodaghabadi, M.; Salehi, M.H. Assess-
ment of different digital soil mapping methods for prediction of soil classes in the Shahrekord plain, Central Iran. Catena 2020,
193, 104648. https://doi.org/10.1016/j.catena.2020.104648. Available online: https://www.sciencedirect.com/science/arti-
cle/pii/S0341816220301983 (accessed on 13 January 2023).
22. Bogrekci, I.; Lee, W.S. Spectral Phosphorus Mapping using Diffuse Reflectance of Soils and Grass. Biosyst. Eng. 2005, 91, 305–
312. https://doi.org/10.1016/j.biosystemseng.2005.04.015. Available online: https://www.sciencedirect.com/science/arti-
cle/pii/S1537511005000814 (accessed on 13 January 2023).
23. Yang, X.; Post, W.M.; Thornton, P.E.; Jain, A. The distribution of soil phosphorus for global biogeochemical modeling. Biogeo-
sciences 2013, 10, 2525–2537. https://doi.org/10.5194/bg-10-2525-2013. Available online:
https://search.proquest.com/docview/1349834631 (accessed on 13 January 2023).
24. Iatrou, M.; Papadopoulos, A.; Papadopoulos, F.; Dichala, O.; Psoma, P.; Bountla, A. Determination of Soil Available Phosphorus
using the Olsen and Mehlich 3 Methods for Greek Soils Having Variable Amounts of Calcium Carbonate. Commun. Soil Sci.
Plant Anal. 2014, 45, 2207–2214. https://doi.org/10.1080/00103624.2014.911304. Available online: https://agris.fao.org/agris-
search/search.do?recordID=US201500196377 (accessed on 13 January 2023).
25. Szara, E.; Sosulski, T.; Szymańska, M.; Szyszkowska, K. Usefulness of Mehlich-3 test in the monitoring of phosphorus dispersion
from Polish arable soils. Environ. Monit. Assess. 2018, 190, 1–10. https://doi.org/10.1007/s10661-018-6685-4. Available online:
https://link.springer.com/article/10.1007/s10661-018-6685-4 (accessed on 13 January 2023).
26. Vares, A.; Jõgiste, K.; Kull, E. Early growth of some deciduous tree species on abandoned agricultural lands in Estonia. Balt. For.
2001, 7, 52–58.
27. Laas, A.; Kull, A. Application Of GIS For Soil Erosion And Nutrient Loss Modelling In A Small River Catchment. WIT Trans.
Ecol. Environ. 2003, 67, 525-534. https://doi.org/10.2495/SPD030491. Available online:
https://search.proquest.com/docview/2261516459 (accessed on 13 January 2023).
28. Vares, A.; Uri, V.; Tullus, H.; Kanal, A. Height growth of four fast-growing deciduous tree species on former agricultural lands
in Estonia. Balt. For. 2003, 9, 2–8.
29. Kull, A.; Kull, A.; Jaagus, J.; Kuusemets, V.; Mander, U. The effects of fluctuating climatic conditions and weather events on
nutrient dynamics in a narrow mosaic riparian peatland. Boreal Environ. Res. 2008, 13, 243–263. Available online:
https://helda.helsinki.fi/bitstream/handle/10138/234736/ber13-3-243.pdf?sequence=1 (accessed on 13 January 2023).
30. Kund, M.; Vares, A.; Sims, A.; Tullus, H.; Uri, V. Early growth and development of silver birch (Betula pendula Roth.) planta-
tions on abandoned agricultural land. Eur. J. Forest Res. 2010, 129, 679–688. https://doi.org/10.1007/s10342-010-0369-0. Available
online: https://agris.fao.org/agris-search/search.do?recordID=US201301854605 (accessed on 13 January 2023).
31. Varik, M.; Aosaar, J.; Ostonen, I.; Lõhmus, K.; Uri, V. Carbon and nitrogen accumulation in belowground tree biomass in a
chronosequence of silver birch stands. For. Ecol. Manag. 2013, 302, 62–70. https://doi.org/10.1016/j.foreco.2013.03.033. Available
online: https://dx.doi.org/10.1016/j.foreco.2013.03.033 (accessed on 13 January 2023).
32. Lutter, R.; Tullus, T.; Tullus, A.; Kanal, A.; Tullus, H. Forest Ecosystem Recovery in 15-Year-Old Hybrid Aspen (Populus Trem-
ula l. × p. Tremuloides Michx.) Plantations on a Reclaimed Oil Shale Quarry. Oil Shale (Tallinn Est. 1984) 2017, 34, 368–389.
https://doi.org/10.3176/oil.2017.4.05. Available online: https://search.proquest.com/docview/1973739673 (accessed on 13 Janu-
ary 2023).
33. Lutter, R.; Tullus, A.; Kanal, A.; Tullus, T.; Vares, A.; Tullus, H. Growth development and plant–soil relations in midterm silver
birch (Betula pendula Roth) plantations on previous agricultural lands in hemiboreal Estonia. Eur. J. Forest Res. 2015, 134, 653–
667. https://doi.org/10.1007/s10342-015-0879-x. Available online: https://link.springer.com/article/10.1007/s10342-015-0879-x
(accessed on 13 January 2023).
34. Uri, V.; Kukumägi, M.; Aosaar, J.; Varik, M.; Becker, H.; Soosaar, K.; Morozov, G.; Ligi, K.; Padari, A.; Ostonen, I.; Karoles, K.
Carbon budgets in fertile grey alder (Alnus incana (L.) Moench.) stands of different ages. For. Ecol. Manag. 2017, 396, 55–67.
https://doi.org/10.1016/j.foreco.2017.04.004. Available online: https://dx.doi.org/10.1016/j.foreco.2017.04.004 (accessed on 13 Jan-
uary 2023).
35. Uri, V.; Kukumägi, M.; Aosaar, J.; Varik, M.; Becker, H.; Morozov, G.; Karoles, K. Ecosystems carbon budgets of differently aged
downy birch stands growing on well-drained peatlands. For. Ecol. Manag. 2017, 399, 82–93.
https://doi.org/10.1016/j.foreco.2017.05.023. Available online: https://dx.doi.org/10.1016/j.foreco.2017.05.023 (accessed on 13 Jan-
uary 2023).
36. Uri, V.; Aosaar, J.; Varik, M.; Becker, H.; Kukumägi, M.; Ligi, K.; Pärn, L.; Kanal, A. Biomass resource and environmental effects
of Norway spruce (Picea abies) stump harvesting: An Estonian case study. For. Ecol. Manag. 2015, 335, 207–215.
https://doi.org/10.1016/j.foreco.2014.10.003. Available online: https://dx.doi.org/10.1016/j.foreco.2014.10.003 (accessed on 13 Jan-
uary 2023).
Agronomy 2023, 13, 1183 23 of 24
37. Uri, V.; Kukumägi, M.; Aosaar, J.; Varik, M.; Becker, H.; Aun, K.; Krasnova, A.; Morozov, G.; Ostonen, I.; Mander, Ü .; Lõhmus,
K.; Rosenvald, K.; Kriiska, K.; Soosaar, K. The carbon balance of a six-year-old Scots pine (Pinus sylvestris L.) ecosystem esti-
mated by different methods. For. Ecol. Manag. 2019, 433, 248–262. https://doi.org/10.1016/j.foreco.2018.11.012. Available online:
https://dx.doi.org/10.1016/j.foreco.2018.11.012 (accessed on 13 January 2023).
38. Varik, M.; Kukumägi, M.; Aosaar, J.; Becker, H.; Ostonen, I.; Lõhmus, K.; Uri, V. Carbon budgets in fertile silver birch (Betula
pendula Roth) chronosequence stands. Ecol. Eng. 2015, 77, 284–296. https://doi.org/10.1016/j.ecoleng.2015.01.041. Available
online: https://dx.doi.org/10.1016/j.ecoleng.2015.01.041 (accessed on 13 January 2023).
39. Aosaar, J.; Mander, Ü .; Varik, M.; Becker, H.; Morozov, G.; Maddison, M.; Uri, V. Biomass production and nitrogen balance of
naturally afforested silver birch (Betula pendula Roth.) stand in Estonia. Silva Fenn. 2016, 50. https://doi.org/10.14214/sf.1628.
Available online: https://www.silvafennica.fi/article/1628 (accessed on 13 January 2023).
40. Aosaar, J.; Drenkhan, T.; Adamson, K.; Aun, K.; Becker, H.; Buht, M.; Drenkhan, R.; Fjodorov, M.; Jürimaa, K.; Morozov, G.;
Pihlak, L.; Piiskop, K.; Riit, T.; Varik, M.; Väär, R.; Uri, M.; Uri, V. The effect of stump harvesting on tree growth and the infection
of root rot in young Norway spruce stands in hemiboreal Estonia. For. Ecol. Manag. 2020, 475, 118425.
https://doi.org/10.1016/j.foreco.2020.118425. Available online: https://dx.doi.org/10.1016/j.foreco.2020.118425 (accessed on 13
January 2023).
41. Aun, K.; Kukumägi, M.; Varik, M.; Becker, H.; Aosaar, J.; Uri, M.; Morozov, G.; Buht, M.; Uri, V. Short-term effect of thinning
on the carbon budget of young and middle-aged Scots pine (Pinus sylvestris L.) stands. For. Ecol. Manag. 2021, 492, 119241.
https://doi.org/10.1016/j.foreco.2021.119241. Available online: https://dx.doi.org/10.1016/j.foreco.2021.119241 (accessed on 13
January 2023).
42. Becker, H.; Aosaar, J.; Varik, M.; Morozov, G.; Kanal, A.; Uri, V. The effect of Norway spruce stump harvesting on net nitrogen
mineralization and nutrient leaching. For. Ecol. Manag. 2016, 377, 150–160. https://doi.org/10.1016/j.foreco.2016.07.005. Available
online: https://dx.doi.org/10.1016/j.foreco.2016.07.005 (accessed on 13 January 2023).
43. Paal, J.; Jürjendal, I.; Suija, A.; Kull, A. Impact of drainage on vegetation of transitional mires in Estonia. Mires Peat 2016, 18, 1–
19. https://doi.org/10.19189/MaP.2015.OMB.183. Available online: http://mires-and-peat.net/media/map18/map_18_02.pdf (ac-
cessed on 13 January 2023).
44. Tullus, T.; Lutter, R.; Randlane, T.; Saag, A.; Tullus, A.; Oja, E.; Degtjarenko, P.; Pärtel, M.; Tullus, H. The effect of stand age on
biodiversity in a 130-year chronosequence of Populus tremula stands. For. Ecol. Manag. 2022, 504, 119833.
https://doi.org/10.1016/j.foreco.2021.119833. Available online: https://dx.doi.org/10.1016/j.foreco.2021.119833 (accessed on 13
January 2023).
45. Kull, A. Buffer zones to limit and mitigate harmful effects of long-term anthropogenic influence to maintain ecological func-
tionality of bogs, stage II. 2016, 183. (In Estonian). Available online: https://4ce0b57b-a630-4e1d-8a88-
d1e1a7a51b96.filesusr.com/ugd/6b6658_446958f4118b44a2a68812820c31119b.pdf (accessed on 13 January 2023).
46. Asi, E.; Timmusk, T. Greenhouse gas emissions inventory studies for the national reporting on forest litter and soil. 2018, 69. (In
Estonian). Available online: file:///C:/Users/Administrator/Downloads/kik_metsanduse_projekt_12654_loppa-
ruanne_koos_lisaga_0-3.pdf (accessed on 13 January 2023).
47. Kulhánek, M.; Balík, J.; Černý, J.; Vaněk, V. Evaluation of phosphorus mobility in soil using different extraction methods. Plant
Soil Environ. 2009, 55, 267–272. https://doi.org/10.17221/43/2009-PSE. Available online: https://www.agriculturejour-
nals.cz/pdfs/pse/2009/07/01.pdf (accessed on 13 January 2023).
48. Wolf, A.M.; Baker, D.E. Comparisons of soil test phosphorus by Olsen, Bray P1, Mehlich I and Mehlich III methods. Commun.
Soil Sci. Plant Anal. 1985, 16, 467–484. https://doi.org/10.1080/00103628509367620. Available online: https://agris.fao.org/agris-
search/search.do?recordID=US8614205 (accessed on 13 January 2023).
49. Kmoch, A.; Kanal, A.; Astover, A.; Kull, A.; Virro, H.; Helm, A.; Pärtel, M.; Ostonen, I.; Uuemaa, E. EstSoil-EH: A high-resolution
eco-hydrological modelling parameters dataset for Estonia. Earth Syst. Sci. Data 2021, 13, 83–97. https://doi.org/10.5194/essd-13-
83-2021. Available online: https://essd.copernicus.org/articles/13/83/2021/ (accessed on 13 January 2023).
50. Kachinsky, N. Soil Physics. 1965; 320 p. (In Russian)
51. Abd-Elsamad, A.; Abdelmoein, N.M.; Mahmoud, A.H.; Rostom, M.; Hassan, S.M.; Gazni, R. Evaluation and comparison of
ordinary kriging and inverse distance weighting methods for prediction of spatial variability of some soil chemical parameters.
Res. J. Biol. Sci. 2009, 4, 93–102. Available online: https://www.researchgate.net/publication/266732920 (accessed on 13 January
2023).
52. Sahu, B.; Ghosh, A.K. Seema Deterministic and geostatistical models for predicting soil organic carbon in a 60 ha farm on In-
ceptisol in Varanasi, India. Geoderma Reg. 2021, 26, e00413. https://doi.org/10.1016/j.geodrs.2021.e00413. Available online:
https://www.sciencedirect.com/science/article/pii/S2352009421000584 (accessed on 13 January 2023).
53. Xue, W.; Pangara, C.; Aung, A.M.; Yu, S.; Tabucanon, A.S.; Hong, B.; Kurniawan, T.A. Spatial changes of nutrients and metallic
contaminants in topsoil with multi-geostatistical approaches in a large-size watershed. Sci. Total Environ. 2022, 824, 153888.
https://doi.org/10.1016/j.scitotenv.2022.153888. Available online: https://dx.doi.org/10.1016/j.scitotenv.2022.153888 (accessed on
13 January 2023).
54. Peet, R.K. Ordination as a tool for analyzing complex data sets. Vegetatio 1980, 42, 171–174. https://doi.org/10.1007/BF00048883.
Available online: https://citeseerx.ist.psu.edu/docu-
ment?repid=rep1&type=pdf&doi=182c33c89a9878641e1de783a1bffb26637f1326 (accessed on 13 January 2023).
Agronomy 2023, 13, 1183 24 of 24
55. Ter Braak, C.J.F.; Prentice, I.C. A Theory of Gradient Analysis. In Advances in Ecological Research; Elsevier Science & Technology:
Amsterdam, The Netherlands, 2004; Volume 34, pp. 235–282.
56. Breiman, L. Bagging predictors. Mach Learn. 1996, 24, 123–140. https://doi.org/10.1023/A:1018054314350.
57. Therneau, T.M.; Atkinson, E.J. An Introduction to Recursive Partitioning Using the RPART Routines. Introd. Recursive Partit.
Using RPART Routines 2022, 61. Available online: https://stat.ethz.ch/R-manual/R-patched/library/rpart/doc/longintro.pdf (ac-
cessed on 13 January 2023).
58. Petersell, V.; Ressar, H.; Carlsson, M.; Mõttus, V.; Enel, M.; Mardla, A.; Täht, K. Geochemical atlas of Estonian agricultural soil.
Eest. Geol. Sver. Geol. Undersökning. Tallinn Upps. Seletuskiri 1997, 75, 171 p. Available online: https://www.di-
gar.ee/viewer/et/nlib-digar:331323/291375/page/109. (In Estonian)
59. Werner, F.; Mueller, C.W.; Thieme, J.; Gianoncelli, A.; Rivard, C.; Höschen, C.; Prietzel, J. Micro-scale heterogeneity of soil
phosphorus depends on soil substrate and depth. Sci. Rep. 2017, 7, 3203–3209. https://doi.org/10.1038/s41598-017-03537-8. Avail-
able online: https://www.ncbi.nlm.nih.gov/pubmed/28600571 (accessed on 13 January 2023).
60. van Leeuwen, J.P.; Saby, N.P.A.; Jones, A.; Louwagie, G.; Micheli, E.; Rutgers, M.; Schulte, R.P.O.; Spiegel, H.; Toth, G.; Creamer,
R.E. Gap assessment in current soil monitoring networks across Europe for measuring soil functions. ERL 2017, 12, 124007.
https://doi.org/10.1088/1748-9326/aa9c5c. Available online: https://iopscience.iop.org/article/10.1088/1748-9326/aa9c5c (ac-
cessed on 13 January 2023).
61. Schillaci, C.; Saia, S.; Lipani, A.; Perego, A.; Zaccone, C.; Acutis, M. Validating the regional estimates of changes in soil organic
carbon by using the data from paired-sites: The case study of Mediterranean arable lands. Carbon Balance Manag. 2021, 16, 19.
https://doi.org/10.1186/s13021-021-00182-7. Available online: https://link.springer.com/article/10.1186/s13021-021-00182-7 (ac-
cessed on 13 January 2023).
62. Keshavarzi, A.; Omran, E.E.; Bateni, S.M.; Pradhan, B.; Vasu, D.; Bagherzadeh, A. Modeling of available soil phosphorus (ASP)
using multi-objective group method of data handling. Model. Earth Syst. Environ. 2016, 2, 1–9. https://doi.org/10.1007/s40808-
016-0216-5. Available online: https://link.springer.com/article/10.1007/s40808-016-0216-5 (accessed on 13 January 2023).
63. Zhang, Y.; Schaap, M.G. Weighted recalibration of the Rosetta pedotransfer model with improved estimates of hydraulic pa-
rameter distributions and summary statistics (Rosetta3). J. Hydrol. 2017, 547, 39–53. https://doi.org/10.1016/j.jhydrol.2017.01.00.
Available online: https://www.sciencedirect.com/science/article/pii/S0022169417300057?casa_to-
ken=Ej7anM0xjK4AAAAA:5r0KM6e6cHxb1FI60ASBz5XveBbA_r-ATIJxgGDs8kXTvUxK1kTbfAf5l9sntbmn6d_KodQSTg (ac-
cessed on 13 January 2023).
64. Dipak, S.; Abhijit, H. Physical and chemical methods in soil analysis. New Age Int. Ltd. New Delhi 2005, 193 p.
65. Tóth, B.; Weynants, M.; Pásztor, L.; Hengl, T. 3D soil hydraulic database of Europe at 250 m resolution. Hydrol Process 2017, 31,
2662–2666. https://doi.org/10.1002/hyp.11203. Available online: https://onlinelibrary.wiley.com/doi/10.1002/hyp.11203 (ac-
cessed on 13 January 2023).
66. Riley, S.J.; DeGloria, S.D.; Elliot, R. Index that quantifies topographic heterogeneity. Intermt. J. Sci. 1999, 5, 23–27. Available
online: http://download.osgeo.org/qgis/doc/reference-docs/Terrain_Ruggedness_Index.pdf (accessed on 13 January 2023).
67. Beven, K.J.; Kirkby, M.J. A physically based, variable contributing area model of basin hydrology/Un modèle à base physique
de zone d’appel variable de l’hydrologie du bassin versant. Hydrol. Sci. J. 1979, 24, 43–69.
https://doi.org/10.1080/02626667909491834. Available online:
https://www.tandfonline.com/doi/abs/10.1080/02626667909491834 (accessed on 13 January 2023).
68. Moore, I.D.; Grayson, R.B.; Ladson, A.R. Digital terrain modelling: A review of hydrological, geomorphological, and biological
applications. Hydrol Process 1991, 5, 3–30. https://doi.org/10.1002/hyp.3360050103. Available online: https://onlineli-
brary.wiley.com/doi/10.1002/hyp.3360050103 (accessed on 13 January 2023).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual au-
thor(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
CAN BOG BREATHING BE MEASURED BY SYNTHETIC APERTURE RADAR INTERFEROMETRY
Tauri Tampuu1∗, Francesco De Zan2, Robert Shau2, Jaan Praks3, Marko Kohv1, Ain Kull1
1Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise Str. 46, Tartu 51014, Estonia 2German Aerospace Center (DLR), Münchener Str. 20, 82234 Weßling, Oberpfaffenhofen, Germany
3School of Electrical Engineering, Aalto University, Maarintie 8, 02150 Espoo, Finland ∗Correspondence: [email protected]
ABSTRACT Accounting for relatively large seasonal and short term peat- land surface vertical displacements with Synthetic Aperture Radar Interferometry (InSAR) poses a problem of possible propagation of ambiguity errors. Notwithstanding, the ab- sence of continuous high temporal resolution peatland sur- face levelling measurements for validation has been some- thing characteristic. Based on the ground levelling from a raised bog, we demonstrate the Sentinel-1 distributed scat- terer (DS) time-series InSAR technique underestimates real surface displacements and hereby we question the accuracy of the approach over peatlands. When the relative surface change from 6-day interferograms is used instead of account- ing for the absolute change, the estimation accuracy improves (Spearman’s rho 0.82, p-value < 0.002) because 6-day in situ surface changes are usually small and do not need InSAR phase unwrapping. Despite a serious unwrapping problem in peatlands, DS time series contain useful signal and differen- tial InSAR (DInSAR) might have potential for assessment of short term peatland surface displacements in favourable con- ditions.
Index Terms— InSAR, Surface deformation, Peatland, Phase ambiguity, Sentinel-1
1. INTRODUCTION
Better understanding of seasonal peat surface displacements initiated by changes in water table (bog breathing) [1] is needed to improve spatial models of greenhouse gas emis- sions [2]. Synthetic Aperture Radar Interferometry (InSAR) is a promising tool for the task in regard to the remote lo- cation and difcult accessibility of majority of the peatlands [3] and the need for a large scale assessment [4]. Neverthe- less, accounting for relatively large peatland surface vertical displacements [5, 6], which are occasionally extremely large and rapid [7], poses a problem of possible propagation of am- biguity errors and causes unreliability of the InSAR results [8, 9]. The concern has been to a great extent overlooked and the absence of high spatial and temporal resolution ground
levelling data for validation has been characteristic to InSAR research in peatlands [10, 8].
In this paper we present the ground levelling measurement results from a raised bog in Estonia, characteristic for the Northern temperate raised bogs, and show how the Sentinel-1 distributed scatterer (DS) time-series InSAR technique [11] underestimates the real magnitude of surface deformations over the ice and snow free period in the year 2016. There- fore, we question the feasibility of the time-series approach to measure (absolute) surface differences with respect to one common master acquisition in peatlands. Instead, we demon- strate that using the relative surface difference of 6-day image pairs form DS time series or single 6-day differential inter- ferograms (conventional DInSAR approach) can yield much less biased results because of the reduced need for unwrap- ping. Also, we argue that the useful signal to capture peat- land surface vertical displacement is contained in the DS time series.
2. METHODOLOGY
The ground levelling measurements are form Umbusi raised bog (58.57°N, 26.18°E) at observation plot 6 (a hummock mi- cro site) and cover the ice and snow free period of the year 2016. The plot situates in the intact portion of the natural bog, 200 m from a deep peat layer penetrating drainage ditch. The inuence of drainage does not affect the studied plot. The thickness of peat at the plot is ∼8 m. The levelling device recording the distance to the ground was attached to a metal bar which penetrates the peat layer and is anchored in the un- derling stable mineral ground. Change in ground height was recorded with hourly interval at 3 mm resolution.
The satellite Sentinel-1 A and B (S1A and S1B) vertical- vertical (VV) polarization ascending orbit (relative orbit num- ber 160) data were used. The DInSAR processing covers the period form 1 July to 29 Oct 2016 being limited by the over- lapping period of SAR acquisitions in early summer and by ground levelling data in late autumn. The DInSAR process- ing done with SARPROZ software resulted in thirteen small
16978-1-6654-2792-0/22/$31.00 ©2022 IEEE IGARSS 2022
IG A R SS
2 0 2 2 - 2 0 2 2 IE EE
In te rn at io n al G eo
sc ie n ce
an d R em
o te
Se n si n g Sy m p o si u m
| 9 7 8 -1 -6 6 5 4 -2 7 9 2 -0 /2 2 /$ 3 1 .0 0 © 2 0 2 2 IE EE
| D O I: 1 0 .1 1 0 9 /I G A R SS 4 6 8 3 4 .2 0 2 2 .9 8 8 3 4 2 1
Authorized licensed use limited to: University of Tartu Estonia. Downloaded on October 03,2022 at 17:57:39 UTC from IEEE Xplore. Restrictions apply.
temporal baseline interferograms. A 12-day temporal base- line available before and a 6-day temporal baseline after the launch of Sentinel-1 B. Interferometric coherence estimated (weighted by the amplitude; calculated before ltering) and Modied Goldstein phase ltering applied in a window size of 10 range (rg) and 3 in azimuth (az) pixels. The resul- tant square footprint on the ground with a side of ∼40 m. Flattening and the topographic phase removal and no multi- looking applied. Thereafter, the interferograms were refer- enced to the locations of the DS reference points to account for atmospheric effects. In that way, the reference would al- ways be set at 0 π and the unambiguous change could only be found among the phase changes not greater than ±1 π (a quarter of the radar wavelength) [12, 13]. In order to widen the span where the unambiguous change can be found, we rotated the ambiguous phase (setting the reference level any- where between ±1 π) to nd the best phase t and identify phase shifts along the transect stretching from the drain at the border of the bog to the central part of the bog where natural conditions are prevailing (the study plot 6 and beyond). In such a way, the dynamics seen along the transect indicated the direction of the peat surface change (subsidence or uplift) and consequently a reference could become set (anywhere be- tween±1 π) so that the subsidence could be found in the span of up to −2 π and the uplift in 2 π.
In the DS time-series processing [11], Sentinel-1 ascend- ing relative orbit number 160 acquisitions from 2014–2020 were included. Thereafter, the part of the time series from 2016 were extracted and referenced to the median of a cluster of stable reference points (8 points available ∼4 km away) to account for atmospheric effects (also the tropospheric phase simulated from ERA5 reanalysis data is removed before the DS calculation step). Only DS points with coherence > 0.9 were included in the analysis. The DS pixel footprint on the ground approximates to a square of a ∼200 m side.
The radar line of sight (LOS) altitude measurements were projected into the vertical direction (uLOS) using the local in- cidence angle of the plot 6, assuming no horizontal motion in the peat [14]. The Spearman’s rank-order correlation (rs) was applied to estimate correlations. The levelling plot 6 is not covered by a DS point, therefore the nearest DS points surrounding the plot (15 points 125–230 m away) are used to represent the plot 6 in calculations. Correlation between the DInSAR time series from the plot 6 and from the DS point locations in the vicinity are 0.83 (p-value < 0.001).
3. RESULTS
Hummocks and ridges are the stable-most micro site elements in the bog while surface uctuations at hollows and lawns are larger. Nevertheless, even during the summer of 2016 (we could not include the spring snow melt induced surface max- imum in our study due to the unavailability of Sentinel-1 data before July 1), the difference between the maximum and min-
Fig. 1. The DS and DInSAR line of sight altitude change projected to vertical direction (uLOS) compared to relative in situ vertical surface deformation between consecutive SAR acquisition dates at the Umbusi plot 6 hummock. The daily precipitation sum corresponds to the latest of the dates of the image pair.
imum in situ recorded surface height on the dates of SAR ac- quisition is more than 5 cm. The surface height difference from the yearly maximum would have been larger and sur- face height differences between years signicantly larger (as indicated by the water table uctuations in the bogs of Endla mire complex [15] 35 km north of Umbusi bog). Contrary, the relative surface height change between the consecutive SAR acquisition dates is considerably smaller and was only once larger than the Sentinel-1 LOS height of ambiguity in 2016 as shown in Figure 1. Nevertheless, we have to consider that the portion of hummocks and ridges versus hollows and lawns varies by different bogs and even by parts of the same bog.
In accordance with the known difculty of correct un- wrapping of the ambiguous phase [9], long temporal baselines and coinciding large in situ surface changes result in the DS time-series approach underestimating the real surface change, in line with what [8] have found. Despite the underestimation, the DS InSAR line of sight deformation projected to vertical dimension (uLOS) is following the trend in the levelling data (rs 0.76, p-value 0.004) (Figure 2a). Similarly, [16, 17] ig- nored the concerns of the absolute accuracy of InSAR and demonstrated the potential of the characteristics of the InSAR time series to be used to quantify peatland condition.
If the temporal baseline is reduced (according to the rec- ommendations by [8]) to the minimal possible (12 or 6 days in our case) via converting the absolute values of DS time series into changes between two consecutive acquisitions, then there is no need anymore for ambiguity resolution in most cases (Figure 2b). The correlation between the relative changes at the plot 6 hummock and the median relative uLOS DS defor- mation at the DS point locations in the vicinity of the plot 6 is 0.77 (p-value 0.005). The conventional DInSAR technique yields similar results. The rs of levelling data with the DIn-
17
Authorized licensed use limited to: University of Tartu Estonia. Downloaded on October 03,2022 at 17:57:39 UTC from IEEE Xplore. Restrictions apply.
(a) (b)
(c) (d)
Fig. 2. Correlation (rs) between the in situ surface deformation at the Umbusi plot 6 hummock and the InSAR line of sight deformation projected to vertical dimension (uLOS) in the ice and snow free period of 2016. Red points represent uLOS values if an ambiguous phase is added/subtracted. (a) The DS time series of absolute uLOS deformation. uLOS calculated as the median of the DS points in the vicinity of the plot 6. 2016-08-18 (the date of the maximum levelling height) taken to be the zero level. (b) The DS time series of relative uLOS deformation. (a, b) The median long term average γ shown. (c) The median DInSAR relative uLOS deformation at the DS point locations in the vicinity of the plot 6. (d) The median DInSAR relative uLOS deformation at the plot 6. A white X on a black background marks a data point of DInSAR coherence (γ) less than 0.4 (indicating unreliable phase estimates).
SAR estimates from the DS locations is 0.55 (p-value 0.077) (Figure 2c) and with the DInSAR pixel accommodating the plot 6, rs is 0.81 (p-value 0.002) (Figure 2d). The DInSAR results have been obtained with the stable reference points around 4 km away from the bog plot. A closer located stable reference points could improve results. We rotated the am- biguous phase along a transect in order to identify the direc- tion of the change. Alternatively, introduction of external data such as precipitation and temperature helps to better account for correct direction of the change [9]. The precipitation in regard to the relative DS and DInSAR surface height change estimates in Umbusi bog in 2016 ate presented in Figure 1.
4. CONCLUSION
A crucial step for application of InSAR in peatlands is the es- timation of the phase ambiguities derived from the relatively large surface height changes. We conclude, based on the in situ levelling data, that the direct application of time-series approach is unreliable in measuring seasonal and short term peatland surface vertical differences with respect to one com- mon date. The DS time series nevertheless contain the useful signal. The simplest way to tackle the ambiguity problem is to reduce the need for unwrapping by reducing temporal baselines. Consequently, we have used the relative surface difference of 6-day image pairs form DS time series or sin-
18
Authorized licensed use limited to: University of Tartu Estonia. Downloaded on October 03,2022 at 17:57:39 UTC from IEEE Xplore. Restrictions apply.
gle 6-day differential interferograms (conventional DInSAR). We conrmed based on our in situ levelling data that such an approach could reduce the estimation bias considerably in bog micro sites dominated by ridges and hummocks or areas of compacted peat which uctuate at less rapid pace and at smaller amplitude.
5. ACKNOWLEDGEMENT
This study is part of a PhD research supported by the Eu- ropean Union from the European Regional Development Fund, the national scholarship program Kristjan Jaak and by grants of Estonian State Forest Management Centre (LL- TOM17250) and Estonian Environmental Investment Centre (SLOOM12006 and SLOOM14103). The authors would like to thank Karsten Kretschmer (DLR) for help with Python and the Sarproz team for providing an excellent software with extremely exible student licensing.
6. REFERENCES
[1] Roulet N. T., “Surface Level and Water Table Fluctua- tions in a Subarctic Fen,” Arctic and Alpine Research, vol. 23, no. 3, pp. 303, 1991.
[2] Dise N. B., “Peatland Response to Global Change,” Sci- ence, vol. 326, no. 5954, pp. 810–811, 2009.
[3] Lees K. J., Quaife T., Artz R. R. E., Khomik M., and Clark J. M., “Potential for using remote sensing to es- timate carbon uxes across northern peatlands – A re- view,” Science of The Total Environment, vol. 615, pp. 857–874, 2018.
[4] Morton P. A. and Heinemeyer A., “Bog breathing: the extent of peat shrinkage and expansion on blanket bogs in relation to water table, heather management and dom- inant vegetation and its implications for carbon stock as- sessments,” Wetlands Ecology and Management, vol. 27, no. 4, pp. 467–482, 2019.
[5] Fritz C., “Surface oscillation in peatlands: How variable and important is it?,” M.S. thesis, The University of Waikato, New Zealand, 2006.
[6] Howie S. A. and Hebda R. J., “Bog surface oscilla- tion (mire breathing): A useful measure in raised bog restoration,” Hydrological Processes, vol. 32, no. 11, pp. 1518–1530, 2018.
[7] Glaser P. H., Chanton J. P., P. Morin, Rosenberry D. O., Siegel D. I., Ruud O., Chasar L. I., and Reeve A. S., “Surface Deformations as Indicators of Deep Ebullition Fluxes in a Large Northern Peatland,” Global Biogeo- chemical Cycles, vol. 18, no. 1, 2004.
[8] Alshammari L., Large D., Boyd D., Sowter A., Ander- son R., Andersen R., and Marsh S., “Long-Term Peat- land Condition Assessment via Surface Motion Mon- itoring Using the ISBAS DInSAR Technique over the Flow Country, Scotland,” Remote Sensing, vol. 10, no. 7, pp. 1103, 2018.
[9] Heuff F. M. G. and Hanssen R. F., “Insar Phase Re- duction Using the Remove-Compute-Restore Method,” 2020, pp. 786–789, ISSN: 2153-7003.
[10] Cigna F. and Sowter A., “The relationship between intermittent coherence and precision of ISBAS InSAR ground motion velocities: ERS-1/2 case studies in the UK,” Remote Sensing of Environment, vol. 202, pp. 177–198, 2017.
[11] Ansari H., De Zan F., and Bamler R., “Efcient Phase Estimation for Interferogram Stacks,” IEEE Transac- tions on Geoscience and Remote Sensing, vol. 56, no. 7, pp. 4109–4125, 2018.
[12] Novellino A., Cigna F., Brahmi M., Sowter A., Bateson L., and Marsh S., “Assessing the Feasibility of a Na- tional InSARGround DeformationMap of Great Britain with Sentinel-1,” Geosciences, vol. 7, no. 2, pp. 19, 2017.
[13] Esch C., Köhler J., Gutjahr K., and Schuh W. D., “On the Analysis of the Phase Unwrapping Process in a D- InSAR Stack with Special Focus on the Estimation of a Motion Model,” Remote Sensing, vol. 11, no. 19, pp. 2295, 2019.
[14] Hoyt A. M., Chaussard E., Seppalainen S. S., and Har- vey C. F., “Widespread subsidence and carbon emis- sions across Southeast Asian peatlands,” Nature Geo- science, vol. 13, no. 6, pp. 435–440, 2020.
[15] Tampuu T., Praks J., and A. Kull, “InSAR Coherence for Monitoring Water Table Fluctuations in Northern Peatlands,” in IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, 2020, pp. 4738–4741, ISSN: 2153-7003.
[16] Alshammari L., Boyd D. S., Sowter A., Marshall C., Andersen R., Gilbert P., Marsh S., and Large D. J., “Use of Surface Motion Characteristics Determined by In- SAR to Assess Peatland Condition,” Journal of Geo- physical Research: Biogeosciences, vol. 125, no. 1, pp. e2018JG004953, 2020.
[17] Bradley A. V., Andersen R., Marshall C., Sowter A., and Large D. J., “Identication of typical eco-hydrological behaviours using InSAR allows landscape-scale map- ping of peatland condition,” Earth Surface Dynamics Discussions, pp. 1–28, 2021.
19
Authorized licensed use limited to: University of Tartu Estonia. Downloaded on October 03,2022 at 17:57:39 UTC from IEEE Xplore. Restrictions apply.
Tartu Ülikool
Loodus- ja täppisteaduste valdkond
Ökoloogia ja maateaduste instituut
Geograafia osakond
Magistritöö loodusgeograafias ja maastikuökoloogias (30 EAP)
Veerežiimi häiringute ja ilmastiku mõju hariliku männi (Pinus sylvestris L.)
radiaalsele juurdekasvule Lehtmetsa soo näitel
Kärt Erikson
Juhendajad: Ain Kull
Alar Läänelaid
Kristina Sohar
Tartu 2022
2
Annotatsioon
Veerežiimi häiringute ja ilmastiku mõju hariliku männi (Pinus sylvestris L.) radiaalsele
juurdekasvule Lehtmetsa soo näitel
Magistritöö eesmärgiks oli koostada Lehtmetsa soo rabamändide aastarõngalaiuste kronoloogiad
ning analüüsida nende abil sealsete puude radiaalse juurdekasvu sõltuvust inimtekkelistest
veerežiimi muutustest ja kohalikest ilmastikutingimustest. Töös uuriti Lehtmetsa soo puude
juurekasvu mööda veetaseme gradienti viiel eriilmelisel proovialal. Sarnaselt varasemate töödega
leidis kinnitust, et üksteisele geograafiliselt lähedal asuvad proovialad erinevad üksteisest
märkimisväärselt. Üks prooviala käitus loodusliku alana, kaks olid mõjutatud nii
klimatoloogilistest kui kuivendusega seotud muutustest, ning kaks prooviala olid tugevalt
mõjutatud inimtekkelistest kuivendustest.
Märksõnad: harilik mänd, Pinus sylvestris L., dendrokronoloogia, dendroklimatoloogia, soo
CERS kood: P510 - füüsiline geograafia, geomorfoloogia, mullateadus, kartograafia,
klimatoloogia
Abstract
Effect of water level change and weather on radial increment of Scots pine
(Pinus Sylvestris L.) in Lehtmetsa Bog
The aim of this Master’s thesis was to create tree-ring chronologies of Scots pine from Lehtmetsa
Bog and to analyse the influence of anthropogenic changes in the water level and climatic variables
on tree-ring growth. Five diverse study sites along a gradient of water table from Lehtmetsa Bog
were used for this study. Compared with previous studies, the thesis confirmed that geographically
close sample plots differed significantly from each other. One sample plot from five behaved like
a natural area, two were influenced by both climate and by changes in the water level in the nearby
peat extraction site or the nearby Lake Kivijärv. And two were mostly influenced by the changes
in water level in the nearby peat extraction site.
Keywords: Scots pine, Pinus sylvestris L., dendrochronology, dendroclimatology, bog
CERS code: P510 - Physical geography, geomorphology, pedology, cartography, climatology
3
Sisukord
1.Teoreetiline sissejuhatus .............................................................................................................. 4
2. Materjal ja metoodika ................................................................................................................. 8
2.1 Uurimisala ............................................................................................................................. 8
2.2 Kliimaandmed ....................................................................................................................... 9
2.3 Puiduproovide kogumine ja mõõtmine ............................................................................... 11
2.4 Andmeanalüüs ..................................................................................................................... 13
3. Tulemused ................................................................................................................................. 15
3.1 Juurdekasvu kronoloogiad ja kuivenduse mõju .................................................................. 15
3.2 Ilmastiku mõju..................................................................................................................... 17
3.2.1 Temperatuuri mõju puude juurdekasvule ..................................................................... 17
3.2.2 Sademete mõju puude juurdekasvule ........................................................................... 19
3.2.3 Ilmastiku mõju puude juurdekasvule enne ja peale kuivendust ................................... 22
3.2.4 Põua ja sademete rohkuse mõju puude juurdekasvule ................................................. 24
4. Arutelu ja järeldused ................................................................................................................. 25
5. Kokkuvõte ................................................................................................................................. 29
6. Summary ................................................................................................................................... 31
Tänuavaldused .............................................................................................................................. 33
Kirjandus ....................................................................................................................................... 34
Lisad .............................................................................................................................................. 39
4
1. Teoreetiline sissejuhatus
Märgalad on olulised bioloogilise mitmekesisuse kandjad ning paljude taime- ja loomaliikide jaoks
ainsad elupaigad (Edvardsson et al. 2019). IMCG (The Global Peatland Database of the
International Mire Conservation Group) hinnangute järgi katavad turbaalad umbes 3%
(Edvardsson et al. 2016) maailma maismaapinnast ning 80% märgaladest asuvad põhjapoolses
parasvöötmes või külmas kliimas (Punttila et al. 2016). Balti riigid, sealhulgas Eesti, on tuntud
oma suhteliselt suure turbaaladega kaetuse poolest võrreldes teiste Euroopa Liidu liikmesriikidega.
Euroopa Liidu keskmine kaetus turbaaladega on hinnanguliselt 2,8% maismaast
(Karofeld et al. 2016). Eestis on aga turvasmuldadega hinnanguliselt kaetud umbes 22,4%
(Orru ja Orru 2008) ja Baltikumis keskmiselt 12,3% (Taminskas et al. 2019). Sooks nimetatakse
turbaalasid, kus turvast jätkuvalt moodustub ja ladestub, ning turbakihi paksus ületab 30
sentimeetrit (Masing 1988, Moen 1995). Turbaalaks nimetatakse kuitahes paksu turbakihiga
kaetud maa-ala, olenemata sellest kas seal turba ladestumine jätkub või mitte
(Paal ja Leibak 2013). Ramsari konventsiooni järgi nimetatakse märgaladeks turbaalasid, soid,
lodusid ja veekogusid (nii looduslikke kui tehislikke, ajutisi või püsivad), mis on küllastunud veega
(seisva, voolava, mageda, riimvee või soolase veega). Märgalade hulka loetakse konventsiooni
järgi ka merealad, kus vee sügavus ei ületa kuute meetrit (sealhulgas saari ja merealasid kus mõõna
ajal ületab veetase kuut meetrit) ja märgalade naabrusesse jäävad kaldaäärseid, rannikualasid ja
saari (Ramsari konventsioon 1971, Paal ja Leibak 2013). Suur osa Euroopa märgaladest ei liigitu
enam soode alla, nii samuti on ka Eestis, kus vaid kolmandik turbaaladest on sood. Hinnanguliselt
katavad Eesti maismaast sood 7–8%. Eelkõige on soode osakaalu vähenemise taga olnud
kuivendamine (Paal ja Leibak 2011, Leibak 2021). Märgalad mängivad olulist rolli süsinikuringes,
akumuleerides orgaanilist süsinikku. Teisalt võivad nad olla aga oluliseks looduslikuks allikaks
metaani, süsihappegaasi ja NOx ühendite puhul (Tamkevičiūtė et al. 2018). Turbaalad talletavad
ligikaudu kolmandiku ülemaailmsest mullasüsinikust (Lucow et al. 2022). Turbaalade
süsinikubilanss sõltub tugevalt turba kogunemise kiirusest ning on seetõttu väga tundlik
turbasambla ja soontaimede vahekorra muutuste suhtes (Edvardsson et al. 2019).
Turba kaevandamise algus Eestis ulatub tagasi 17. sajandisse. 19. sajandil oli turbaalade
kuivendamine või nende põletamine üheks kõige levinumaks põllumajandusega kaasnevaks
teguriks märgaladel (Vasander et al. 2003). Mitmed turbaalade ökosüsteemid on sajandeid olnud
5
tugevasti mõjutatud inimtegevuse poolt. 70% kahjustustest Soome märgalades on tingitud
inimtegevusest, Poolas loetakse inimtegevuse tagajärjel kahju saanud turbaalade suuruseks 80%
(Cedro ja Sotek 2016). Sarnaselt eelmainitud riikidega on hinnanguliselt 70% Eesti märgaladest
tugevalt mõjutatud inimtegevusest (Vasander et al. 2003, Paal ja Leibak 2011). Majanduslikus
mõttes on Baltikumi rabad olnud inimeste huviorbiidis 18. sajandist alates. Märgalasid kuivendati,
et hõlbustada turba kaevandamist või kiirendada metsa kasvu. Need protsessid intensiivistusid 19.
sajandil ning olid kõige laiemalt levinud 20. sajandil. Turba kaevandamise, põlengute või
põllumajandusliku ettevalmistamise tõttu on mitmetes märgalades esialgsed elupaigad täielikult
hävinud (Cedro ja Sotek 2016). Põhilisteks Eestis märgalasid mõjutavateks tegevusteks on
metsandus, põllumajandus ja turbakaevandamine (Vasander et al. 2003, Paal ja Leibak 2011).
Hinnanguliselt on enim kahjustada saanud ja hävinud madalsookooslused, mille varasem osakaal
Eesti sookooslustest oli kolmandik, kuid tänasel päeval moodustavad nad 10–20% soodest Eestis
(Leibak 2021).
Dendroklimatoloogilised uuringud põhinevad teadmisel, et puidu kasvu mõjutab ümbritsev
keskkond ja sarnases keskkonnas kasvavatel puudel on sarnane kasvukiirus.
Keskkonnatingimused nagu päikesevalgus, temperatuur, vesi, toitainete varu, tuul, mehaanilised
toimed ning õhu ja mulla saastatus kõik mõjutavad puu kasvu. Dendrokronoloogilised uuringud
annavad puude aastarõngaste kaudu väärtuslikku informatsiooni uuritava ala inimtegevusest ja
kliimamuutustest tulenevate muutuste kohta (Schweingruber 1996). Põhjapoolsetel laiuskraadidel
ei toimu puudel aasta läbi pidevat kasvu, vaid kasv jälgib iga-aastast neljatsüklilist jaotust, mille
tingivad fotoperioodi pikkus ja termilise režiimi muutused. Nendeks tsükli faasideks on kasvu
taasaktiveerimine, meristeemilise aktiivsuse periood, kasvu aeglustumine ja talvine puhkeperiood
(Sarvas 1972). Tsükli jooksul tekitavad hemiboreaalsetes tingimustes kasvavad puud igal aastal
ühe radiaalse juurdekasvu rõnga (puu aastarõnga) (Schweingruber 1996).
Märgaladel kasvavate puude kasv on tugevalt seotud sealse veetaseme kõrgusega. Kõrgem veetase
põhjustab tavaliselt puude juurdekasvu vähenemist ning madal veetase suuremat puude
juurdekasvu (Linderholm et al. 2002, Edvardsson et al. 2016, Tamkevičiūtė et al. 2018). Samuti
sõltub puude kasv klimaatilistest tingimustest ja antropoloogilistest teguritest
(Cedro ja Lamentowicz 2008). Puude aastarõngalaiused on head hindamaks limiteerivate tegurite
mõju puu kasvule (Fritts 1976). Kui veetase on kõrge, on väiksem puudele kättesaadav hapniku
hulk ja puudel on oht hapnikuvaeguseks ehk hüpoksiaks või hapnikupuuduseks ehk anoksiaks.
6
Kui veetase on aga madal, siis võivad rabas kasvavate mändide pinnalähedased juured olla altid
põua poolt tekitatavale kahjule (Braekke 1983, Dang ja Lieffers 1989, Pepin et al. 2002). Seega
seni, kuni on võimalik eristada madala veetaseme mõjudest tingitud häiringuid kõrge veetaseme
omast, annavad rabamändide aastarõngalaiuste read hea ülevaate märgala veetaseme ajaloost
(Smiljanić et al. 2014, Läänelaid et al. 2014). Varasemad uuringud on näidanud, et turbaalade
puude radiaalne juurdekasv sõltub niiskustingimustest veega küllastumata alal. Seeläbi juhivad
turbaalade veetaseme kõikumised puu juurdekasvumustrite teket (Tamkevičiūtė et al. 2018).
Varasemad dendrokronoloogilised uuringud rabamändidega on näidanud, et rabamändide
keskmisi aastaseid juurdekasvuridasid saab kasutada märgaladel veetaseme muutuste uurimiseks
(Smiljanić et al. 2014). Need tööd toovad aga välja, et puude aastarõngaste kasutamisel
veetingimuste uurimiseks esineb viivitusi puu reageeringus. Need on tingitud taimede
füsioloogilistest protsessidest või teistest mitte-hüdroloogilistest piiravatest teguritest, näiteks
temperatuurist või sademetehulgast. Kuigi need viivitused võivad segada aasta täpsusega
tulemuste saamist, kasutatakse aastarõngaste laiust siiski tihti hüdroloogiliste tingimuste
uurimiseks (Edvardsson et al. 2019).
Rabas kasvavatel puudel on mitmeid omadusi, mis raskendavad nende kasutamist
dendrokronoloogilistes uuringutes. Märgalades kui liigniisketes, madala toitainesisaldusega ja
halva mullahingamisega aladel võivad puude eluead olla lühemad kui kuivematel aladel. Lisaks
esineb tihti väga kitsaid rõngaid ning osaliselt või täielikult puuduvaid rõngaid
(Dauškane et al. 2011, Smiljanić et al. 2014). Varasemad uuringud toovad samuti välja, et
kuivendamine mõjub positiivselt puude kasvule ja seeläbi on suuremad ka keskmised
aastarõngalaiused (Macdonald ja Fengyou 2001, Choi et al. 2007).
Harilik mänd (Pinus sylvestris L.) on üks levinumaid puuliike Kesk- ja Ida-Euroopas. Teda leidub
ka Lääne-Euroopa mägismaadel ning tema levila katab Põhja-Euroopa kuni Siberi idaosani
(Gardner 2013). Samuti on ta üheks levinumaks puuliigiks Eestis ja kasutatud
dendrokronoloogilistes uuringutest (Läänelaid ja Eckstein 2003, Metslaid 2017) Tegemist on
Eestile pärismaise igihalja okaspuuga, mis kuulub männiliste sugukonda ja perekonda mänd (e-
elurikkus). Kogu Eesti metsamaast katab harilik mänd pisut üle 34%. Kuna mänd on mulla
toitainesisalduse suhtes vähenõudlik, on ta suuteline kasvama väga erinevates tingimustes. Harilik
mänd on võimeline kasvama nii liivastel ja savistel muldadel kui ka liigniisketes rabakooslustes.
7
Lisaks talub ta hästi nii põuda kui ka pakast. Harilik mänd on aga väga valgusnõudlik. Soostunud
aladel kasvab umbes 15% kõikides Eestis asuvatest männikutest ning soodes pisut alla 20%
männikutest. Kõige rohkem leidub männimetsi Põlva- ja Harjumaal ning kõige vähem Jõgeva- ja
Tartumaal (Sibul 2014).
Harilik mänd on lülipuiduline. Lülipuit on puutüve keskmises osas asuv surnud rakkudest koosnev
puidu osa, mis ei võta osa vedelike transpordist. Puid, millel selline kiht esineb nimetatakse
lülipuidulisteks. Lülipuit on kollakas- või punakaspruun ning see hakkab tekkima peale puu 40.
eluaastat. Männi maltspuit on heledama värvusega kui lülipuit, ning see koosneb vedelikke
juhtivatest elusrakkudest. Hariliku männi puhul on tegemist kiirekasvulise puuliigiga. Heades
kasvutingimustes võib puu juba 70-aastaselt ületada 30 meetri kõrguse. Rabades võib männi
kõrgus aga jääda kuni 2 meetrini. Maksimaalne eluiga jääb männil 400 ja 500 aasta vahele.
Aastarõngad harilikul männil on hästi nähtavad (Sibul 2014).
Käesoleva magistritöö eesmärkideks on:
1. Hinnata inimtegevusest tingitud veetaseme häiringute mõju hariliku männi radiaalsele
juurdekasvule endisel turbakaevandusalal. Vaatluse all on puude aastane juurdekasv nii
looduslikus seisundis kui raba kuivendamisel ja turba kaevandamisel, mahajätmisel ja
taastamisel.
2. Analüüsida ilmastikutingimuste (temperatuur, sademed) mõju rabamändide radiaalsele
juurdekasvule.
Tööd alustades on püstitatud järgnevad hüpoteesid:
1. Inimtegevusest tingitud kuivenduse poolt on tugevasti mõjutatud kaevandusele
geograafiliselt lähimal asuvad proovialad üks ja kaks.
2. Looduslike aladena käituvad Lehtmetsa soos keskosas asuvad proovialad kolm ja neli.
3. Viies, kõige Kivijärve poolseim prooviala, on samuti tugevalt mõjutatud inimtegevusest,
kuid mõjud tulenevad veetaseme muutustest järves.
8
2. Materjal ja metoodika
2.1 Uurimisala
Uuritav ala asub Jõgevamaal Jõgeva vallas Laiusevälja külas Lehtmetsa soos (joonis 1). Jääksoo
paikneb voortevahelises nõos, mida võib mõjutada surveline põhjavesi (Lode et al. 2015). Soo
läänepiiriks on Laiuse mägi (145 meetri kõrgune voor) ja põhjaserv ulatub Jõgeva-Mustvee
maanteeni (Eesti Turbauuringute Andmebaas. Kivijärve turbaala). Lehtmetsa soo alal asub
tervikuna mahajäetud freesturba tootmise ala. Piirkonnast kaevandati freesalusturvast aastatel
1969–1996. Kaevandamine lõpetati raskete kuivendamistingimuste ja vähelagunenud turba kihi
ammendumise tõttu (Ramst et al. 2006). Endine kaevandusala asub kahe plokina raba
põhjaosas – põhja- ja lõunapoolse plokina. Põhjapoolse ploki suuruseks on 15 hektarit ja
kuivendusdreenid kulgevad lõuna-põhjasuunaliselt suundudes põhjaservas asuvasse kogujakraavi.
Lõunapoolse ploki suuruseks on 17 hektarit ja ala kuivendusdreenid kulgevad põhja-
lõunasuunaliselt suubudes ala lõunaservas paiknevasse kogujakraavi. Kahe ploki vahelt kulgeb
väljaveotee, mis on ümbritsevast ca 1 meetri võrra kõrgem (Lode et al. 2015).
Joonis 1. Lehtmetsa soo – mändide radiaalkasvu uurimisala (Maa-ameti kaardi ja ortofoto
hübriid).
9
2005. aastast alates on ala arvatud passiivse tarbevaru hulka. Ala suuruseks on keskkonnaregistri
andmetel 31,91 hektarit. Passiivse turbavaru koguhulk on 168 000 tonni, millest 165 000 tonni on
hästilagunenud turvast ning 3000 tonni on vähelagunenud turvast. Hästilagunenud kihi keskmiseks
paksuseks on 2,8 meetrit ja vähelagunenud turbakihi keskmiseks paksuseks on pool meetrit.
Lehtmetsa soo hästilagunenud turba keskmine lagunemisaste on 31% ja keskmine tuhasus on 5,9%
ning sealse vähelagunenud turba keskmine lagunemisaste on 6%, keskmine tuhasus 1,8%.
Turbakihid asuvad kuni poole meetri paksusel järvelubjakihil, mis omakorda lasub jääjärvelistel
liivadel ja moreenil (Ramst et al. 2006).
Lehtmetsa soo on väga eriilmeline. Endisel kaevandamise alal kasvavad turbavallidel sookased
(Betula pubescens), mille kõrguseks on 1–2 meetrit. Samblarinde kogukatvusega 60–70%
moodustavad raba-karusammal (Polytrichum strictum), nõtke karusammal
(Polytrichum longisetum), longus pirnik (Pohlia nutans), pugu-kaksikhambake
(Dicranella cerviculata), hammas-karviksammal (Pohlia bulbifera) ja väävel-porosamblik
(Cladonia deformis). Samuti leidub endisel kaevandusalal Eestis üliharuldast saagjat kübesammalt
(Ephemerum serratum) (Ramst et al. 2006).
2013. aastal esines endisel kaevanduse alal üleujutus seoses kobraste paisutustegevusega. Alates
2019. aastast tõsteti endisel kaevanduse alal veetaset märgala taastamise eesmärgil.
Uuritava ala lõunapiiriks on Kivijärv. Kivijärve puhul on tegemist endise voortevahelise järvega,
millest tänaseks päevaks on säilinud ainult järve sügavaim osa (kunagine järve suurus on olnud
hinnanguliselt 500 hektarit). Järve pindala on praeguseks 19 hektarit ning keskmine sügavus 1,5
meetrit. Järve suurim sügavus ulatub 2 meetrini. Kivijärv kuulub kihistumata kalgiveeliste
segatoiteliste järvede hulka. Aastatel 1928–1929 alandati järve veetaset ühe meetri võrra, mis
kiirendas oluliselt järve pindala vähenemist. Aastal 1951 oli järve taimestik liigirikas ja taimestik
kattis pea kogu järve (Mäemets 1977). Teine katse järve veetaset alandada toimus 1973. aastal
(Maaparandussüsteemide register). Järve kalastik on pigem liigivaene, kuid järve kasutavad
pesitsuspaigana mitmed veelinnud (Eesti Entsüklopeedia).
2.2 Kliimaandmed
Jõgevamaa kuulub Mandri-Eesti kliimavaldkonna Sise-Eesti allvaldkonda. Uurimisalale lähimaks
ilmajaamaks on Jõgeva meteoroloogiajaam (N 58°44´59´´ E 26°24´54´´ H= 70,29 m), mis asub
10
uuritavast alast edelas kaheksa kilomeetri kaugusel. Jõgeva ilmajaama asukoht on aja jooksul
muutunud. Kui ajavahemikul 1922–1964 asus meteoroloogiajaama Jõgeva alevikus sees, siis
hiljem kolis see Jõgeva idaservale (Riigi Ilmateenistus. Jõgevamaa kliimast). Töös kasutati
dendroklimatoloogiliseks analüüsiks Jõgeva ilmajaama temperatuuri (1922–2019) ja sademete
mõõtmisandmeid (1945–2019). Üksikuid puuduvaid vaatlusandmeid temperatuuri
aegreas interpoleeriti lineaarse regressioonimudeli järgi Tartu ilmajaama andmetest.
Lisaks kasutati töös professor Jaak Jaaguselt saadud Jõgeva meteoroloogiajaama andmetel
arvutatud standardiseeritud sademete aurustumise indeksit (SPEIP). Kasutatud indeksid olid
ajaperioodi 1948–2018 kohta (Jaagus et al. 2021).
Ajavahemiku 1945–2019 keskmine õhutemperatuur Jõgevamaal on 5 °C ja aasta keskmine
sademete hulk on 655 mm. Klimaatiline kevad algab Jõgeval keskmiselt 21. aprillil. Sügise
keskmiseks alguskuupäevaks on 2. september (Riigi Ilmateenistus. Jõgevamaa kliimast).
Kõrgeima keskmise temperatuuriga kuuks on juuli (keskmine temperatuur 16,8 °C) ning
külmimaks kuuks on veebruar (keskmine temperatuur –6,4 °C) (joonis 2). Ajaperioodil 1991–
2020 algas klimaatiline kevad Jõgeval keskmiselt 21. aprillil ja sügise keskmiseks
alguskuupäevaks on 2. september (Riigi Ilmateenistus. Jõgevamaa kliimast).
Joonis 2. Uuritava ala lähima Jõgeva meteoroloogijaama kuukeskmised temperatuurid (1922–
2019) ja sademetehulgad (1945–2019).
11
2.3 Puiduproovide kogumine ja mõõtmine
2019. aasta oktoobris koguti puidu puurproove viielt proovialalt Lehtmetsa soos. Uurimisala jäi
endise kaevandusala ja Kivijärve vahele (joonis 3). Puurproovid koguti juurdekasvupuuriga
kasvavatest harilikest mändidest. Sama liigi sealsetest kändudest saeti ristlõikekettad. Proovialad
paiknevad mööda vee gradienti endisest kaevandusalast Kivijärveni. Ristlõikekettaid kändudest
koguti ainult esimeselt proovialalt, ülejäänud proovialadelt koguti mändidelt ainult
puursüdamikke. Igalt proovialalt koguti vähemalt 12 puiduproovi ning kõik proovid välja arvatud
ristlõikekettad kändudest võeti ca 30 sentimeetri kõrguselt maapinnast. Kännuproovid saeti
madalamalt, vastavalt kännu kõrgusele. Juurdekasvupuuriga puuriti radiaalselt läbi säsi, et saada
ühe puurimisega tüvest kaks vastasraadiust.
Joonis 3. Proovialade paiknemine piki veegradienti Lehtmetsa soos (aluseks Maa-ameti ortofoto).
Esimene prooviala asus endisel kaevandusalal ning sealselt alalt koguti proove nii kändudest kui
elavatest puudest. Proovid võeti seitsmest elusast männist ja kümnest männikännust. Teine
prooviala asus endist kaevandusala ümbritseva kuivenduskraavi kõrval. Proove koguti sealsetest
mändidest 13. Kolmas prooviala asus 150 meetrit endist kaevandusala ümbritsevast
kuivenduskraavist Kivijärve poole. Proove saadi sealt kokku 15 männist. Kolm esimest prooviala
12
on eelduste kohaselt kõige rohkem mõjutatud veetaseme muutustest endisel freesturbakaevanduse
alal. Neljas prooviala asus uurimisala kõige puutumatumas osas. Viies prooviala asus raba
lõunaosas Kivijärve läheduses. Eeldatavasti ei ole see piirkond mõjutatud veetaseme muutustest
endisel freesturbaväljal, vaid seda prooviala mõjutavad veetaseme muutused järves. Neljandalt ja
viiendalt proovialalt koguti mõlemalt 12 proovi.
Kogutud proovid mõõdeti kasutades mõõtmisaparaati LINTAB ja Leica S4E mikroskoopi. Enne
mõõtmist lõigati puurproovi või kännu ristlõikeketta pind žiletiga risti puidukiude siledaks ja
hõõruti parema nähtavuse saavutamiseks kriidiga. Iga puurproov ja mõõdetud tüveraadius sai
unikaalse koodi. Mõõtmistulemused salvestati programmis TSAP-Win™ (Rinn 2003).
Aastarõngalaiused mõõdeti mikromeetrites (µm), kuna valdavalt liigniiskete tingimuste tõttu on
aastarõngad sageli väga kitsad. Iga proov mõõdeti kahest vastasraadiusest koorest säsini.
Vastasraadiustest mõõdetud juurdekasvuridade ühtimisel need keskmistati puu kronoloogiaks.
Proovialade keskmised aastarõngalaiuste kronoloogiad on saadud valdavalt (mõne erandiga)
nende keskmistatud raadiuste aastarõngalaiuste ridade keskmistamisel. Kui männi vastasraadiused
omavahel ei sünkroniseerunud, kuid ühe raadiuse aastarõngarida ühtis hästi prooviala teiste puude
aastarõngakronoloogiatega, kasutati ala üldkeskmises männi ühe raadiuse andmeid. Rabamändidel
esineb niinimetatud puuduvaid aastarõngaid. Tegemist on nähtusega, kus aastarõngas on olemas
ühes raadiuses, kuid puudub sama puu vastasraadiuses. Taolisi olukordi on võimalik tuvastada
vastasraadiuste omavahelisel või erinevate puurproovide juurdekasvuridade võrdlemisel
joongraafikutel. Kõige tõenäolisematesse puuduvate rõngaste asukohtadesse lisati programmis
TSAP-Win™ lisa-aastarõngalaius väärtusega 1. Mõõdetud proovide omavahelist sarnasust
kontrolliti programmiga COFECHA (Holmes 1983, Grissino-Mayer 2001). See on spetsiaalne
arvutiprogramm, mida kasutatakse hindamaks aastarõngalaiuste mõõtmistulemuste täpsust ja
ristdateeringu kvaliteeti.
Programmis TSAP-Win™ (Rinn 2003) leiti individuaalsete puude juurdekasvukronoloogiate ja
keskmiste juurdekasvukronoloogiate omavahelised statistilised sarnasusnäitajad: Pearsoni
korrelatsioonikordaja, dendrokronoloogias kasutatav sarnasusnäitaja tBP-väärtus
(Baillie ja Pilcher 1973) ning samasuunaliste kõikumiste protsent Glk (Gleichläufigkeit)
(Eckstein ja Bauch 1969). Usaldusväärselt sarnaseks loetakse aastarõngakronoloogiaid, mille
omavaheline sarnasusnäitaja tBP > 3,5 (tabelis märgitud kui TVBP, lisa 1).
13
2.4 Andmeanalüüs
Programmis ARSTAN (Cook ja Holmes 1986) loodi kronoloogiatest standardiseeritud ja mürast
puhastatud kronoloogiad. Standardiseerimine on klassikaline dendrokronoloogia meetod, mille
eesmärgiks on vähendada endogeensete häiringute ja puu vanuse mõju juurdekasvu ridadele.
Programmiga ARSTAN eemaldatakse mõõdetud juurdekasvuridadest trendid – iga kasvuaasta
reaalne väärtus jagatakse vastava negatiivse eksponentsiaal- või lineaarfunktsiooni trendijoone
väärtusega. Protsessi käigus eemaldatakse aegridadest autokorrelatsioon, mis on eelduseks
dendroklimaatilisele korrelatsioonanalüüsile. Tulemuseks on standardiseeritud
aastarõngakronoloogia, kus aastarõnga laiused on suhtelistes ühikutes ning keskväärtusega
1 (Cook ja Kairiukstis 1990) (lisa 2).
Aastarõngalaiuste kronoloogia ja kliimaandmete (temperatuuri ja sademetehulga) seoste
uurimiseks kasutati DENDROCLIM2002 programmi (Biondi ja Waikul 2004).
Aastarõngalaiustega võrdlemise perioodi alguseks võeti kasvuperioodile eelneva aasta oktoober
ning lõpuks kasvuperioodi aasta september. Seega vaadeldava perioodi pikkuseks oli 12 kuud.
Dendroklimatoloogilises analüüsis kasutati Pearsoni korrelatsioonikordajat (r), mis leiti
aastarõngaste kronoloogia ning määratud kalendrikuude sademetehulkade ja keskmiste
temperatuuride vahel. Korrelatsioonikordajate usalduspiirid (p < 0,05) leiti bootstrap-meetodil.
Kliimaandmetega võrreldi programmis DENDROCLIM2002 nii standardiseeritud
aastarõngalaiuste kronoloogiaid kui ka standardiseerimata kronoloogiaid. Programmiga
DENDROCLIM2002 jälgiti ka korrelatsiooni muutumist ajas, milleks kasutati 24-aastast aja-akent
1-aastase sammuga. Lisaks analüüsiti programmiga DENDROCLIM2002 standardiseeritud
kronoloogiaid kahes osas. Vaadeldavateks perioodideks olid aastarõngaste kronoloogiate
algusaastad (temperatuuriandmete puhul oli alguseks 1923. aasta ja sademete puhul 1946. aasta)
kuni 1970. aasta ja 1971–2019. aasta. Sellise kahes osas analüüsi eesmärgiks oli näha kuivendusest
tingitud erinevusi, kuna perioodi esimene pool jääb turbakaevanduse eelsesse aega ja teine
perioodi, mil proovialade läheduses on toimunud kuivendust.
Lisaks korreleeriti standardiseeritud kronoloogiaid uuritava ala lähima ilmajaama liidetud kuude
ilmastikuandmetega. Viimasteks oli märtsi ja aprilli, mai ja juuni ning augusti ja septembri
vastavad keskmised temperatuurid ning sademete summad.
14
Viimaseks võrreldi keskmistatud alade aastarõngakronoloogiaid Jõgeva ilmajaama sademete ja
õhutemperatuuri andmetel arvutatud standardiseeritud sademete aurustumise indeksitega ehk
SPEIP indeksitega (Standardized Precipitation and Evapotranspiration Indexi Penman-Monteith
meetod). Antud indeksi kasutatakse sageli põua uuringutes ja nende iseloomustamiseks, kuna
indeks põhineb temperatuuri ja sademete andmetel. Negatiivsed SPEIP väärtused kirjeldavad
keskmisest põuasemat olukorda ja positiivsed liigniiskemat olukorda (Vicente-Serrano et
al., 2010, Metslaid 2019, Jaagus et al. 2021).
15
3. Tulemused
3.1 Juurdekasvu kronoloogiad ja kuivenduse mõju
Esimeselt proovialalt saadud seitsme puu juurdekasvuread sünkroniseerusid omavahel ning
keskmise kronoloogia kogupikkuseks on 15 aastat. Sealsel proovialal on puud hakanud kasvama
peale turbakaevandamise lõppu 1990. aastatel looduslikult (ilma istutamata). Keskmine
aastarõngalaius esimesel proovialal on 3,37 mm (joonis 4a). Esimeselt proovialalt kogutud
kännuproovide juurdekasvuridu ei õnnestunud omavahel sünkroniseerida ja sellepärast neid
edaspidises analüüsis ei kasutatud.
Teise prooviala keskmistatud kronoloogia koosneb seitsme puurproovi aastarõngaste
juurdekasvureast ja on 43 aasta pikkune. Keskmine aastarõngalaius teisel proovialal on 2,06 mm.
Antud proovialas on keskmine aastarõngalaius vähenenud alates 1990. aastatest (joonis 4b).
Kolmandalt proovialalt kogutud seitsme omavahel sünkroniseeritud ja seejärel keskmistatud
kronoloogia pikkuseks on 172 aastat. Keskmine aastarõngalaius sealsel proovialal on 0,61 mm.
Kolmandal proovialal kasvavate puude juures on iseloomulikuks aastarõngalaiuste suurenemine
alates 1970. aastate keskpaigast 1980. aastate keskpaigani (joonis 4c).
Neljandalt proovialalt sünkroniseeriti ja keskmistati üheksa puu puurproovi juurdekasvuread.
Saadud kronoloogia pikkuseks on 157 aastat ja keskmine aastarõngalaius neljandal proovialal on
0,47 mm. Antud proovialal on näha keskmiste aastarõngalaiuste suurenemist ajavahemikus
2011 – 2015 (joonis 4d).
Viiendalt proovialalt sünkroniseerusid omavahel üheteistkümne puu puurproovi juurdekasvuread,
mis seejärel keskmistati. Saadud kronoloogia pikkuseks on 115 aastat ja keskmine aastarõngalaius
on 1,31 mm. Üldkronoloogial on näha keskmiste aastarõngalaiuste kasvu 1930. ja 1970. aastatel,
mille järel on näha stabiilset aastarõngalaiuste kahanemist. 1980. ja 1990. aastatel esinevad selged
lühiajalised suuremate aastarõngalaiustega perioodid (joonis 4e).
16
Joonis 4. Proovialade aastarõngaste kronoloogiad (tumedad jooned) ning individuaalsete puude
juurdekasvuread (heledad jooned).
17
3.2 Ilmastiku mõju
3.2.1 Temperatuuri mõju puude juurdekasvule
Esimese prooviala kohta dendroklimaatilist analüüsi temperatuuriandmetega sealse kronoloogia
lühiduse (15 aastat) tõttu läbi ei viidud.
Teise prooviala dendroklimaatilise analüüsi ajalise ulatuse 1977–2019 määras sealse prooviala
aastarõngakronoloogia pikkus. Temperatuuriandmetega võrreldi nii standardiseeritud kui
standardiseerimata aastarõnga kronoloogiaid. Nii standardiseerimata (novembri r = –0,334,
detsembri r = –0,456) kui standardiseeritud (novembri r = –0,311, detsembri r = –0,327)
aastarõngakronoloogiatega ilmnesid statistiliselt olulised negatiivsed korrelatsioonid kasvuaastale
eelneva novembri ja detsembri temperatuuridega. Samuti tuli korrelatsioonianalüüsist välja
statistiliselt oluline seos kasvuaasta veebruari temperatuuriga (r = –0,334). Standardiseerimata
aastarõngakronoloogiat analüüsides tulid esile statistiliselt olulised negatiivsed korrelatsioonid
kasvuaasta augusti ja septembri temperatuuridega (august r = –0,329, sept r = –0,489) (joonis 5a).
Kolmanda prooviala dendroklimaatiline korrelatsioonianalüüs katab ajaperioodi 1923–2019.
Statistiliselt olulised ja positiivsed seosed esinevad standardiseerimata juurdekasvu ning jaanuari
(r = 0,218), märtsi (r = 0,234) ja aprilli (r = 0,219) temperatuuridega. Standardiseeritud
kronoloogia kliimaandmetega võrdlemisel kaovad aga nende kuude mõjud ja esile tulevad
statistiliselt olulised negatiivsed korrelatsioonid augusti (r = –0,207) ja septembri (r = –0,192)
temperatuuridega (joonis 5b).
Neljanda prooviala dendroklimaatiline korrelatsioonanalüüs katab perioodi 1923–2019.
Standardiseeritud kronoloogia võrdlemisel temperatuuriandmetega ei esinenud ühegi kuu
keskmise temperatuuriga statistiliselt olulist seost. Standardiseerimata kronoloogia ja kuu
keskmiste temperatuuride vahel ilmnesid statistiliselt olulised negatiivsed korrelatsioonid jaanuari
(r = –0,264), augusti (r = –0,280) ja mai (r = –0,193) temperatuuridega (joonis 5c).
Viienda prooviala dendroklimaatiline korrelatsioonianalüüs katab perioodi 1923–2019.
Standardiseerimata kronoloogia ja temperatuuriandmete analüüsil tulevad esile statistiliselt
olulised negatiivsed korrelatsioonid aprilli (r = –0,307) ja mai (r = –0,191) temperatuuridega.
Standardiseeritud kronoloogia ja temperatuuriandmeid korreleerides olid statistiliselt olulise
18
positiivse väärtusega märtsi (r = –0,277) ja juuli (r = 0,217) temperatuurid (joonis 5d). Libiseva
standardiseeritud aastarõngakronoloogia korrelatsioonanalüüsist võib näha, et märtsi ja juuli
statistiliselt olulised tulemused on aja jooksul oma olulisuse kaotanud (lisa 5).
Lisaks arvutati liidetud kuude (märts–aprill, mai–juuni, august–september) keskmise temperatuuri
ja eri proovialade keskmiste juurdekasvuridade omavaheline korrelatsioon. Liidetud kuude
tulemustest tuli esile statistiliselt oluline negatiivne korrelatsioon kolmanda prooviala juurdekasvu
ning augusti–septembri keskmise temperatuuriga (r = –0,246) (joonis 6).
Joonis 5. Proovialade standardiseeritud ja standardiseerimata aastarõngakronoloogiate ja kuu
keskmiste temperatuuride omavaheline korrelatsioon kasvuaastale eelnenud oktoobrist
kasvuaasta septembrini. Täidetud tulbad näitavad statistiliselt olulisi seoseid (p < 0,05).
19
Joonis 6. Liidetud kuude keskmiste temperatuuride ja rabamändide aastarõngalaiuste
omavaheline korrelatsioon. Täidetud tulbad näitavad statistiliselt olulisi seoseid (p < 0,05).
3.2.2 Sademete mõju puude juurdekasvule
Esimese prooviala kohta dendroklimaatilist analüüsi sademete andmetega aastarõngalaiuste
kronoloogia lühiduse (15 aastat) tõttu läbi ei viidud.
Teise prooviala dendroklimaatilise korrelatsioonianalüüsi sademetega ajalise ulatuse määras
sealse prooviala aastarõngakronoloogia pikkus 1977–2019. Standardiseeritud
aastarõngakronoloogiaga ja sademete vahel esinenud ühtegi statistiliselt olulist seost.
Standardiseerimata kronoloogiaga aga esinesid statistiliselt olulised positiivsed seosed juuli
(r = 0,407) ja septembri (r = 0,306) sademetega (joonis 7a).
Kolmanda prooviala dendroklimaatilise korrelatsioonianalüüsi ajaline ulatus oli 1946–2019.
Standardiseerimata kronoloogia ja sademeterea korreleerimisel tuvastati statistiliselt olulised
positiivsed seosed jaanuaris (r = 0,289) ja märtsis (r = 0,234). Standardiseeritud kronoloogiaga
analüüsides tuli esile statistiliselt oluline negatiivne seos juuli (r = 0,272) sademetega. Nii
standardiseerimata kui standardiseeritud kronoloogia ning aprilli sademete vahel ilmnes
statistiliselt oluline negatiivne seos (vastavalt r = –0,230 ja r = –0,169) (joonis 7b).
Neljanda prooviala dendroklimaatilise korrelatsioonianalüüsi sademete andmetega ajaline ulatus
oli 1946–2019. Standardiseeritud kronoloogia ja sademete vahel ilmnes statistiliselt oluline
positiivne korrelatsioon juunikuus (r = 0,255). Nii standardiseeritud kui standardiseerimata
kronoloogia ja aprilli sademete vahel esines positiivne ja statistiliselt oluline korrelatsioon
20
(vastavalt r = 0,247 ning r = 0,246) (joonis 7c). Neljanda prooviala libiseva korrelatsioonanalüüsi
(lisa 4) tulemusi vaadates võib näha, et ajas on nõrgenemas positiivne seos standardiseeritud
juurdekasvu ja aprilli sademete vahel ning on tugevnenud seos juuni sademetega.
Viienda prooviala dendroklimaatilise korrelatsioonianalüüsi ajaline ulatus oli 1946–2019.
Standardiseerimata krononoloogia analüüsil ilmnes statistiliselt oluline ja negatiivne seos jaanuari
(r = –0,277) sademetega ja statistiliselt oluline positiivne seos aprilli (r = 0,217) sademetega.
Standardiseeritud aastarõngakronoloogia võrdlusel sademetega tulid tugevalt esile positiivsed
statistiliselt olulised seosed juunis (r = 0,556) ja augustis (r = 0,410) (joonis 7d).
Libisevat standardiseerimata kronoloogia korrelatsioonanalüüsist (lisa 5) on näha, et positiivne
statistiliselt oluline seos juuni sademetega on olnud ajas püsiv ning sama seos augusti sademetega
on saanud statistiliselt oluliseks pigem vaatlusperioodi lõpuosas.
Liidetud kuu sademete ja eri proovialade keskmiste juurdekasvuridade analüüsi tulemuseteks olid
statistiliselt olulised positiivsed korrelatsioonid neljandal ja viiendal proovialal märtsi ja aprilli
sademete (vastavalt r = 0,237 ja r = 0,265) ning mai ja juuni sademetega (vastavalt r = 0,238, ja
r = 0,489). Samuti on näha statistiliselt olulist positiivset korrelatsiooni viienda prooviala
keskmiste aastarõngalaiuste ning augusti ja septembri liidetud sademete hulga vahel (r = 0,407)
(joonis 8).
21
Joonis 7. Proovialade standardiseeritud ja standardiseerimata aastarõngakronoloogiate ja kuude
sademetehulga korrelatsioon kasvuaastale eelnenud oktoobrist kasvuaasta septembrini. Täidetud
tulbad tähistavad statistiliselt olulisi seoseid (p < 0,05).
Joonis 8. Liidetud kuude keskmiste sademetehulkade ja aastarõngalaiuste korrelatsioon. Täidetud
tulbad näitavad statistiliselt olulisi seoseid (p < 0,05).
22
3.2.3 Ilmastiku mõju puude juurdekasvule enne ja peale kuivendust
Programmis DENDROCLIM2002 korreleeriti kolmanda, neljanda ja viienda prooviala
aastarõngalaiuste kronoloogiaid temperatuuri ning sademete andmetega enne ja peale kuivendust
ehk perioodidel 1923(T)/1946(S)–1970 ning 1971–2019. Seejärel võrreldi saadud korrelatsioone
omavahel.
Tulemustest on näha, et kuivenduse tõttu on kolmandal proovialal juuli ja augusti temperatuuride
mõju muutunud varasemast statistiliselt oluliseks ja negatiivseks (enne kuivendust juuli r = 0,121
ja peale kuivendust r = –0,233; enne kuivendust augusti r = –0,150 ja peale kuivendust r = –0,365).
Samuti on kuivenduse mõjul statistiliselt oluliseks saanud juuli sademete mõju (enne r = 0,273 ja
peale r = 0,331) (joonis 9a).
Neljanda prooviala tulemustest on näha, et enne kuivenduse algust on statistiliselt olulist
negatiivset mõju avaldanud kasvuaastale eelneva novembri ja detsembri temperatuurid (vastavalt
r = –0,343 ja r = –0,265). Peale kuivendust on korrelatsioonid nõrgalt positiivsed, kuid statistiliselt
ebaolulised. Samuti on kuivendamine vähendanud märtsi sademete mõju (enne kuivendust
r = 0,347; peale kuivendust r = 0,0438). Samuti on kuivendamise järgselt vähenenud teiste
kevadkuude sademete mõju neljanda ala mändide juurdekasvule (aprill enne kuivendust
r = 0,370 ja peale kuivendust r = 0,196; mai enne r = 0,153 ja pärast r = 0,003). Tunduvalt on
vähenenud ja muutnud korrelatsiooni suunda ka septembri sademete mõju juurdekasvule (enne
kuivendust r = 0,454 (p < 0,05) ja peale kuivendust r = –0,036) (joonis 9b).
Viiendal proovialal ilmnevad suurimad muutused aastarõngakronoloogiate ja
temperatuurinäitajate omavahelistes seostes jaanuaris ja märtsis (vastavalt jaanuar enne
kuivendust r = –0,007 ja peale r = –0,277 (p < 0,05) ning märts enne kuivendust r = 0,335
(p < 0,05) ja peale kuivendust r = 0,085). Juuni ja augusti kuu sademete mõjud on mõlemal
ajaperioodil statistiliselt olulised ja positiivsed (juuni enne kuivendust r = 0,458 ja peale
kuivendust r = 0,606; augusti eelnevalt r = 0,478 ja peale r = 0,385) (joonis 9c).
23
Joonis 9. Proovialade aastarõngaste kronoloogiate korrelatsioonid temperatuuride (T) ja
sademetega (S) kasvuaastale eelnenud oktoobrist kasvuaasta septembrini enne ja peale kuivendust
ning kogu vaatlusperioodi jooksul. Mustad täpid tähistavad statistiliselt olulisi seoseid (p < 0,05).
24
3.2.4 Põua ja sademete rohkuse mõju puude juurdekasvule
Võrreldes aastarõngalaiuste kronoloogiaid SPEIP1 väärtustega (SPEIP1 tähistab indeksi
arvutamist ühe kuu kohta (Jaagus et al. 2021)) tulevad esile statistiliselt olulised positiivsed seosed.
Analüüsi ajaliseks pikkuseks oli teisel proovialal 1976–2018 ja teistel proovialadel 1949–2018.
Teise prooviala rabamändide juurdekasvu ja põuasuse indeksi vahel ei esinenud statistiliselt olulisi
seoseid. Kolmanda prooviala puude juurdekasvu ja juulikuu SPEIP1 indeksi vahel esineb
statistiliselt oluline positiivne seos (r = 0,256, p < 0,05). Neljanda prooviala puude
aastarõngalaiuste ja kuude põuasuse indeksi vahel esinesid statistiliselt olulised positiivsed seosed
aprillis ja juunis (vastavalt r = 0,245 ja r = 0,271, p < 0,05). Viiendal proovialal ilmnesid
statistiliselt olulised seosed aastarõngalaiuste ning juuni ja augusti põuasuse indeksi vahel
(vastavalt r = 0,557 ja r = 0,396, p < 0,05). Kui vaadata lisas 5 välja toodud proovialade keskmiste
aastarõngakronoloogiate ja aastaseid põuaindekseid tuleb samuti esile viienda prooviala vastavate
aegridade sarnane käik. Seos teiste alade näitajate vahel ei ole nii selge.
Joonis 10. Proovialade aastarõngakronoloogiate ja SPEIP omavahelised korrelatsioonid märtsist
augustini. Täidetud tulbad tähistavad statistiliselt olulisi seoseid (p < 0,05).
25
4. Arutelu ja järeldused
Teise prooviala keskmistatud aastarõngalaiuste kronoloogia ja ülejäänud proovialade
kronoloogiate vahel ei ilmne suurt sarnasust (joonis 11). Sealsel proovialal kasvavate puude
aastarõngalaiused on olnud langustendentsis alates 1990. aastatest, mis kattub endise
turbakaevandusala mahajätmise ja veetaseme langetamise lõpuga. Varasemad uuringud on
leidnud, et veetaseme tõustes langeb puude iga-aastane radiaalne juurdekasv
(Linderholm et al. 2002, Edvardsson et al. 2016, Tamkevičiūtė et al. 2018). Samuti on
kronoloogiast näha 2013. aastal kobraste poolt ala üleujutamise tõttu tingitud puude juurdekasvu
vähenemine (joonis 4b). Teise prooviala standardiseeritud ja standardiseerimata
aastarõngakronoloogiate dendroklimaatilise analüüsi tulemuste erinevus (joonis 5a) on
põhjendatav kuivenduse mõjuga. Prooviala asus endise kaevandusala kõrval ja seetõttu oli ta ka
oodatult väga tugevasti mõjutatud kaevandamisega seotud kuivendamisest. Standardiseerimise
käigus eemaldatakse kronoloogiast trendid. Kui põhiliseks trendiks on olnud kuivenduse mõju,
siis trendi eemaldamisel kaovad ka sellele viitavad korrelatsioonid. Tulemuste põhjal saab öelda,
et teise prooviala kasvutingimused olid enim mõjutatud kuivendamisest, kuid rolli mängisid ka
kasvuperioodile eelnenud kuude temperatuurid (november, detsember), kuid ka augusti ja
septembri temperatuur (joonis 5a). Ka hüdroloogiliselt sarnaseid kuid omavahel liites ei ilmne
teisel proovialal ilmastikutingimuste suuremat mõju sealsete mändide juurdekasvule (joonis 6).
Kolmanda prooviala aastarõngaste kronoloogia näitab keskmiste aastarõngalaiuste suurenemist
1970. aastate keskpaigast 1980. aastateni. Sellist radiaalse juurdekasvu suurenemist on tinginud
veetaseme langetamine 150 kuni 200 meetri kaugusel asuval turbakaevandamise alal. Pärast seda
on märgata stabiilset keskmiste aastarõngalaiuste kahanemise tendentsi, mis kattub endise
kaevandusala maha jätmise ja selle loomuliku taastumisega (joonis 4c). Prooviala
aastarõngakronoloogiate analüüsil kliimaandmetega saadud mõneti erinevad tulemused on samuti
põhjendatavad standardiseerimisel kaduva kuivendamise trendi mõjuga (joonis 9b). Samas viitab
püsiv samasuunaline ja statistiliselt oluline korrelatsioon selle prooviala mändide kasvu seotusele
aprilli sademetega (joonis 7b). Juuli sademete positiivne mõju ja sealsete mändide juurdekasvudele
(joonis 7b) on kolmandal proovialal ühine ühe Dauškane et al. (2011) uuringus olnud proovialaga.
Neljas prooviala, mis asub uuritava jääksoo kõige niiskemas osas, ei ole eelduste kohaselt
mõjutatud veetaseme muutustest turbakaevandusalal, vaid kajastab klimaatiliste muutujate
26
mõjusid. Sealse ala juurdekasvurõngaste kronoloogia näitab vaadeldaval ajal ühtlast puude
juurdekasvu, erandiga ajavahemikus 2011–2015. Selle ajaperioodi keskmiste aastarõngalaiuste
kasvu saab põhjendada keskmisest soojemate aastatega (Kotta et al. 2018) (joonis 4d). Prooviala
aastarõngalaiuste ja ilmastiku andmete võrdlemine kinnitab eeldust, et alal kasvavad puud on kõige
vähem mõjutatud nii veetaseme muutustest endisel turbakaevanduse alal kui lähedal asuvast
Kivijärvest (joonis 5c). Neljas prooviala käitub loodusliku alana ja sealsete mändide juurdekasvu
mõjutavad positiivselt aprilli sademed. Sarnaseid seoseid aprilli sademete mõjuga on varasemalt
leidnud ka Dauškane et al. (2011) uurides Läti rabamände. Kui võrrelda sademete mõju prooviala
mändidele kahel ajaperioodil (enne 1970. aastat ja pärast 1971. aastat), on näha sademete olulisuse
vähenemist ning kasvuaastale eelneva novembri ja detsembri temperatuuri mõju vähenemist ja
suuna muutust (joonis 9b).
Viiendal proovialal, mis asub endisest kaevandusalast kõige kaugemal ning Kivijärve vahetus
läheduses, oli näha mõjutusi juurdekasvule, mis tulenesid veetaseme muutustest järves. Kivijärve
veetaset on tugevalt alandatud kahel korral. Esimesel korral aastatel 1928–1929 ja teisel korral
1973. aastal. 1928.–1929. aastate veetaseme langetamise (ühe meetri võrra) mõju on kronoloogias
näha järsu keskmiste aastarõngalaiuste kasvu ja peale seda stabiilset aastarõngalaiuste kahanemise
(1930. aastatel) näol (joonis 4e). Hilisem aastarõngalaiuste kahanemine on seletatav puude šokiga
selle muutuse suhtes. Väiksemad lühiajalised keskmiste aastarõngalaiuste tipud 1970. ja 1980.
aastatel on tingitud sekundaarsetest kuivendustest ja väetiste ekstensiivsest kasutusest ümbruses
asuvatel põldudel (Sults 2003) (joonis 4e). Prooviala kronoloogiaid ja ilmastiku andmeid
omavahel võrreldes selgub, et viies prooviala käitub pigem loodusliku alana ning on enim
mõjutatud suveperioodi alguse sademeist (joonis 8). Sarnaselt Dauškane et al. (2011) uuringuga
tuleb ka viiendas proovialas esile juuli temperatuuri positiivne mõju mändide juurdekasvule
(joonis 5d). Juuni ja augusti sademete mõju mändide kasvule ei muutu oluliselt ka järve veetaseme
alandamisel peale 1973. aastat (joonis 9c). Võrreldes selle prooviala puude aastarõngalaiuseid
SPEIP indeksi väärtustega (joonis 10 ning lisad 5 ja 6) ilmneb nende aegridade sarnane käik.
Tulemus võib viidata rabamändide kasvu suurele sõltuvusele õhuniiskusest. Rabas kui muidu
liigniiskes keskkonnas võib temperatuuri tõustes puud jääda füsioloogilise põua kätte. Kui aasta
on keskmisest niiskem, siis on selle võimalus väiksem ja ka puude aastane juurdekasv on selle
võrra parem.
27
Võrreldes kõigi proovialade kronoloogiaid on näha, et suurim kattuvus esineb kolmanda ja
neljanda prooviala aastarõngalaiuste vahel (joonis 11). Suuremad erinevused nende kahe
kronoloogia vahel algavad 1970. aastate keskelt, mis ühtib veetaseme langetamisega endisel
turbakaevandusalal. Veetaseme langetamine turba kaevandamise eesmärgil mõjutas veetaset ka
kolmandal proovialal ja selle tulemusel suurenes puude aastane juurdekasv.
Kolmanda, neljanda ja viienda prooviala keskmistatud aastarõngalaiuste kronoloogiatel on kõigil
näha sarnast kiiret juurdekasvu suurenemist ja seejärel selle järsku langust 1920. aastatel
(joonis 11). Selle muutuse põhjusteks saab pidada klimaatilisi tingimusi ning viiendal proovialal
on muutuse tinginud 1928. aasta veetaseme langetamine ühe meetri võrra lähedal asuvas
Kivijärves. Kõigil viiel proovialal saab täheldada aastarõngalaiuste ühtlast kahanemist viimastel
aastatel. Aastarõngalaiuste kahanemine on aga märgatavalt suurem aladel, mis olid rohkem
mõjutatud endisel kaevandusalal toimunud veetaseme muutustest.
Joonis 11. Proovialade mändide keskmised aastarõngalaiuste kronoloogiad.
Nagu ka mitmetes varasemates uuringutes tuleb ka selles töös esile suured erinevused proovialade
kasvutingimuste ja ilmastikutingimuste omavahelistes seostes. Edvardsson et al. 2015. aasta töö
toob välja, et sealsetes rabamuldadel kasvavtel mändidel ilmnevad nõrgad seosed juurdekasvude
ja ilmastikutingimuste vahel. Sama toob esile ka Cedro ja Soteki 2016. aastal ilmunud uuring.
Dauškane et al. 2011. aasta töö leidis, et Läti rabamändide kasvu ja aastarõngalaiuste vahel esines
positiivset korrelatsiooni juuni ja juuli temperatuuriga ja osadel uuringualadel esinesid seosed
augusti ja septembri temperatuuriga. Veebruari ja märtsi madalamate temperatuuridega ilmnesid
28
olenevalt proovialast nii positiivseid kui ka negatiivseid korrelatsioone ja osadel proovialadel
ilmnesid positiivsed korrelatsioonid veebruari, aprilli, mai ja juuli sademetega
(Dauškane et al. 2011). Käesolev töö leidis sarnaseid tulemusi kolmandal ja viiendal proovialal,
kus kolmanda prooviala puude juurdekasvu mõjutab positiivselt juuli sademed ja viienda
prooviala puude kasvu mõjutab positiivselt juuli temperatuur.
Töö alguses püstitatud hüpoteesidest sai töö käigus kinnitatud esimene ja kolmas hüpotees. Teine
hüpotees osutus ebatäpseks, kuna kolmandal proovialal mõjutasid puude juurdekasvu nii
ilmastikutingimused kui ka kuivendamine. Ilmastikutingimustest mängisid sealsel proovialal
suurimat rolli aprilli ja juuli sademed ning augusti ja septembri temperatuurid. Samuti tuli
juurdekasvukronoloogias esile aastarõngalaiuste suurenemine alates 1970. aastatest, mis langeb
kokku turbakaevandamise ja kuivenduse algusega.
Samuti viitavad käesoleva uuringu tulemused, et Lehtmetsa soos kasvavate mändide kasv
reageerib kiiremini veetaseme langemisele kui selle tõusule, kuid see aspekt vajab edaspidiseid
uuringuid.
29
5. Kokkuvõte
Lehtmetsa soo on väga eriilmeline. Põhjast piirab uuritavat ala endine turbakaevandusala, kus
turba kaevandamiseks ajavahemikul 1969–1996 alandati tugevalt sealset veetaset pinnases.
Lõunast piirab uurimisala Kivijärv, mille veetaset on sammuti kahel korral märkimisväärselt
alandatud (aastatel 1928–1929 ja 1973). Töö eesmärkideks oli analüüsida veerežiimi häiringute
ning ilmastikutingimuste mõjusid rabamändide radiaalsele juurdekasvule mööda veetaseme
gradienti viiel proovialal. Selleks koguti rabamändide puurproovid, millelt mõõdeti
aastarõngalaiused. Koostatud juurdekasvukronoloogiaid võrreldi nii turbakaevandustegevuse kui
ka temperatuuri, sademete ning põuasuse andmetega.
Kokkuvõtvalt saab Lehtmetsa soo mändide kasvu ja sealsete proovialade kohta öelda, et sealsete
mändide juurdekasvud olid tugevalt mõjutatud turbakaevandamisega kaasnevast veetaseme
langusest, seda nii kuivenduskraavi vahetus läheduses kui ka sellest kuni 150 meetri kaugusel
(vastavalt teisel ja kolmandal proovialal). Nende proovialade aastarõngakronoloogiatel on näha
keskmiste aastarõngalaiuste suurenemist 1970. aastatel, mis on põhjendatav lähedalasuva
kuivendatava ala mõjuga. Peale suurt aastarõngalaiuste tõusu sumbub see vaikselt. Endisest
turbakaevandusalast kaugemad neljas (aladest kõige niiskem) ja viies prooviala käitusid nagu
looduslikud sooalad, kusjuures viiendal proovialal ilmnesid lühiajalised reageeringud järve
veetaseme suurtele muutustele. Aastarõngakronoloogiate omavahelisel võrdlusel osutusid kõige
sarnasemaks kolmas ja neljas prooviala. Lahknevus nende kronoloogiate sarnasuses algab 1970.
aastate keskpaigast ja on selgitatav kolmanda prooviala puude kasvu mõjutava turbakaevandusala
veerežiimi muutusega.
Üksteisele geograafiliselt lähedal asuvad proovialad erinevad Lehtmetsa soos märkimisväärselt.
Dendrokliimaatilise analüüsi tulemused viitavad soo enda tugevale hüdroloogilisele survele
rabamändide kasvu formeerumisel. Kõigi proovialade aastarõngakronoloogiates täheldati
kuivenduse järgsel ajal mõningaid muutusi võrreldes kuivenduse eelse perioodiga, kuid need
muutused on proovialati erinevad. Proovialade aastarõngakronoloogiate ja kliimaandmete
analüüsil võib näha, et eri proovialade puude juurdekasvu mõjutavad erinevad ilmastikutegurid eri
ajal. Kui kolmanda prooviala mändide juurdekasv oli mõjutatud aprilli ja juuni sademetest ning
augusti ja septembri temperatuuridest, siis neljandal proovialal mõjutasid mändide juurdekasvu
30
vaid aprilli ja juuni sademed. Viiendal proovialal mõjutasid mändide juurdekasvu jällegi nii
sademed kui ka temperatuur, vastavalt juuni ja augusti sademed ja märtsi ning juuli temperatuurid.
Töö tulemustest ilmneb, et Lehtmetsa soos kasvavate mändide kasv reageerib kiiremini veetaseme
langemisele kui selle tõusule, kuid see aspekt vajab veel edaspidiseid uuringuid.
31
6. Summary
Effect of water level change and weather on radial increment of
Scots pine (Pinus sylvestris L.) in Lehtmetsa Bog
Kärt Erikson
The Master’s thesis had three main objectives. The first one was to create tree-ring chronologies
of pine trees from different areas of the Lehtmetsa Bog. The second aim was to analyse the
chronologies with climate data, and the third objective was to see the effect of anthropogenic water
level changes on the increment of Scots pines.
The Lehtmetsa Bog is located in central Estonia. On the northern border of the study area lies an
old peat extraction site (active from 1969 to 1996) and on the southern border there is Lake
Kivijärv, a lake that has had its water level lowered twice. Diverse study sites were chosen along
a gradient of water table. Tree-ring samples from five different study sites were measured, dated
and analysed.
The first study site was located directly at the former peat extraction site. The second was located
right next to it. The third was at the distance of about 150 meters from the former peat extraction
site. The fourth site was located in the most pristine part of the bog and the fifth was located near
the southern part of the peatland near the bog lake. The first study site was not included in the
analysis due to shortness of chronology (15 years). Length of chronologies at the different sites
was the following: 43 years at the second site, 172 years at the third site, 157 years at the fourth
site, and 115 years at the fifth site.
Chronologies from the third and fourth study site were the most similar. The differences between
them started from the 1970s and coincided with the start of the peat extraction, influencing tree
growth in the third study site. The second and the third study site were mostly influenced by the
changes in the water level in the former peat extraction area, but the third site was also influenced
by weather. The fourth study site acted like a natural site and the area was mostly influenced by
April and June precipitations. The fifth study site was also mostly influenced by weather but it also
showed an increase of tree growth after lowering the water level of the nearby lake Kivijärv. This
32
work showed that geographically close sample plots differed from each other significantly.
Correlations between tree growth and climatic variables varied greatly by study site. The growth
of pines at the third study site was influenced by precipitation in April and June and temperatures
of August and September. In contrast, the fourth study site tree-ring growth was only influenced
by precipitation of April and June. The fifth study site was again influenced by both temperature
and precipitation, respectively from March, July, June and August.
There were some similarities between previous work on peatland pines in Latvia and results from
this study, like correlation between tree-ring growth at study site five and precipitation in June and
temperature in July. But as mentioned in the previous paragraph, there were also a lot of differences
in the results that do not occur in previous works and can be explained by peatlands own strong
hydrological pressure on tree-ring growth and anthropogenic changes on water level and therefore
change in their growth patterns. Previous studies have also shown that Scots pine growth on
peatlands is strongly influenced by local factors.
33
Tänuavaldused
Soovin tänada oma juhendajaid Alar Läänelaidu, Kristina Soharit ja Ain Kulli, kes olid kogu töö
kirjutamise aja äärmiselt koostööaltid ja kelleta poleks töö valmimine võimalik olnud. Tänan neid
konstruktiivse tagasiside, tulemuste mõtestamise ja tehniliste nõuannete eest.
34
Kirjandus
Baillie, J.C.G., Pilcher, J.R., 1973. A simple crossdating program for tree-ring research. Tree-
Ring Bulletin, 33:7–14.
Biondi, F., Waikul, K., 2004. DENDROCLIM2002: AC++ program for statistical calibration of
climate signals in tree-ring chronologies. Computers & Geosciences, 30(3):303–311.
Cedro, A., Lamentowicz, M., 2008. The Last Hundred Years - Dendroecology of Scots Pine
(Pinus sylvestris L.) on a Baltic Bog in Northern Poland: Human Impact and Hydrological
Changes, Baltic Forestry, 14 (1): 2633.
Cedro, A., Sotek, Z., 2016. Natural and Anthropogenic Transformations of A Baltic Raised Bog
(Bagno Kusowo, North West Poland) in the Light of Dendrochronological Analysis of Pinus
sylvestris L. Forests, 7(9): 202. https://doi.org/10.3390/f7090202
Choi, W.-J., Chang, S.X., Bhatti, J.S., 2007. Drainage affects tree growth and c and n dynamics
in a minerotrophic peatland. Ecology, 88(2): 443–453.
https://doi.org/10.1890/00129658(2007)88[443:DATGAC]2.0.CO;2
Cook, E.R., Holmes, L.R., 1986. A R S T A N Guide for computer program ARSTAN. Laboratory
of Tree-Ring Research, The University of Arizona. Pp: 50–65.
Cook, E.R., Kairiukstis, L.A., 1990. Methods of Dendrochronology: applications in the
environmental science. International Institute for Applied Systems Analysis. Kluwer Academic
Publishers, Dordrecht. Lk 104 -105. http://dx.doi.org/10.1007/978-94-015-7879-0
Dauškane, I., Brumelis G., Elferts D., 2011. Effect of climate on extreme radial growth of Scots
pine growing on bogs in Latvia. Estonian Journal of Ecology, 60: 236–248.
Eckstein, D., Bauch, J., 1969. Beitrag zu Rationalisierung eines dendrochronologischen
Verfahrens und zu Analyse seiner Aussagesicherheit. Forstwissenschaftliches Centralblatt,
88:230–250.
Edvardsson, J., Stoffel, M.,Corona, C., 2016. CLIMPEAT: Climate and hydrology over the last
10,000 years – A story told by peat, pollen, and tree rings. Peatlands International, 2015:4, 32-35.
35
Edvardsson, J., Rimkus, E., Corona, C., Šimanauskienė, R., Kažys, J., Stoffel, M., 2015.
Exploring the impact of regional climate and local hydrology on Pinus sylvestris L. growth
variability – A comparison between pine populations growing on peat soils and mineral soils in
Lithuania. Plant Soil, 392, 345–356. https://doi.org/10.1007/s11104-015-2466-9
Edvardsson, J., Baužienė, I., Lamentowicz, M., Šimanauskienė, R., Tamkevičiūtė, M.,
Taminskas, J., Linkevičienė, R., Skuratovič, Ž., Corona, C., Stoffel, M., 2019. A multi-proxy
reconstruction of moisture dynamics in a peatland ecosystem: A case study from Čepkeliai,
Lithuania. Ecological Indicators, 106. https://doi.org/10.1016/j.ecolind.2019.105484
Fritts, H.C., 1976. Tree Rings and Climate. Academic Press. London.
Grissino-Mayer, H.D., 2001. Evaluating crossdating accuracy: A manual and tutorial for the
computer program COFECHA. Tree-Ring Research, 57(2):205-
221.https://repository.arizona.edu/handle/10150/251654
Holmes, R.L., 1983. Computer-assisted quality control in tree-ring dating and measurement. –
Tree-Ring Bulletin, 43: 69–78.
Jaagus, J., Aasa, A., Aniskevich, S., Boincean, B., Bojariu, R., Briede, A., Danilovich, I.,
Castro, F.D., Dumitrescu, A., Labuda, M., et al. 2021. Long-term changes in drought indices
in eastern and central Europe. International Journal of Climatology, 42: 225–249.
https://doi.org/10.1002/joc.7241
Karofeld, E., Jarašius, L., Priede, A., Sendžikaitė, J., 2016. On the after-use and restoration of
abandoned extracted peatlands in the Baltic countries. Restoration Ecology, 25(2).
https://doi.org/10.1111/rec.12436
Kotta, J., Herkül, K., Jaagus, J., Kaasik, A., Raudsepp, U., Alari, V., Arula, T., Haberman,
J., Järvet, A., Kangur,K., Kont, A., Kull, A., Laanemets, J., Maljutenko, I., Männik, A.,
Nõges, P., Nõges, T., Ojaveer, H., Peterson, A., Reihan, A., Rõõm, R., Sepp, M., Suursaar,
Ü., Tamm, O., Tamm, T., Tõnisson, H., 2018. Linking atmospheric, terrestrial and aquatic
environments in high latitude: Regime shifts in the Estonian regional climate system for the past
50 years. PLoS ONE, 13 (12), ARTN e0209568−20.
https://doi.org/10.1371/journal.pone.0209568
36
Läänelaid, A., Eckstein, D., 2003. Development of a Tree-ring Chronology of Scots Pine (Pinus
sylvestris L.) for Estonia as a Dating Tool and Climatic Proxy. Baltic Forestry, 9 (2): 7682.
Läänelaid, A., Sohar, K., Kull, A., 2014. Kuivenduse mõju ulatus Tellissaare rabas mändide
jämeduskasvu järgi. Publicationes Instituti Geographici Universitatis Tartuensis. 111.
Leibak, E., 2021. Sood Eestis ja Euroopas. In Leibak, E. (editor) Soode kaitse ja taastamine.
Eestimaa Looduse Fond, Tartu, pp. 9-11.
Linderholm, H.W., Moberg, A., Grudd, H., 2002. Peatland pines as climate indicators? A
regional comparison of the climatic influence on Scots pine growth in Sweden. Canadian Journal
of Forest Research, 32. https://doi.org/10.1139/x02-071
Lode, E., Sepp, K., Truus, L., Ilomets, M., Pajula, R., 2015. Korrastatavate jääksoode valik.
Aruanne. Ökoloogia keskus Loodus- ja terviseteaduste Instituut, Tallinna Ülikool.
Macdonald, S.E., Fengyou, Y., 2001. Factors influencing size inequality in peatland black spruce
and tamarack: evidence from post-drainage release growth. Journal of Ecology, 87(3): 404–412.
https://doi.org/10.1046/j.1365-2745.1999.00370.x
Metslaid, S., 2017. Assessment of Climate effects on Scots pine (Pinus sylvestris L.) growth in
Estonia. Eesti Maaülikool. http://doi.org/10.15159/emu.11
Metslaid, S., 2019. Erineva päritoluga hariliku kuuse populatsioonide kasvureaktsioon,
vastupanu- ja taastumisvõime põua järgselt – dendrokronoloogiline analüüs. Lõpparuanne.
https://dspace.emu.ee/xmlui/bitstream/handle/10492/6305/2018.a.%20projekt%20nr.%2015234.
pdf?sequence=1&isAllowed=y
Orru, M., Orru, H., 2008. Sustainable use of Estonian peat reserves and environmental
challenges. Estonian Journal of Earth Sciences, 57(2): 87–93.
https://doi.org/10.3176/earth.2008.2.04
Paal, J., Leibak, E., 2011. Estonian mires: inventory of habitats. Regio Ltd, Tartu: 173 p.
https://issuu.com/elfond/docs/estonian_mires_inventory
Paal, J., Leibak, E., 2013. Eesti soode seisund ja kaitstus. Tartu.
https://issuu.com/elfond/docs/eesti_soode_seisund_ja_kaitstus
37
Pepin, S., Plamondon, A.P., Britel, A., 2002. Water relations of black spruce trees on a peatland
during wet and dry years. Wetlands, 22(2): 225–33.
Punttila, P., Autio, O., Kotiaho, J.S., Kotze, D.J., Loukola, O.J., Noreika, N., Vuori, A.,
Vepsäläinen, K., 2016. The effects of drainage and restoration of pine mires on habitat structure,
vegetation and ants. Silva Fennica, 50(2). http://dx.doi.org/10.14214/sf.1462
Ramsari konventsioon, 1971.
https://www.ramsar.org/sites/default/files/documents/library/current_convention_text_e.pdf.
Ramst, R., Orru, M., Salo, V., Halliste, L., 2006. Eesti mahajäetud turbatootmisalade revisjon.
2. etapp. Ida-Viru, Lääne-Viru, Jõgeva, Järva ja Tartu maakond. (Revision of abandoned peat
production areas in Estonia. Stage 2. Ida-Viru, Lääne-Viru, Jõgeva, Järva and Tartu counties) OÜ
Eesti Geoloogiakeskus, rakendusgeoloogia ja maavarade osakond.
Rinn, F., 2003. TSAP-Win. Time series analysis and presentation for dendrochronology and
related applications. Version 0.53 for Microsoft Windows. User Reference. Rinntech Heidelberg,
Heidelberg.
Sarvas, R., 1972. Investigations on the annual cycle of development of forest trees. Active period.
Communicationes Instituti Forestalis Fenniae, 76: 1–110.
Schweingruber, F.H., 1996. Tree rings and Environment. Dendroecology. Birmensdorf, Swiss
Federal Institute for Forest, Snow and Landscape Research. Berne, Stuttgart, Vienna, Haupt
Sibul, I., 2014. Männi üldiseloomustus. Mänd Eestis. Koostaja ja toimetaja: Malle Kurm. Tartu:
Eesti Maaülikool. Valli Press OÜ. Lk 41-49.
Smiljanić, M., Seo, J-W., Läänelaid, A., Van der Maaten-Theunissen, M., Stajić, B.,
Wilmking, M., 2014. Peatland pines as a proxy for water table fluctuations: Disentangling tree
growth, hydrology and possible human influence. Science of The Total Environment, 500–501:
52–63. https://doi.org/10.1016/j.scitotenv.2014.08.056.
Sults, Ü., 2003. Rohke- ja liigtoitelisus: veekogude loomuliku vananemise kiirendaja. Käsiraamat
eutrofeerumise mõjudest pinnaveekogudele. (Nutrient rich Eutrophication and over-nutrient
hypertrophication: accelerator of natural aging of water bodies. Handbook on the effects of
eutrophication on surface waters). Peipsi Koostöö Keskus, Tartu: 32p.
38
Taminskas, J., Edvardsson, J., Linkevičienė, R., Stoffel, M., Corona, C., Tamkevičiūtė, M.,
2019. Combining multiple proxies to investigate water table fluctuations in wetlands: A case study
from the Rėkyva wetland complex, Lithuania. Palaeogeography, Palaeoclimatology,
Palaeoecology, 514: 453–463. https://doi.org/10.1016/j.palaeo.2018.11.004
Tamkevičiūtė, M., Edvardsson, J., Pukiene, R., Taminskas J., Stoffel, M., Corona, C.,
Kibirkštis, G., 2018. Scots pine (Pinus sylvestris L.) based reconstruction of 130 years of water
table fluctuations in a peatland and its relevance for moisture variability assessments. Journal of
Hydrology, 558. https://doi.org/10.1016/j.jhydrol.2018.01.067
Vasander, H., Tuittila, E.-S., Lode, E., Lundin, L., Ilomets, M., Sallantaus, T., Heikkilä, R.,
Pitkänen, M.-L., Laine, J., 2003. Status and restoration of peatlands in northern Europe.
Vicente-Serrano, S.M., Beguería, S., Lopez-Moreno, J.I., 2010. A multiscalar drought index
sensitive to global warming: the standardized precipitation evapotranspiration index. Journal of
Climate, 23, 1696–1718.
Internetiallikad
e-elurikkus. https://elurikkus.ee/bie-hub/species/6318. (Viimati vaadatud 02.11.2021)
Eesti Entsüklopeedia, Kivijärv, http://entsyklopeedia.ee/artikkel/kivij%C3%A4rv (Viimati
vaadatud 19.05.2022)
Eesti Turbauuringute Andmebaas. Kivijärve turbaala. https://turba.geoloogia.info/turbaala/52.
(Viimati vaadatud 12.01.2022)
Gardner, M., 2013. Pinus sylvestris. The IUCN Red List of Threatened Species 2013:
e.T42418A2978732. https://dx.doi.org/10.2305/IUCN.UK.2013-1.RLTS.T42418A2978732.en
(Viimati vaadatud 25.04.2022)
Maaparandussüsteemide register (Register of Land Amelioration Systems)
https://portaal.agri.ee/avalik/#/maaparandus/systeem/2020552000020 (Viimati vaadatud
30.03.2021)
Riigi Ilmateenistus. Jõgevamaa kliimast. https://www.ilmateenistus.ee/2022/02/jogevamaa-
kliimast/ (Viimati vaadatud 24.04.2022)
39
Lisad
40
Lisa 1. Mändide juurdekasvuproovide ja vastavate proovialade keskmiste
juurdekasvukronoloogiate (ref.) vahelised statistilised sarnasusnäitajad. OVL = kattuvus, Glk =
samasuunaliste muutuste protsent, GSL = eelmise usaldusnivoo, %CC = korrelatsioonikordaja ,
TV = t-väärtus, TVBP =Baillie–Pilcher t-väärtus , TVH = Hollsteini t-väärtus, CDI =
kattuvdateerimise indeks.
41
Lisa 2. Standardiseeritud ja standardiseerimata kronoloogiad.
42
Lisa 3. DENDROCLIM2002 väljundfailid. Kolmanda prooviala standardiseeritud (ülal) ja
standardiseerimata (all) kronoloogiate ning kuu keskmiste temperatuuride (T, vasemal) ja
sademete summa (P, paremal) libisevad korrelatsioonid (p < 0,05) 24-aastastes intervallides.
43
Lisa 4. DENDROCLIM2002 väljundfailid. Neljanda prooviala standardiseeritud kronoloogiate
ning kuu keskmiste temperatuuride (T) ja sademete summa (P) libisevad korrelatsioonid (p < 0,05)
24-aastastes intervallides.
44
Lisa 5. DENDROCLIM2002 väljundfailid. Viienda prooviala standardiseeritud kronoloogiate ja
kuu keskmiste temperatuuride (T) ja sademete summa (P) libisevad korrelatsioonid (p < 0,05) 24-
aastastes intervallides.
45
Lisa 6. Proovialade standardiseeritud aastarõngaste kronoloogiad ja Jõgeva ilmajaama andmetel
arvutatud põuasuse indeksi (SPEIP) aegrida. Negatiivsed SPEIP väärtused kirjeldavad
veepuuduse olukorda ja positiivsed liigniiskust.
46
Lisa 7. Viienda prooviala mändide standardiseeritud aastarõngalaiuste ja juunikuu põuasuse
indeksi (SPEIP) vaheline seos.
47
Lisa 8. Magistritöö materjali põhjal avaldatud konverentsiteesid.
Erikson, K., Kull, A., Läänelaid, A., Sohar, K., 2021. Effect of water level change on radial
increment of Scots pine in a restored peatland in Estonia. Peatlands and Peat – Source of
Ecosystem Services. Extended abstracts: 16th International Peatland Congress 2021, Tallinn,
Estonia. International Peatland Society, 235−240.
Lihtlitsents lõputöö reprodutseerimiseks ja üldsusele kättesaadavaks tegemiseks
Mina, Kärt Erikson,
1. annan Tartu Ülikoolile tasuta loa (lihtlitsentsi) minu loodud teose
Veerežiimi häiringute ja ilmastiku mõju hariliku männi (Pinus sylvestris L.)
radiaalsele juurdekasvule Lehtmetsa soo näitel,
mille juhendajad on Ain Kull, Alar Läänelaid ja Kristina Sohar
reprodutseerimiseks eesmärgiga seda säilitada, sealhulgas lisada digitaalarhiivi DSpace kuni
autoriõiguse kehtivuse lõppemiseni.
2. Annan Tartu Ülikoolile loa teha punktis 1 nimetatud teos üldsusele kättesaadavaks Tartu
Ülikooli veebikeskkonna, sealhulgas digitaalarhiivi DSpace kaudu Creative Commonsi
litsentsiga CC BY NC ND 4.0, mis lubab autorile viidates teost reprodutseerida, levitada
ja üldsusele suunata ning keelab luua tuletatud teost ja kasutada teost ärieesmärgil, kuni
autoriõiguse kehtivuse lõppemiseni.
3. Olen teadlik, et punktides 1 ja 2 nimetatud õigused jäävad alles ka autorile.
4. Kinnitan, et lihtlitsentsi andmisega ei riku ma teiste isikute intellektuaalomandi ega
isikuandmete kaitse õigusaktidest tulenevaid õigusi.
Kärt Erikson
30.05.2022
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
1
Relationship between ground levelling measurements and radar satellite
interferometric estimates of bog breathing in ombrotrophic northern bogs
Tauri Tampuu1, Jaan Praks2, Francesco De Zan3, Marko Kohv 1, Ain Kull1
1 Institute of Ecology and Earth Sciences, University of Tartu, Estonia
2.School of Electrical Engineering, Aalto University, Finland 3 Remote Sensing Technology Institute, German Aerospace Centre DLR, Germany
_______________________________________________________________________________________
SUMMARY
Understanding the seasonal oscillation of peatland surface height initiated by changes in water table level
(known as ‘bog breathing’) is key to improving spatial models of the water balance of bogs and their
greenhouse gas exchanges with the atmosphere. Bog breathing has been studied locally via point-based
measurements by telmatologists as well as over wider areas by the remote sensing community, in the latter
case often without or with limited ground-truth validation. We aim to bring the two disciplines together by
assessing the feasibility of validating Synthetic Aperture Radar (SAR) data from the Sentinel-1 satellite with
in situ ground levelling data from nanotopes with different drainage status in two hemiboreal raised bogs. We
demonstrate the continuous measurement of bog breathing using automatic ultrasonic levelling devices which
shows that, during one growing season, bog breathing amounted to 11.6–14.7 cm in hollows, 6.9–7.5 cm in
hummocks and 9.5–11.6 cm in haplotelmic nanotopes. Accounting for such relatively large vertical surface
deformations remotely using the SAR Differential Interferometry (DInSAR) technique is prone to estimation
errors owing to the so-called estimation ambiguity that occurs when deformation exceeds half the wavelength
of the radar signal (2.77 cm for Sentinel-1). We approach the ambiguity problem by estimating deformation
between consecutive SAR acquisitions (time separation 6 or 12 days) only. Remote and in situ measurements
of bog breathing correlate moderately to very strongly (rs = 0.82–0.93 in hummocks) even though, from time
to time, all nanotopes except hummocks show surface deformations in just a single day that exceed the
ambiguity threshold. This indicates that DInSAR surface deformation estimates contain useful information
despite under-estimating larger changes, and DInSAR has high potential for the assessment of bog breathing.
Our findings imply that DInSAR estimates in peatlands without ground validation should be interpreted with
caution. To take full advantage of the plentiful data from Sentinel-1, the introduction of contextual information
(e.g. temperature, precipitation and/or evapotranspiration data) could guide ambiguity resolution, as we
demonstrate up to moderate correlation (rs = 0.59 in a hollow) between precipitation and bog surface height.
KEY WORDS: bog drainage, InSAR, peatland, Sentinel-1, surface deformation
_______________________________________________________________________________________
INTRODUCTION
Northern peatlands are significant pools of stored
carbon (Leifeld & Menichetti 2018, Nichols & Peteet
2019), but peatlands can switch from being net sinks
of greenhouse gases (GHG) to emitters (Blodau
2002) depending on the water regime, which is
vulnerable to both climate change and direct human
disturbance (Gorham 1991, Ojanen et al. 2010, Yu
2012, Webster et al. 2018). The porous sponge-like
nature of peat, which allows it to adsorb and release
water and trap gases, causes the peatland surface to
fall and rise following the dynamics of the water table
(WT) (Roulet 1991, Kellner & Halldin 2002, Dise
2009). GHG exchange is largely determined by peat
moisture content, which is closely related to WT
(Heikurainen et al. 1964); as well as by temperature,
vegetation and nutrient status, all being in turn
affected by WT. Therefore, understanding the
seasonal oscillation of peatland surface height and
volume, often referred to as ‘bog breathing’ (Roulet
1991, Kellner & Halldin 2002), is key to improving
spatial models of greenhouse gas (GHG) exchange
(CO2, N2O, CH4) with the atmosphere (Fritz 2006,
Dise 2009). Despite being a well-known
phenomenon (Strack et al. 2006) and significant for
climate change (Blodau 2002, Dise 2009, Howie &
Hebda 2018), bog breathing has not been
exhaustively studied and understood (Fritz 2006,
Fritz et al. 2008, Morton & Heinemeyer 2019). Nor
has the spatiotemporal variability of bog surface
deformations been documented definitively (Fritz
2006, Bradley et al. 2022, Marshall et al. 2022).
Assessment of bog breathing over vast and remote
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
2
peatland areas is feasible with the satellite Synthetic
Aperture Radar (SAR) and differential
interferometry technique (Differential InSAR or
DInSAR) (Lees et al. 2018, Morton & Heinemeyer
2019). DInSAR has become a proven tool for
globally quantifying surface displacements at
millimetre level in many domains (Ferretti et al.
2001, Ferretti et al. 2007, Crosetto et al. 2016,
Osmanoğlu el al. 2016, Biggs & Wright 2020). Thus,
the availability of SAR data potentially opens a new
era in the study of bog breathing by finally enabling
the estimation of temporal (including seasonal) bog
surface movements with sufficient spatial resolution
over wide areas, in contrast to the few point
measurements that were previously available from
accessible study sites only.
SAR measures the amplitude and phase of the
backscatter of the transmitted electromagnetic signal
(Ferretti et al. 2007). In DInSAR, SAR phase images
from the same orbital position (zero-baseline) at
different times are combined (Bamler & Hartl 1998).
When two phase images are combined in DInSAR,
the resultant phase difference or phase change image
known as interferogram indicates the change in the
targets during the time interval between two SAR
acquisitions (Ferretti et al. 2007). Assuming that the
dielectric properties of a target remain stable, the
interferometric phase (DInSAR phase, i.e. phase
change) becomes a sensitive measure of the surface
elevation change (vertical deformation) (Bamler &
Hartl 1998, Ferretti et al. 2007). In DInSAR, an
elevation change smaller than half of the radar
wavelength can be translated to a phase change inside
a 360 degree or 2π radian circle, called the phase
cycle, and can be measured precisely. Unfortunately,
if the elevation change is larger than half of the
wavelength, the phase jumps from 2π radian to 0 and
repeats the cycle. This happens after every half-
wavelength and, thus, possibly many times. Such
periodicity brings 2π ambiguity to the relationship
between the elevation change and the phase change.
Therefore, in an interferogram, the real surface
deformation is ambiguously wrapped in 2π phase
cycles. The amount of height change that leads to a
2π change in the sensor’s line of sight (LOS) phase
change (Rosen et al. 2000) is referred as the LOS
ambiguity threshold in this paper. Whenever the
surface deformation is larger than the LOS ambiguity
threshold, phase unwrapping, i.e. addition of the
correct number of phase cycles to the phase change,
is needed to resolve the phase ambiguity and
reconstruct the true elevation change (Ferretti et al.
2007). The unwrapping issue is central to obtaining
accurate deformation estimates. The key to the
correct ambiguity resolution is a visible fringe pattern
in the interferogram. A fringe occurs when the phase
jumps from +1π radian to −1π radian or vice versa,
often referred to as a phase jump. Assuming that the
true phase gradient is continuous because the surface
of a natural terrain is elastic, a phase jump visible in
the interferogram is not caused by an abrupt shift of
a part of the observed surface (Bamler & Hartl 1998,
Ferretti et al. 2007). For further explanation of the
theory of DInSAR in the context of studying northern
bogs, refer to Tampuu (2022).
Despite having revolutionised measurement of the
Earth’s surface deformation globally (Biggs &
Wright 2020), DInSAR has seen limited application
over peat. This is because the majority of DInSAR
methods are suitable primarily for non-vegetated
surfaces where DInSAR coherence, which describes
local phase stability (Ferretti et al. 2007), is high
(Alshammari et al. 2018). Therefore, DInSAR studies
in northern peatlands have been concerned mainly
with long-term peatland subsidence using advanced
DInSAR techniques, referred to collectively as
InSAR time series analysis, which can mitigate the
coherence issue (Zhou 2013, Cigna & Sowter 2017,
Alshammari et al. 2018, Fiaschi et al. 2019).
A limited number of publications have only
recently dealt with bog breathing. To the best of our
knowledge, the inclusive list contains Alshammari et
al. (2020), Tampuu et al. (2020), Tampuu et al.
(2021a), Bradley et al. (2022), Marshall et al. (2022)
and Tampuu et al. (2022). Among these, only
Marshall et al. (2022), Tampuu et al. (2021a) and
Tampuu et al. (2022) possessed ground levelling
measurements to compare with the surface oscillation
derived using DInSAR. The last two describe our
own preliminary research, and use the same ground
validation data as this study but larger windows in
coherence estimation and phase filtering. In the
current article we have reduced the estimation
windows as a trade-off between minimising
averaging-out of the useful signal and increased noise.
The absence of ground levelling data for validation
has been characteristic for the entire field of peatland
DInSAR (Cigna & Sowter 2017, Alshammari et al.
2018). Another issue, outwith the scope of this study,
is the representativeness of point measurements for
spatially much larger footprints, such as a SAR pixel,
in bogs with high nanotope-level heterogeneity
(Alekseychik et al. 2021, Marshall et al. 2022).
The estimates of bog breathing sensed with
DInSAR by Alshammari et al. (2020), Tampuu et al.
(2020) and Bradley et al. (2022) and not validated
with in situ levelling were internally consistent and
related to peatland ecohydrology, but were
considerably smaller in magnitude than the known
possible amplitude of bog breathing (Glaser et al.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
3
2004, Fritz 2006, Howie & Hebda 2018). Alhough
Zhou (2013), Alshammari et al. (2018) and Tampuu
et al. (2020) anticipated the risk of unwrapping errors
in bogs, research has only recently provided some
evidence, based on ground data, for the unreliability
of C-band (Conventional band; wavelength ~5.6 cm)
DInSAR time series analysis in peatlands. Marshall
et al. (2022) demonstrated under-estimation during a
drought period and in more dynamic areas of blanket
bog, while Heuff & Hanssen (2020) and Conroy et al.
(2022) showed that erroneously resolved phase
ambiguities made estimates for grasslands on peat
unreliable, and Umarhadi et al. (2021) showed the
peat subsidence rate based on C-band time series was
underestimated compared to results based on L-band
(Long band; ~24 cm) and on surface deformations
modelled from the WT in tropical peatlands.
Phase unwrapping in natural landscapes can be
very complicated and, when incorrectly solved, leads
to wrong deformation estimates (Alshammari et al.
2018). To avoid the ambiguity issue, Marshall et al.
(2022) recommended using interferometry in less
dynamic parts of the peatland. To tackle the
ambiguity issue, Conroy et al. (2022) proposed the
introduction of contextual information, i.e.
temperature and precipitation, to guide unwrapping.
The estimation of elevation changes smaller than
the LOS ambiguity threshold is intrinsically least
error-prone (Conroy et al. 2022). Reducing the
temporal baseline (i.e. time interval) between the
image pairs constituting an interferogram is the
simplest way to reduce the magnitude of change
(Alshammari et al. 2018, Conroy et al. 2022).
Tampuu et al. (2020) and Tampuu et al. (2021b) have
shown that coherence is preserved in open raised
bogs over short to medium time intervals (i.e. days to
months), indicating that applying conventional short
temporal baseline DInSAR (using only radar
acquisitions whose time separation is short, i.e. days)
is possible in open bogs and can produce reliable
deformation results. Until now, ground
measurements have only been compared with
deformation estimates from InSAR time series
analysis (Alshammari et al. 2018, Marshall et al.
2022). The conventional DInSAR approach has been
largely ignored and, to the best of our knowledge,
used only by Tampuu et al. (2021a) and Tampuu et
al. (2022). However, limiting the analysis to only the
shortest baselines may capture surface movements
that are beyond the scope of advanced DInSAR.
Bog breathing has been studied locally via point-
based measurements by telmatologists, as well as
over wider areas by the remote sensing community,
in the latter case often without or with limited
ground-truth validation. We aim to bring the two
disciplines together by assessing the feasibility of
using Sentinel-1 C-band SAR data validated with
time series of in situ ground levelling data from raised
bog nanotopes with different drainage status. The
objectives of this study are:
1. to assess the accuracy of Sentinel-1 conventional
short temporal baseline DInSAR in estimating bog
breathing by comparing the results with
automatically measured continuous in situ ground
levelling measurements from different nanotopes
in two raised bogs in Estonia over the growing
season of 2016;
2. to discuss implications for the applicability of
DInSAR to monitoring bog breathing; and
3. to publish the in-situ ground levelling data from
different nanotopes along a gradient of decreasing
drainage influence.
Continuous records of in situ bog breathing are rare
in any case, and particularly rare if they address the
gradient of drainage influence. Only the shortest
available time intervals (6 or 12 days) are used for
estimating DInSAR surface deformation, to minimise
the need for phase unwrapping, and we hypothesise
that this can allow accurate assessment of the
magnitude of bog breathing because the magnitude of
surface deformations will be small enough to remain
within the convenient range for C-band radar.
MATERIALS
Study area
This study is concerned with two medium-sized
ombrotrophic mires (bogs) in Estonia (Figure 1).
Umbusi Bog (58.57 °N, 26.18 °E) and Laukasoo Bog
(58.43 °N, 27.00 °E), located 50 km apart, are
characteristic of hemiboreal (northern temperate)
raised bogs. Both bogs have a deep peat layer and a
pool system in the central part of the bog. The peat
thickness at our study transects is ~ 8 m in Umbusi
and ~ 5 m in Laukasoo. At each site, the Sphagnum-
dominated open bog is a ridge-hollow-hummock
ecotope while the central part is a ridge-pool ecotope
dominated by Pinus sylvestris trees up to 5 m tall. The
growing season in Estonia usually lasts from early
May to the end of October. Precipitation (annual
norm 672 mm) is unevenly distributed with a
minimum in winter–spring and a maximum in
summer–autumn. The bog water table (WT) is
highest during snowmelt in April, lowers rapidly in
May and June due to high subsurface discharge and
evapotranspiration, and usually reaches a minimum
level in July or August. In autumn, the peat pore
water recharges due to decreasing evapotranspiration
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
4
Figure 1. Plan views of the transects on Umbusi Bog (left) and Laukasoo Bog (right) in relation to the milled
peat extraction areas. At each site a transect of plots with automatically and manually measured bog water
table (WT) and automatic surface levelling measurements follows the gradient of decreasing drainage
influence from the drainage ditch to the interior of the open natural bog. Each transect is prolonged towards
the reference plots as two virtual transects crossing peat extraction fields. The letters “U” for Umbusi and “L”
for Laukasoo, together with the plot identifiers, denote measurement plots. The levelling device recording both
a hollow and a hummock nanotope at Plot U6 is shown in the inset image for Umbusi Bog. The location of
Tooma Meteorological Station is shown in the inset map.
and increased precipitation, and a rise of the WT
follows. The bog surface freezes and snow cover
establishes in December, and the snow persists until
April (Estonian Environment Agency 2023).
In both bogs, the natural bog area is truncated by
an active milled peat extraction area with a system of
double-ditch drainage. The border between the peat
extraction field and natural bog is formed by the main
drainage ditch which cuts through the entire peat
layer. Running parallel to that main drain, at
15–20 m in the direction of the centre of the bog,
there is a secondary drainage ditch (~ 0.5 m deep)
which penetrates the acrotelm. The margins of the
bog sections that are affected by drainage are wooded
(Pinus sylvestris).
Measurement transects along the gradient of
drainage influence
A transect of automatic surface levelling and WT
measurements has been established along the
gradient of decreasing drainage influence in both
Umbusi Bog and Laukasoo Bog (Figure 1). Both
transects, consisting of seven measurement plots
(plots 1–7), stretch from the main drainage ditch into
the intact portion of the natural bog. The transects
were initially established to measure bog WT
manually in WT sampling wells, at monthly
intervals. Automatic WT measurement and surface
levelling devices were installed later, at some of the
plots. The naming convention employed for the
measurement plots follows a spatial logic: uppercase
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
5
letter denotes the bog (U for Umbusi Bog, L for
Laukasoo Bog); the plot identifier (integer) denotes
the position of the plot in relation to the gradient of
decreasing drainage influence (plots numbered 1–7,
with 1 closest to and 7 farthest from the drainage
ditch). The levelling measurements at Plots U6 and
L6 are recorded in both a hollow and a hummock
nanotope, with the distance between the measuring
devices being 2 m and distinguished by the respective
subscripts “hol” and “hum”. The transects used in
this study are part of a wider study focusing on the
effects of drainage on bogs and transitional mires.
Detailed descriptions of the layout of transects and
other studied factors are provided by Paal et al.
(2016).
In this article we are concerned only with the six
measurement plots (U2, U4, U6 in Umbusi Bog and
L2, L4, L6 in Laukasoo Bog) for which surface
levelling data are available over the 2016 growing
season (15 Apr to 31 Oct). Automatic WT data from
the selected (levelling) plots are also considered if
available. If automatic WT data are not available
from the same plot, the manually measured WT data
are used to decide whether automatic WT
measurements from a neighbouring plot can be used
to represent the WT at the levelling plot.
The haplotelmic plots U2 and L2 (characterised
by no acrotelm and compacted peat) are located
around 15 m from the main drain and are severely
affected by drainage. Plots U4 (lawn nanotope) and
L4 (hollow nanotope) experience a significant
influence of drainage, being located ~ 50 m and ~ 40
m from the main drain in Umbusi Bog and Laukasoo
Bog, respectively. Plots U6 and L6 are located in
ridge-hollow-hummock microtopes within zones
which are intact in terms of most factors but
experience a weak influence of drainage according to
some of our metrics. This assessment is based on 107
indicators (characteristics of vegetation, nutrients,
water regime, water chemistry, nanotope structure,
landscape metrics, etc.). All of the indicators are
given in a report in Estonian (Kull 2016) and some of
them (for transitional mires) have been published in
English (Paal et al. 2016). As a simplification in this
study we consider Plots U6 and L6 to be natural sites,
lying 200 m and 75 m from the main drains in
Umbusi Bog and Laukasoo Bog, respectively.
Automatic and manual water table measurements
The automatic WT measurements were obtained
using automatic piezometers (Geotech AB, Göteborg,
Sweden) and values are recorded as daily averages for
2012–2018. In the case of Umbusi Bog, automatic WT
measurements are available from Plots U2, U4, U5and
U7 except that data for the period 26 May–25 June
2016 are missing. With regard to Laukasoo Bog, the
automatic WT data are from Plots L1, L2, L3 and L7.
The piezometers were installed in the stable peat
layer at 1.3 m depth (from the surface at the time of
installation) and measure the pressure of the water
column. The barometrically compensated WT level
relative to the peatland surface was modelled from
the water and air pressures. The WT data calculated
from piezometer readings were validated using
manual WT measurements in sampling wells located
near the piezometers. The sampling wells were
anchored in the stable peat layer 1.3 m below the mire
surface at the time of installation. WT in the sampling
wells was recorded relative to the peatland surface at
monthly (or bi-monthly) intervals year-round in
2012–2016.
Automatic surface levelling measurements
From the plots numbered 2, 4 and 6 in both bogs (U2,
U4, U6 in Umbusi Bog and L2, L4, L6 in Laukasoo
Bog), we have time series of automatic ground
levelling measurements over the growing season of
2016. For this we built devices that measure the time
an ultrasound wave takes to travel along the path
sensor–ground–sensor and convert this time into the
distance between the sensor and the ground
(resolution 0.25 mm, repeatability ± 0.2 % / ± 1 mm).
Each device was attached to a T-shaped metal bar that
penetrated through the peat layer and was anchored
in the underlying stable mineral ground. At Plots U6
and L6, one device recorded the surface elevation in
a hollow (U6hol and L6hol) and another was placed in
a hummock (U6hum and L6hum), with the distance
between the two devices being 2 m. At Plots U2, L2,
U4 and L4, only one device was used. Since only
daily average data were available for L2, all of the
hourly levelling data were aggregated to daily data.
The time series end when the levelling devices were
removed after the formation of snow cover, which
prevents the signal from reaching the ground.
Additionally, fog and intense rain may occasionally
cause inaccuracies in the measurements.
Meteorological data
Precipitation data (2012–2018) were provided by the
Estonian Environment Agency as daily totals from
Tooma Meteorological Station (Estonian Environment
Agency 2021), which is located 35 km north of
Umbusi Bog and 65 km northwest of Laukasoo Bog
(Figure 1 inset) and considered to be representative
of both locations. The growing season (15 Apr to 31
Oct) of 2016, which is in the focus of this study, was
climatologically close to normal except that May and
September were drier than normal with 16 mm vs.
42 mm and 28 mm vs. 64 mm of rainfall, respectively.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
6
SAR data
The SAR data were collected by the radar satellite
mission Sentinel-1 (S1), consisting of a constellation
of two identical satellites, S1A and S1B. The images
we used cover the period 01 Jul to 29 Oct 2016,
which is limited by the availability of S1 summer
acquisitions between periods when S1 operated in
Extended Wide (EW) swath mode for sea-ice
monitoring and ground levelling data in autumn. We
used 14 S1A and S1B ascending orbit (relative orbit
number 160) vertical-vertical (VV) polarisation
Interferometric Wide (IW) swath mode Single Look
Complex (SLC) images. As S1B started operation in
April 2016 and the first image over our study areas is
dated 29 Sep 2016, we could use three S1B
acquisitions in our stack. The acquisition time is
around 15.56 UTC, corresponding roughly to 19.00
EEST, the local time of our study area. The local
incidence angle in the sub-swath IW2 is 38.41° over
Umbusi Bog and 40.96° over Laukasoo Bog. The
pixel spacing in IW2 approximates to 4 m × 14 m
(range (rg) × azimuth (az)) on the ground and the
spatial resolution is 3.1 m × 22.7 m (rg × az) (CLS
2016). Thus, the S1 data are much coarser than a
single nanotope on the bog and the backscattering
response is formed at the level of the microtope,
which consists of a pattern of nanotopes (Lindsay
2010). The interferometric baselines (the distance
between the image acquisitions) vary roughly from
-5 to 140 metres.
METHODS
Correlating in situ measurements of bog
breathing and DInSAR deformation estimates
The aim of the DInSAR analysis was to correlate the
measured seasonal vertical surface deformation on
the open natural bog with Sentinel-1 DInSAR
deformation estimates. The deformation estimates
were calculated using the conventional short
temporal baseline (6 or 12 days) DInSAR method to
take into account possible large magnitudes of bog
breathing. The DInSAR deformation estimates were
compared to ground levelling measurements from
different nanotopes along the drainage gradient in the
raised bog.
In Umbusi Bog, 13 interferograms for Plot U4 and
11 interferograms for Plots U6hol and U6hum were
used in the analysis. In Laukasoo Bog, 11
interferograms for Plots L4 and L6hum and 13
interferograms for plot L6hol were used. Plots U2 and
L2 were excluded from the DInSAR analysis because
they were located a mere 15 m from the drain and were
thus potentially susceptible to pixel contamination by
the drainage ditch and the peat extraction field. Also,
significant tree cover and shrubs resulting from the
direct effects of drainage could potentially cause
decorrelation of the DInSAR phase. Conventionally,
coherence (γ) thresholds have been set to extract
more reliable pixels (Berardino et al. 2002, Jiang &
Lohman 2021). However, despite its common use
there is no prescribed value for such a threshold, as
the coherence bias depends upon the resolution
(sensor-dependent) and the applied coherence
estimation window size (Morishita & Hanssen 2015).
Here we use a threshold of γ = 0.4 to ensure more
reliable phase estimates in concordance with previous
research (Weydahl 2001, Mohammadimanesh et al.
2018, Braun & Veci 2020).
DInSAR processing
The interferograms, which contain unprocessed
interferometric phase and coherence estimates, were
computed using SARPROZ software (Perissin 2021).
For the interferometric coherence estimation
(weighted by the amplitude; calculated before
filtering) and Modified Goldstein phase filtering
(Goldstein & Werner 1998), a window of ten pixels
in range and three in azimuth direction was used,
approximating to a 40 m square footprint on the
ground. The 40 m length for the footprint was chosen
to best address the trade-off between the coherence
estimation bias towards higher values and the loss of
spatial resolution (Touzi et al. 1999). Flattening and
topographic phase removal were applied. No multi-
looking was applied in order to preserve the original
pixel’s spatial resolution, as natural bog displays high
nanotope and microtope level heterogeneity with
hummocks, ridges and hollows occurring at a spatial
scale of 0.5–10 m.
Formation of DInSAR deformation transects for
further analysis
For the interferograms, only the pixels corresponding
to the measurement transects were selected for
analysis, and the pixels located over the ground
levelling measurement plots were correlated with the
ground levelling measurements. To calibrate the
DInSAR phase measurements, a known stable
reference point is needed. In the absence of ground
calibration data, locations from the stable mineral soil
or compacted haplotelmic organic soil were used as
stable reference points where zero displacement in
any direction (vertical or horizontal) is assumed
(following the example by Liu et al. 2010). To study
the influence of the coherence of the chosen reference
point on the final result, we used multiple reference
points for both bogs. In Umbusi Bog the independent
reference points used were a cluster of small
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
7
buildings and an extension to the causeway, while in
Laukasoo Bog a larger building, a junction and a
causeway were used (Figure 1).
This simplified approach of assuming zero
displacement at a stable reference point in all the
images allowed us to also neglect the atmosphere
induced errors in the phase, as the atmosphere should
stay relatively constant over a few kilometres
(Webley et al. 2004, Foster et al. 2006, Bekaert et al.
2015). The maximum distance from any point in the
transect to the stable reference point was 560 m in
Umbusi and 960 m in Laukasoo. The DInSAR line of
sight (LOS) deformation (i.e. altitude change)
measurements were projected into the vertical
direction (uLOS) using local incidence angles,
assuming no horizontal motion in the peat (following
the work by Hoyt et al. 2020). Regarding minimal
temporal baselines and corresponding negligible
crustal motions (Lidberg et al. 2010), we were not
concerned with horizontal ground motion (Fuhrmann
& Garthwaite 2019). All calculations were carried
out in the SAR coordinates.
We extended each ground levelling transect
virtually over the peat extraction fields to create
continuous phase value paths (assuming a
deformation gradient along the elevation change
profile) from two reference points to the
measurement points (Figure 1). The phase difference
between the points along a transect was treated as a
difference in the deformation between imaging dates,
assuming the reference point was stable. To avoid
phase ambiguity, the whole phase dynamics between
a reference point and measurement points were fitted
into a single phase cycle (2π radians, which
corresponds to half of the wavelength of the sensor,
being ~2.77 cm for Sentinel-1). This was done by
adding a constant to the complex phase value, which
does not affect the calculated phase difference. In this
way, the linear relationship between the assumed
deformation gradient and the phase could be
maintained and assured with a visual interpretation of
the DInSAR deformation profile along the virtual
transect without the need for phase unwrapping as
long as the profile did not indicate a phase jump. If
the transect-and-phase-rotation based approach had
not been implemented, the phase at the reference
point would have been set to zero and, consequently,
the unambiguous change would have been confined
to being found in ±1π (a quarter of the radar
wavelength) (Novellino et al. 2017, Esch et al. 2019).
The Umbusi virtual transects 1 and 2 and
Laukasoo transect 2 crossed active peat extraction
fields whereas Laukasoo transect 1 extended over an
abandoned peat milling field. Laukasoo reference
point 3 was not connected to the measurement points
via a transect because there was intervening forest
which caused decorrelation in the radar signal. For
Laukasoo reference point 3 the phase rotation from
the nearest transect, i.e. Laukasoo transect 1, was used.
Data analysis
All the data were analysed in Python programming
language version 3.6 (Python Software Foundation
2021). The hourly ground levelling data were cleaned
of outliers using the Tukey’s fence method with a
multiplier of 1.5 (Tukey 1977) in a discrete 3-day
window, and presented as daily median (except for
Plot L2 where only daily averages were available).
As the data were not normally distributed according
to the Shapiro-Wilk test for normality (Shapiro &
Wilk 1965), the Spearman’s rank-order correlation
(Spearman 1904) was applied to estimate the
correlation between the levelling, DInSAR and WT
data. Reporting of the strength of correlation follows
the convention: negligible (0.0–0.3), weak (0.3–0.5),
moderate (0.5–0.7), strong (0.7–0.9), very strong
(0.9–1.0). The threshold of statistical significance is
p-value < 0.05.
To enable comparison of the calculated DInSAR
uLOS deformation estimates between two consecutive
SAR acquisitions with the in situ surface levelling
measurements, we calculated the ground-measured
vertical deformation of the peatland surface between
the dates corresponding to each interferometric pair.
The time separation was either 6 or 12 days
depending on the interferometric pair. Additionally,
to better understand how large and rapid the short-
term peatland surface deformations caused by bog
breathing can be, we calculated the vertical surface
deformation for every possible 1-day and 6-day
period from the levelling measurements. The
overlapping measurement period for all of the
levelling devices was 25 Apr to 07 Oct 2016.
RESULTS
Bog breathing
The range of bog breathing at our study plots during
the growing season (15 Apr to 31 Oct) of 2016 is
shown in Table 1, where the surface elevation is
given relative to the maximum surface height during
the measurement period. The largest surface
deformations were recorded in natural hollow
nanotopes, at Plot L6hol on Laukasoo Bog (median
surface level -4.3 cm, minimum level -14.7 cm) and
at Plot U6hol on Umbusi Bog (median -5.7 cm,
minimum -12.6 cm). The haplotelmic plots U2 and
L2 also displayed large surface fluctuations (median
surface level -5.4 cm, minimum -9.5 cm at U2;
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
8
Table 1. The magnitude of bog breathing and water table (WT) changes in Umbusi Bog and Laukasoo Bog
during the growing season (15 Apr–31 Oct) of 2016 and the growing seasons in 2012–2018. The daily median
for the period is given with the range of changes in parentheses. Bog breathing is given relative to the maximum
surface height recorded during the period, and WT is shown relative to the peatland surface. The letters “U”
and “L” (together with the plot identifier) denote measurement plots along the drainage gradient in Umbusi
Bog and Laukasoo Bog, and levelling measurements for hollow and hummock nanotopes at Plots U6 and L6
are denoted by the subscripts “hol” and “hum”, respectively. The WT values from Plots U5, L3 and L7 are
applied, respectively, to Plots U6, L4 and L6, where WT was not measured.
Plot Bog breathing 2016 WT 2016 WT 2012–2018
median (range) in cm median (range) in cm median (range) in cm
U2 -5.4 (-9.5 to 0.0) -53.1 (-68.3 to -38.6) -74.6 (-114.2 to -38.6)
U4 -5.7 (-10.8 to 0.0) -26.9 (-41.7 to -13.4) -35.4 (-63.0 to -6.0)
U5 -29.0 (-36.9 to -22.4) -33.5 (-57.7 to -13.7)
U6hol -5.7 (-12.6 to 0.0) U5 U5
U6hum -3.3 (-7.5 to 0.0) U5 U5
L2 -3.9 (-8.0 to 0.0) -29.2 (-45.6 to -22.8) -40.3 (-84.3 to -12.7)
L3 -21.8 (-27.5 to -16.0) -24.9 (-48.6 to -9.0)
L4 -3.3 (-11.6 to 0.0) L3 L3
L6hol -4.3 (-14.7 to 0.0) L7 L7
L6hum -2.2 (-6.9 to 0.0) L7 L7
L7 -14.2 (-18.2 to -6.7) -16.1 (-39.5 to 2.0)
median -3.9 cm, minimum -8.0 cm at L2). The
smallest surface fluctuations were recorded in natural
hummock nanotopes, at L6hum (median -2.2 cm,
minimum -6.9 cm) and U6hum (median -3.3 cm,
minimum -7.5 cm). In other words, bog breathing
larger in range than the Sentinel-1 sensor’s LOS
ambiguity threshold (~ 2.77 cm) was recorded in all
observed nanotopes of both bogs indifferently of the
nanotope’s disturbance status.
The largest short-term deformations were
recorded at Plot L4 in Laukasoo. Here the change in
surface level during a single day exceeded ~2.77 cm
on 28 occasions (Table 2), with subsidence occurring
on 13 of these occasions (range -3.2 to -7.8 cm,
median -5.0 cm) and surface rise occurring on 15
occasions (range 3.0 to 6.6 cm, median 4.1 cm); and
during 6-day periods there were 58 instances of
changes larger than the ambiguity threshold. Only the
natural hummock plots U6hum and L6hum exhibited
surface fluctuations that were consistently less than
the Sentinel-1 sensor’s LOS ambiguity threshold of
~ 2.77 cm.
Water table fluctuations
The WT (relative to the peatland surface at the time
of measurement) recorded by the automatic
piezometers was highest and changed least at Plot L7
(Table 1). At the natural bog plot L7, the median WT
during the 2016 growing season was -14.2 cm, range
-18.2 to -6.7 cm. The long-term (2012–2018) median
growing season WT at Plot L7 was -16.1 cm,
range -39.5 to 2.0 cm. The WT was lowest and
fluctuated over the greatest range at the haplotelmic
plot U2 (median level -53.1, range -68.3 to -38.6 in
2016; median level -74.6, range -114.2 to -38.6 in
2012–2018).
Correlation between water table and bog breathing
To study the connection between WT and bog surface
height we correlated the automatic piezometer
readings with the levelling measurements. As we had
levelling measurements but no automatic WT
measurements from Plots U6, L4 and L6, we used the
manual (2012–2016) WT readings from the sampling
wells (average values shown in Table 3) to judge
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
9
Table 2. Occasions when bog breathing exceeded the Sentinel-1 sensor’s line of sight ambiguity threshold
(~ 2.77 cm) in Umbusi Bog and Laukasoo Bog in any given 1-day and 6-day period during 2016 (the
overlapping measurement period for the levelling devices was 25 Apr to 07 Oct). The total count (N) of such
occasions is given, with subsidence (-) distinguished from rise (+). The data presented are medians of bog
surface change with the range in parentheses. The letters “U” and “L” (together with the plot identifier) denote
measurement plots along the drainage gradient in Umbusi Bog and Laukasoo Bog, and levelling measurements
from the hollow and hummock nanotopes at Plots U6 and L6 are denoted by the subscripts “hol” and “hum”,
respectively.
Plot Nanotope 1-day period 6-day period
N N- Subsidence (cm) N+ Rise (cm) N N- Subsidence (cm) N+ Rise (cm)
U2 haplotelmic 1 1 -2.9 7 1 -3.0 6 3.3
(3.0 to 5.0)
L2 haplotelmic 6 3 -4.1
(-4.8 to -3.0) 3
4.3
(2.9 to 5.1) 10 2 (-3.9 to -3.7) 8
3.2
(3.1 to 4.7)
U4 lawn 0 4 4 3.3
(3.0 to 3.6)
L4 hollow 28 13 -5.0
(-7.8 to -3.2) 15
4.1
(3.0 to 6.6) 58 30
-4.7
(-8.6 to -3.0) 28
4.3
(3.0 to 8.4)
U6hol hollow 0 7 7 3.5
(3.0 to 4.1)
L6hol hollow 4 1 -3.0 3 3.2
(3.0 to 3.5) 19 1 -3.2 18
3.6
(3.0 to 5.9)
L6hum hummock 0 0
L6hum hummock 0 0
Table 3. Weighted averages of monthly (2012–2016) manual water table (WT) measurements in the sampling
wells at Umbusi Bog and Laukasoo Bog. WT is recorded relative to the peatland surface. Distance from the
main drainage ditch is shown. The letters “U” and “L” (together with the plot identifier) denote measurement
plots along the drainage gradient in Umbusi Bog and Laukasoo Bog, respectively.
Plot U1 U2 U3 U4 U5 U6 U7
Umbusi Bog Distance (m) 10 16 26 51 101 201 365
Average WT (cm) -77.2 -64.2 -42.9 -26.8 -11.8 -12.2 -2.9
Plot L1 L2 L3 L4 L5 L6 L7
Laukasoo Bog Distance (m) 3 13 28 38 50 75 125
Average WT (cm) -114.9 -26.5 -7.0 -16.6 -5.6 -0.4 -7.5
which of the piezometers could best represent the WT
at each of these three plots. Guided by this exercise,
we applied the piezometer data from Plot U5 to Plot
U6, that from L3 to L4, and that from L7 to L6 in
further analysis (Table 1).
Figure 2 shows the dynamics of surface height in
adjacent (2 m apart) natural hollow and hummock
nanotopes in relation to WT measured by the
corresponding piezometers in Umbusi Bog (Plots
U6hol, U6hum and U5) and Laukasoo Bog (Plots L6hol,
L6hum and L7). The dynamics for haplotelmic and
drainage affected plots are shown in Figure A1 in the
Appendix. In general, the relationship between WT
and bog breathing can be expected to differ between
natural areas with intact acrotelm and haplotelmic
parts of the bog. In bog with an acrotelm, the bulk
density of the surface layer is low, leading to smooth
surface movements that correlate directly with WT
changes. In haplotelmic parts, bulk density is higher,
rewetting takes longer and fast reactions to WT
changes are rare, but seasonal surface height changes
can be of similar magnitude to those in natural bog.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
10
Figure 2. The daily median of peatland surface height and the daily average of water table depth (WT) for
Plot U6 in Umbusi Bog and Plot L6 in Laukasoo Bog where levelling data are available, along with daily
precipitation during the growing season (15 Apr–31 Oct) in 2016. The levelling data are relative to the
maximum surface height during the period. Levelling measurements from hollow and hummock nanotopes
are denoted by the subscripts “hol” and “hum”, respectively. The WT data are relative to the peatland surface
at the time of measurement. The WT measurements from Plot U5 are used to represent WT at U6 in Umbusi
and the WT measurements from Plot L7 are used to represent WT at L6 in Laukasoo.
The correlation between surface height and WT
(Table 4) in Umbusi Bog was strong at Plots U4 and
U6hol and moderate at Plots U6hum and U2. The
correlations were weaker overall in Laukasoo Bog,
being strong at Plot L2 (haplotelmic), moderate at
Plots L6hol and L6hum, and weak at Plot L4. All
correlations were statistically significant at both sites.
The intra-seasonal dynamics of WT and surface
height are detailed in Figure A2.
Correlation between precipitation, water table
and bog breathing
The correlation between the surface height, WT and
cumulative precipitation of the preceding time
periods during the growing season in 2016 is shown
in Table 4. Both surface height and WT correlated
most strongly to the cumulative precipitation of the
preceding 12 days. The correlation with 12-day
precipitation was statistically significant at all
measurement plots. In Laukasoo the correlation
between surface level and 12-day precipitation was
moderate at Plot L6hol and weak at Plot L6hum,
whereas in Umbusi this correlation was weak at U6hol
and negligible at L6hum. The correlation between WT
and 12-day precipitation was moderate at Plot U4 in
Umbusi Bog and at Plots L2 and L3 in Laukasoo Bog.
The correlations of WT with precipitation sums for
less than nine preceding days were variable, ranging
from negligible to moderate. In general, precipitation
showed stronger explanatory power over WT than
over surface height. Also, the explanatory power was
stronger for hollows than for hummocks (Table 4).
Over the long term (growing seasons 2012–2018),
the correlation of WT with the precipitation sum of
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
11
Table 4. Correlations between the daily median of automatically measured bog surface level, water table (WT)
in the corresponding wells and the cumulative daily precipitation sum (Pr.) of the preceding 1, 6, 9, 12, 15 and
18 days during the growing season (15 Apr–31 Oct) of 2016. WT is relative to the peatland surface. 118 days
have been correlated. The letters “U” and “L” (together with the plot identifier) denote measurement plots
along the drainage gradient in Umbusi Bog and Laukasoo Bog, and levelling data from hollow and hummock
nanotopes at Plots U6 and L6 are distinguished by the subscripts “hol” and “hum”, respectively. Significance
thresholds: * p < 0.05 and ** p < 0.001.
Umbusi Bog Laukasoo Bog Cumulative precipitation
Levelling WT Levelling WT
U2 U4 U6hol U6hum U2 U4 U5 L2 L4 L6hol L6hum L2 L3 L7 Pr.1 Pr.6 Pr.9 Pr.12 Pr.15 Pr.18
U m
b u
si B
o g
L ev
el li
n g
U2 1.00
U4 0.93** 1.00
U6hol 0.67** 0.81** 1.00
U6hum 0.64** 0.80** 0.98** 1.00
W T
U2 0.60** 0.68** 0.83** 0.79** 1.00
U4 0.83** 0.82** 0.78** 0.69** 0.77** 1.00
U5 0.80** 0.82** 0.77** 0.68** 0.79** 0.89** 1.00
L au
k as
o o
B o
g
L ev
el li
n g
L2 0.47** 0.60** 0.58** 0.55** 0.55** 0.54** 0.61** 1.00
L4 0.09 0.10 0.26* 0.18 0.23* 0.27* 0.37** 0.18* 1.00
L6hol 0.45** 0.56** 0.83** 0.75** 0.79** 0.73** 0.70** 0.59** 0.38** 1.00
L6hum 0.55** 0.71** 0.96** 0.94** 0.82** 0.71** 0.70** 0.63** 0.28* 0.89** 1.00
W T
L2 0.68** 0.75** 0.81** 0.73** 0.85** 0.84** 0.88** 0.71** 0.33** 0.87** 0.83** 1.00
L3 0.68** 0.67** 0.58** 0.49** 0.58** 0.75** 0.86** 0.62** 0.38** 0.68** 0.58** 0.85** 1.00
L7 0.78** 0.79** 0.61** 0.57** 0.52** 0.70** 0.82** 0.61** 0.26* 0.53** 0.56** 0.76** 0.89** 1.00
C u
m u
la ti
v e
p re
ci p
it at
io n
Pr.1 -0.04 -0.11 0.06 -0.05 0.11 0.21* 0.12 0.01 0.19* 0.31** 0.06 0.15 0.22* 0.03 1.00
Pr.6 0.13 0.05 0.24* 0.12 0.37** 0.45** 0.37** 0.08 0.27* 0.55** 0.29* 0.47** 0.46** 0.20* 0.61** 1.00
Pr.9 0.29* 0.19* 0.32** 0.19* 0.46** 0.61** 0.54** 0.18 0.30** 0.59** 0.36** 0.58** 0.57** 0.31** 0.51** 0.90** 1.00
Pr.12 0.36** 0.24* 0.34** 0.20* 0.50** 0.65** 0.62** 0.23* 0.31** 0.56** 0.36** 0.62** 0.61** 0.37** 0.43** 0.78** 0.91** 1.00
Pr.15 0.36** 0.22* 0.31** 0.18 0.48** 0.62** 0.60** 0.24* 0.33** 0.48** 0.31** 0.59** 0.57** 0.34** 0.41** 0.69** 0.81** 0.92** 1.00
Pr.18 0.34** 0.23* 0.28* 0.16 0.45** 0.59** 0.60** 0.24* 0.28* 0.43** 0.27* 0.57** 0.55** 0.34** 0.39** 0.68** 0.76** 0.86** 0.94** 1.00
the preceding 18-day period was weak but
statistically significant; and stronger than the
correlation between WT and the precipitation sum of
less than nine preceding days, which ranged from
negligible to weak (Table 5). This is consistent with
existing knowledge about WT and surface changes in
peatlands. The difference in peat bulk density
between hollows and ridges or hummocks, as well as
between diplotelmic (with acrotelm) and haplotelmic
parts of the bog, is the main factor facilitating a fast
surface uplift response to rainfall in hollows and the
diplotelmic parts of bogs. The main role in WT and
surface level lowering is played by
evapotranspiration combined with subsurface flow,
but this is a slower process and more uniform across
different vegetation complexes except for hollows
where, due to the low bulk density, WT and surface
height are strongly correlated.
Temporal behaviour of interferometric coherence
We had 13 consecutive interferograms (seven with
12-day and six with 6-day temporal baseline) over the
study period. Umbusi reference 1 (buildings) showed
minimum coherence (γmin) 0.79 and maximum
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
12
Table 5. Correlations between the daily median of automatically measured peatland water table depth below
surface (WT) and the cumulative daily precipitation sum (Pr.) of the preceding period of 1, 6, 9, 12, 15 and 18
days during the growing seasons (15 Apr–31 Oct) of the years 2012–2018. WT was measured relative to the
peatland surface. 1041 days have been correlated. The letters “U” and “L” (together with the plot identifier)
denote measurement plots along the drainage gradient in Umbusi Bog and Laukasoo Bog. Significance
thresholds: * p < 0.05 and ** p < 0.001.
Umbusi Bog Laukasoo Bog Cumulative precipitation
U2 U4 U5 L2 L3 L7 Pr.1 Pr.6 Pr.9 Pr.12 Pr.15 Pr.18
U m
b u
si
b o
g
U2 1.00
U4 0.75** 1.00
U5 0.79** 0.94** 1.00
L au
k as
o o
b o
g
L2 0.76** 0.78** 0.89** 1.00
L3 0.61** 0.84** 0.89** 0.91** 1.00
L7 0.66** 0.75** 0.85** 0.89** 0.88** 1.00
C u
m u
la ti
v e
p re
ci p
it at
io n
Pr.1 0.02 0.10* 0.09* 0.07* 0.09* 0.06 1.00
Pr.6 0.08* 0.23** 0.23** 0.20** 0.25** 0.13** 0.52** 1.00
Pr.9 0.14** 0.30** 0.30** 0.26** 0.31** 0.16** 0.45** 0.88** 1.00
Pr.12 0.20** 0.35** 0.36** 0.31** 0.35** 0.20** 0.37** 0.77** 0.91** 1.00
Pr.15 0.26** 0.38** 0.40** 0.35** 0.38** 0.23** 0.31** 0.66** 0.81** 0.93** 1.00
Pr.18 0.31** 0.40** 0.43** 0.38** 0.40** 0.25** 0.27** 0.59** 0.73** 0.85** 0.94** 1.00
coherence (γmax) 0.97 (if the image pair of 23–29 Oct,
which may be affected by snowfall, was excluded),
whereas Umbusi reference 2 (road extension) had γmin
0.24 and γmax 0.63. Plot U4 had γmin 0.29, γmax 0.75
and Plot U6 had γmin 0.18, γmax 0.85. At Laukasoo
reference 1 (junction) γmin was 0.26, γmax 0.83 and at
reference 2 (causeway) γmin was 0.19 and γmax was
0.76, whereas reference 3 (building) had γmin 0.86 and
γmax 0.98. Plot L4 had γmin 0.16, γmax 0.82, and Plot L6
had γmin 0.46, γmax 0.91.
In Umbusi, the number of image pairs that
retained minimum reliable coherence (γ > 0.4) at U4,
U6 and reference 2 (road extension) was 10 (out of
13), and at reference 1 (buildings) it was 12. In
Laukasoo, the number of image pairs that attained
γ > 0.4 was 11 at L4 but 13 at L6 and reference 3
(building). At Laukasoo references 1 (junction) and 2
(causeway), nine image pairs had γ > 0.4. In both
Umbusi and Laukasoo (Figure 3 and Figure 4), the
image pair of 13–25 Jul showed low coherence and
moderately large bog surface subsidence, coinciding
with a transition from a rainy period to a dry one. In
the image pair of 11–23 Sep, coherence was reduced
in all Umbusi plots after a dry period roughly three
weeks long and a relatively large drop in surface
elevation. Introduction of the 6-day temporal
baseline starting with the image pair of 23–29 Sep
coincided with higher γ values at all plots (except 23–
29 Oct in Umbusi Bog).
Correlation between bog breathing and DInSAR
measurements
There were considerable differences in the behaviour
of DInSAR phase and coherence between the open
bog, peat extraction areas and reference plots as
illustrated by Figures 5, 6, A3 and A4. The radar
often recorded a distinctive phase change (i.e.
deformation) over the peat extraction fields along
Umbusi transect 1 (illustrated by Figure 5c), which
was referenced to the buildings, even though milled
peat fields are characterised by compacted peat,
effective drainage and relative stability. This phase
change was often present even in image pairs of
relatively good coherence, where it could not be
caused solely by phase decorrelation (for further
discussion, see Tampuu et al. 2021a). However, the
phenomenon was not present in many of the autumn
images (as illustrated by Figure 6).
The correlation between the uLOS deformation
estimated using the DInSAR technique and the
vertical peatland surface deformations measured in
situ is shown in Figure 7, where we have referenced
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
13
Figure 3. DInSAR line of sight deformation projected to vertical direction (uLOS) (i.e. altitude change)
compared to vertical surface deformation measured in situ by levelling (shown with standard deviation)
between consecutive SAR acquisition dates at Plots U4 and U6 in Umbusi Bog. Levelling measurements from
hollow and hummock nanotopes at U6 are denoted by the subscripts “hol” and “hum”, respectively. Daily
precipitation sum corresponds to the date of the second image of each pair. Coherence threshold γ > 0.4 set for
the pixel corresponding to the levelling plot indicates whether the quality of the phase measurement can be
trusted.
Plots U4, U6hol and U6hum to the buildings 560 m
away (Umbusi reference 1). It must be remembered
that the resolution of S1 (satellite) data is relatively
coarse and one pixel is much larger than a single
nanotope. Therefore, the backscattering response is
formed at microtope level and it is not known which
nanotopes dominate in the response. The Spearman’s
correlation coefficients (rs) between DInSAR
estimates and surface levelling measurements
corresponding to the period covered by an
interferogram were moderate (rs 0.53; p-value 0.061)
at U4, strong (rs 0.85; p-value < 0.001) at U6hol and
very strong (rs 0. 93; p-value < 0.001) at U6hum
(Figure 7a1–a3). We were not able to reliably detect
deformations close to or more than the uLOS
ambiguity threshold. None of the interferograms
displayed a fringe pattern, which occurs when the
deformation phase jumps from +1π radian to -1π
radian or vice versa, when the levelling data
confirmed deformations larger than the LOS
ambiguity threshold (illustrated by Figure 5b). The
transects in Figure 5c are noisy but clearly do not
display any phase jumps. Addition or subtraction of
a phase cycle (correctly resolved ambiguity) could
have improved our results (Figure 7). When
referenced to the road extension (Umbusi
reference 2), Plots U4, U6hol and U6hum displayed
statistically insignificant negligible to weak
correlation. Referencing the plots to the mean value
of the active peat milling field along transect 1
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
14
Figure 4. DInSAR line of sight deformation projected to vertical direction (uLOS) (i.e. altitude change)
compared to vertical surface deformation measured in situ by levelling (shown with standard deviation)
between consecutive SAR acquisition dates at Plots L4 and L6 in Laukasoo Bog. Note the wider span of the
y-axis for L4. Levelling measurements from hollow and hummock nanotopes at L6 are denoted by the
subscripts “hol” and “hum”, respectively. Daily precipitation sum corresponds to the date of the second image
of each pair. Coherence threshold γ > 0.4 set for the pixel corresponding to the levelling plot indicates whether
the quality of the phase measurement can be trusted.
resulted in statistically insignificant negligible and
weak correlations at U4 and U6hol, respectively, and
a near to significant moderate correlation (rs 0.58;
p-value 0.063) at U6hum. Umbusi plots referenced to
the mean value of the active peat milling field along
transect 2 gave negligible correlations at U4 and
U6hol and a weak correlation at U6hum, although all of
the correlations were not statistically significant.
When Laukasoo levelling plots were referenced to
the building slightly less than one kilometre away
(Laukasoo reference 3), the correlation between the
DInSAR results and surface levelling measurements
corresponding to the period covered by an
interferogram was statistically insignificant and weak
to moderate for L4 and L6hol. At Plot L6hum, the
correlation was statistically significant and strong
(rs 0.82; p-value 0.002) (Figure 7b1–b3). The plots in
Laukasoo Bog, when referenced to the junction
(Laukasoo reference 1) or to the causeway (Laukasoo
reference 2), displayed negligible negative or positive
and mainly statistically insignificant correlations.
When referenced to the mean value of the abandoned
peat milling field along transect 1, Plots L4, L6hol and
L6hum displayed statistically insignificant weak
correlation coefficients. Referencing the levelling
plots to the mean value for the active peat milling
field along transect 2 gave statistically insignificant
negligible or weak negative correlations.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
15
Figure 5. Umbusi Bog: InSAR coherence γ (a) and phase (b) images and radar line of sight deformation along
the virtual transects (c) in radar coordinates for the image pair of 06–18 Aug 2016. Stable reference points
used in DInSAR processing and the plots (U4 and U6; ground levelling data available) used in validation of
DInSAR deformation estimates are indicated. The geographic coordinates of the scene are shown in Figure 1.
(a)
(b)
(c)
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
16
Figure 6. Umbusi Bog: InSAR coherence γ (a) and phase (b) images and radar line of sight deformation along
the virtual transects (c) in radar coordinates for the image pair of 29 Sep–05 Oct 2016. Stable reference points
used in DInSAR processing and the plots (U4 and U6; ground levelling data available) used in validation of
DInSAR deformation estimates are indicated. The geographic coordinates of the scene are shown in Figure 1.
(a)
(b)
(c)
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
17
Figure 7. Correlations (rs) between the vertical surface deformation measured in situ by levelling (x-axis,
shown with standard deviation) and DInSAR line of sight deformation projected to vertical dimension (uLOS)
at (a1–a3) Umbusi plots U4, U6hol (hollow) and U6hum (hummock) and (b1–b3) Laukasuo plots L4, L6hol
(hollow) and L6hum (hummock). A white X on black background marks a data point of DInSAR coherence
γ > 0.4 (indicating unreliable phase estimate). Red points represent uLOS values if a phase cycle is
added/subtracted in the cases where the in-situ value is close to the Sentinel-1 uLOS ambiguity threshold. Notice
the wider span of the x-axis in (b1).
(a1) (b1)
(a3)
(a2) (b2)
(b3)
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
18
The amplitude of bog breathing in the observation
periods of 6–12 days differ among nanotopes but
usually lie in the range of the uLOS ambiguity
threshold (3.54 cm in Umbusi and 3.67 cm in
Laukasoo). Only the natural hummock nanotope
displayed surface deformations between the radar
acquisition dates that were consistently less than the
uLOS ambiguity threshold (Figures 3–4 and Figure 7).
DISCUSSION
Bog breathing and its spatiotemporal variability
We studied bog breathing along the gradient of
decreasing drainage influence in two raised bogs
continuously, using automatic levelling devices
exploiting the travel time of ultrasound between the
sensor and the ground. Automated systems have been
recommended for monitoring bog breathing
(Marshall et al. 2022). In other peatland
environments, the use of extensometers (Van Asselen
et al. 2020, Conroy et al. 2022) and cameras (Evans
et al. 2021) has recently been demonstrated. We used
the transect-based approach to gauge not only how
bog breathing proceeds through time, but also how it
develops along the drainage gradient. We also
studied the behaviour of different nanotopes along
the transect; most significantly, we compared the
dynamics of a natural hollow and a hummock just
two metres apart. The variability of bog breathing in
blanket bogs in the UK was recently studied by
Marshall et al. (2022) who showed the interior parts
of blanket bogs being the most dynamic and
undergoing large surface deformation in extreme
drought conditions. Additionally, we have presented
a preliminary analysis of how the mechanisms
underlying bog breathing (in this study precipitation
and WT in regard to nanotopes) (Roulet 1991,
Kellner & Halldin 2002, Morton & Heinemeyer
2019) are correlated to the recorded surface motion.
We measured the range of bog breathing in two
raised bogs as 11.6–14.7 cm at natural and disturbed
hollow nanotopes and 6.9–7.5 cm at natural
hummock nanotopes during the growing season of
2016. The haplotelmic nanotopes oscillated in the
range 9.5–11.6 cm. Our levelling measurements
agree with preceding in situ research by others. A
literature review by Fritz (2006) found that bog
breathing amounts to 2.6–11 cm in natural bogs (11
plots from 9 studies) and 0.7–13 cm in disturbed
peatlands (14 plots in 12 studies). Howie & Hebda
(2018) recorded a multiyear average surface
oscillation of 11.7 cm at plots in natural raised bog,
whereas disturbed plots had a mean of 9.1 cm.
Marshall et al. (2022) measured an inter-seasonal
range of 3–12 cm in the interior part of a low-lying
blanket bog.
We recorded the largest surface elevation changes
over a 1-day period in a disturbed hollow nanotope:
up to 6.6 cm of uplift and up to 7.8 cm of subsidence.
The short-term vertical surface changes in hummocks
were modest and below the Sentinel-1 LOS
ambiguity threshold (2.77 cm) in any given 6-day
period. Glaser et al. (2004) observed oscillations at a
raised bog plot that exceeded 20 cm over a period of
four hours. The time needed for the peat matrix to
equilibrate with the changed WT may range from
hours to days (Fritz et al. 2008).
Bog breathing is generally smallest, and rapid
surface movements are rare, when the peat has
become uniformly saturated with rainwater. The
surface elevation changes in different nanotopes are
most uniform for periods of very dry or fully water-
saturated peat conditions. Regarding fully saturated
peat, when pores are filled with rainwater up to the
surface, any excess rainwater simply flows away
across the peat surface and through the acrotelm
(Holden et al. 2004). The surface fluctuations are
caused by relatively uniform changes in WT and the
magnitudes of change in hollows and hummocks
harmonise (illustrated by U6hol and U6hum in late
September and October; Figure 3). Regarding dry
peat, the rainwater pours freely through the pores of
the upper peat layers without being trapped there and
the surface fluctuations depend solely on the uniform
expansion or shrinkage of the lower-lying layers
(Evans et al. 1999). The amplitude of bog breathing
is largest during periods when the peat is nearly
saturated with water. The water-filled pore space is a
poor conductor of water, therefore the surface
fluctuations become determined by the amount of
rainfall and where the rainwater accumulates on the
peat surface (Kellner & Halldin 2002). Consequently,
the disparity in the magnitude of bog breathing in
hollows and hummocks is considerable (illustrated
by L6hol and L6hum in autumn; Figure 4). Accordingly,
hollows and hummocks fluctuated at different
magnitudes during most of the summer (Plot U6 in
Figure 3, Plot L6 in Figure 4).
Perspectives for DInSAR in northern raised bogs
Our results show that DInSAR uLOS deformation
estimates correlate moderately to strongly with bog
breathing measured in situ. The highest correlations
were recorded in natural hummock nanotopes, which
are especially suitable for applying DInSAR because
their surface fluctuations in any given 6-day period
were always below the LOS ambiguity threshold.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
19
Thus, despite under-estimating the larger surface
height changes, the bog breathing estimates from
conventional DInSAR contain a useful deformation
signal. However, it was only when buildings were
used as stable reference points for the DInSAR phase
measurements that the correlation between remotely
estimated and in situ measured surface deformation
values became strong and statistically significant.
Using causeways as reference points gave only a
statistically insignificant weak correlation. This
indicates that further complications can be expected
when using C-band conventional DInSAR in
peatlands which are located in remote areas.
A useful deformation signal is also contained in
advanced DInSAR time series, as demonstrated by
Tampuu et al. (2022) using the method of Ansari et
al. (2018). This is in accordance with Marshall et al.
(2022) who showed, using an advanced DInSAR
time series approach by Cigna & Sowter (2017), that
despite under-estimation of the magnitude of
oscillation, the timing of peaks and valleys in the time
series and the direction of change were mostly
correct. Even without precise bog breathing
magnitudes, that information could allow a wide
scale assessment of the status of peatlands, as
demonstrated recently (Alshammari et al. 2020,
Bradley et al. 2022, Islam et al. 2022).
Our approach to overcoming the ambiguity issue
and deriving precise magnitudes of peatland surface
elevation changes was twofold. First, limiting the
time interval between radar images to the minimum
available, and thereby reducing the magnitude of
surface displacement that had to be sensed
(Alshammari et al. 2018), aimed to eliminate the
need for unwrapping. Secondly, establishing the
virtual transect stretching from stable ground to the
levelling plots helped us to predict the dominant
direction of elevation change along the transect,
which we presumed could guide phase ambiguity
resolution when unwrapping was needed. However,
none of the interferograms displayed fringes
(recognisable phase jumps) when the levelling data
confirmed deformations larger than the LOS
ambiguity threshold, and we were not able to reliably
estimate such deformations.
On one hand, the loss of DInSAR coherence at the
bog margins, caused by the tree cover, makes a phase
jump in a marginal bog area invisible to C-band
(Tampuu et al. 2020). On the other hand, even the
haplotelmic plots only 15 m from the drainage ditch
experienced bog breathing of up to 9.5 cm and rapid
surface displacements that exceeded the LOS
ambiguity threshold in just a single day. Thus,
Sentinel-1 pixels with spatial resolution around 3 m
× 22 m in IW mode (CLS 2016) may be too large to
capture the deformation gradient. The maximum
detectable deformation gradient is one fringe per
pixel (Massonnet & Feigl 1998, Rosen et al. 2000),
being affected also by DInSAR coherence and the
number of looks (Jiang et al. 2011). We used a phase
filtering window with a ~ 40 m footprint on the
ground. Alshammari et al. (2018) and Marshall et al.
(2022) used multilooked pixels with a side of ~ 80 m
and an advanced DInSAR method, which allows
estimates to be retrieved also in wooded peatland
areas (Alshammari et al. 2018). Future research
could consider working with unfiltered
unmultilooked pixels to preserve the maximum
spatial resolution and avoid the possibility of
averaging over the deformation gradient occurring
across such a limited spatial extent.
To tackle the ambiguity issue, future research
should consider the introduction of contextual
information, e.g. temperature, precipitation and
evapotranspiration (Roulet 1991, Lhosmot et al.
2021) to guide unwrapping, following the work by
Heuff & Hanssen (2020) and Conroy et al. (2022) on
peatland grasslands. This could make C-band
DinSAR reliable also in estimating larger changes, as
we have demonstrated up to moderate statistically
significant relationships between precipitation, WT
and bog surface height. Alternatively, Zhou et al.
(2010) and Zhou (2013) have recommended using
longer wavelength radar, which would mitigate the
unwrapping problem by allowing the surface
displacement to fit into one phase cycle (Hoyt et al.
2020) and also by increasing penetration through the
forest cover (Wei & Sandwell 2010, Hoyt et al. 2020,
Umarhadi et al. 2021) at bog margins. The imminent
L-band (wavelength 24 cm) missions NISAR (NASA
2021) and ROSE-L (Davidson & Furnell 2021) could
make this possible.
Importance of the availability of levelling
measurements to verify DInSAR deformation
estimates
The magnitudes of bog breathing recorded by us and
reported in the literature indicate the possibility that
the rather modest DInSAR bog surface deformation
estimates that have not been validated with ground
levelling data could be underestimates (Zhou et al.
2010, Zhou 2013, Cigna et al. 2014, Cigna & Sowter
2017, Fiaschi et al. 2019, Alshammari et al. 2020,
Tampuu et al. 2020, Bradley et al. 2022). The
underestimation, shown in our comparison with
ground levelling data, has been previously confirmed
by Alshammari et al. (2018), and by Marshall et al.
(2022) under drought conditions in the most intact
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
20
parts of a blanket bog. We have shown that
underestimation can occur in any microtope within a
raised bog under normal climatic conditions.
C-band DInSAR studies in permafrost regions are
also characterised by a lack of in situ ground
validation data (Iwahana et al. 2021), posing a similar
question of reliability. For example, the foundational
DInSAR study by Liu et al. (2010) measured
subsidence of 1–4 cm in tundra during the thawing
season, whereas Iwahana et al. (2021) reported
average seasonal thaw settlements of 5.8–14.3 cm
based on in situ measurements. De la Barreda-Bautista
et al. (2022) measured maximum subsidence of
25 cm in permafrost peatlands using digital elevation
models (DEMs) derived from multispectral and true
colour RGB imagery captured from Unmanned
Aerial Vehicles (UAVs), whereas the maximum
subsidence detected by DInSAR was 1.5 cm.
We highlight that the ability of DInSAR to
accurately estimate bog breathing is limited, and we
were not able to detect estimation errors by
inspection of the DInSAR data alone. The errors were
found only by comparison with ground-based
validation data. As it is uncertain whether, when and
where peatland surface height changes can be
presumed to remain within a convenient range for C-
band observations, we emphasise the need for caution
when interpreting DInSAR bog breathing estimates
without ground validation. The uncertainty regarding
conditions under which sufficiently small surface
height changes can be presumed has also been
demonstrated in blanket bogs by Marshall et al.
(2022). Also, the proportion of hummocks and ridges
versus hollows within microtopes, and consequently
within SAR pixels, varies between different parts of
a bog and between bogs, complicating the process of
relating SAR data to ground-based data.
An aspect not covered in this study is what exactly
governs C-band radar backscatter in peatlands, which
goes beyond merely estimating the proportion of
nanotopes in an image pixel or averaging window
(Morrison 2013). We found that DInSAR estimates
correlated better to ground-based measurements on
hummocks than in hollows if both were found in an
image pixel. However, it is not known how that
reflects the possible predominance of signal from
hummocks in backscatter, the areal dominance of
hummocks in a particular pixel, or the struggle of
DInSAR to measure larger changes (Marshall et al.
2022). The question of how faithfully in situ point
measurements can represent surface deformations
across the much larger areas of SAR pixels is a
concern (Alshammari et al. 2018, Alshammari et al.
2020) that is yet to be resolved (Marshall et al. 2022).
ACKNOWLEDGEMENTS
This study formed part of a PhD project supported by
the European Union from the European Regional
Development Fund. The research was funded by
Estonian Environmental Investment Centre grants
SLOOM12006 and SLOOM14103, Estonian State
Forest Management Centre grant LLTOM17250, and
national scholarship program Kristjan Jaak which is
funded and managed by the Archimedes Foundation
in collaboration with the Ministry of Education and
Research (Estonia). The authors thank Karsten
Kretschmer (DLR) for helping with Python, Philip
Conroy for giving a native speaker’s touch to the
manuscript, and the SarProz team for their excellent
software and extremely flexible student licensing.
AUTHOR CONTRIBUTIONS
Conceptualisation and general design of the field
studies: AK; DInSAR methodology: TT, JP; field
data and data curation: AK, MK, TT; formal analysis:
TT, AK, JP; resources and supervision of the
research: AK, JP, FDZ; writing - preparation of first
draft: TT; writing - review and editing: JP, AK, FDZ;
funding acquisition: AK.
REFERENCES
Alekseychik, P., Korrensalo, A., Mammarella, I.,
Launiainen, S., Tuittila, E.S., Korpela, I. Vesala,
T. (2021) Carbon balance of a Finnish bog:
temporal variability and limiting factors based on
6 years of eddy-covariance data. Biogeosciences,
18(16), 4681–4704.
Alshammari, L., Large, D.J., Boyd, D.S., Sowter, A.,
Anderson, R., Andersen, R., Marsh, S. (2018)
Long-term peatland condition assessment via
surface motion monitoring using the ISBAS
DInSAR technique over the Flow Country,
Scotland. Remote Sensing, 10(7), 1103, 24 pp.
Alshammari, L., Boyd, D.S., Sowter, A., Marshall,
C., Andersen, R., Gilbert, P., Marsh, S. Large,
D.J. (2020) Use of surface motion characteristics
determined by InSAR to assess peatland
condition. Journal of Geophysical Research:
Biogeosciences, 125(1): e2018JG004953, 15 pp.
Ansari, H., De Zan, F., Bamler, R. (2018) Efficient
phase estimation for interferogram stacks. IEEE
Transactions on Geoscience and Remote Sensing,
56(7), 4109–4125.
Bamler, R., Hartl, P. (1998) Synthetic aperture radar
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
21
interferometry. Inverse Problems, 14(4), R1–R54.
Bekaert, D.P.S., Walters, R.J., Wright, T.J., Hooper,
A.J., Parker, D.J. (2015) Statistical comparison of
InSAR tropospheric correction techniques.
Remote Sensing of Environment, 170, 40–47.
Berardino, P., Fornaro, G., Lanari, R., Sansosti, E.
(2002) A new algorithm for surface deformation
monitoring based on small baseline differential
SAR interferograms. IEEE Transactions on
Geoscience and Remote Sensing, 40(11), 2375–
2383.
Biggs, J., Wright, T.J. (2020) How satellite InSAR
has grown from opportunistic science to routine
monitoring over the last decade. Nature
Communications, 11(1), 3863, 4 pp.
Blodau, C. (2002) Carbon cycling in peatlands – A
review of processes and controls. Environmental
Reviews, 10(2), 111–134.
Bradley, A.V., Andersen, R., Marshall, C., Sowter,
A., Large, D.J. (2022) Identification of typical
ecohydrological behaviours using InSAR allows
landscape-scale mapping of peatland condition.
Earth Surface Dynamics Discussions, 10(2), 261–
2771.
Braun, A., Veci, L. (2020) TOPS Interferometry
Tutorial. Sentinel-1 Toolbox, ESA / Skywatch
Space Applications Inc., New York. Online at:
http://step.esa.int/docs/tutorials/S1TBX%20TOP
SAR%20Interferometry%20with%20Sentinel-
1%20Tutorial_v2.pdf, accessed 01 Oct 2020.
Cigna, F., Sowter, A. (2017) The relationship
between intermittent coherence and precision of
ISBAS InSAR ground motion velocities: ERS-1/2
case studies in the UK. Remote Sensing of
Environment, 202, 177–198.
Cigna, F., Sowter, A., Jordan, C.J., Rawlins, B.G.
(2014) Intermittent Small Baseline Subset
(ISBAS) monitoring of land covers unfavourable
for conventional C-band InSAR: proof-of-
concept for peatland environments in North
Wales, UK. In: SAR Image Analysis, Modeling,
and Techniques XIV, SPIE Proceedings Volume
9243, International Society for Optics and
Photonics, 924305.
CLS (2016) Sentinel-1 Product Definition, Issue 2.7,
S1-RS-MDA-52-7440. Collecte Localisation
Satellites (CLS), European Space Agency (ESA).
Online at: https://dragon3.esa.int/web/sentinel/
user-guides/sentinel-1-sar/document-library/-
/asset_publisher/1dO7RF5fJMbd/content/sentine
l-1-product-definition, accessed 08 Nov 2021.
Conroy, P., van Diepen, S.A.N, van Asselen, S.,
Erkens, G., van Leijen, F.J., Hanssen, R.F. (2022)
Probabilistic estimation of InSAR displacement
phase guided by contextual information and
artificial intelligence. IEEE Transactions on
Geoscience and Remote Sensing, 60, 5234611,
11 pp.
Crosetto, M., Monserrat, O., Cuevas-González, M.,
Devanthéry, N., Crippa, B. (2016) Persistent
scatterer interferometry: A review. ISPRS Journal
of Photogrammetry and Remote Sensing, 115, 78–
89.
Davidson, M.W.J., Furnell, R. (2021) ROSE-L:
Copernicus L-Band SAR Mission. In: IGARSS
2021, 2021 IEEE International Geoscience and
Remote Sensing Symposium, IEEE, 872–873.
De la Barreda-Bautista, B., Boyd, D.S., Ledger, M.,
Siewert, M.B., Chandler, C., Bradley, A.V., Gee,
D., Large, D.J., Olofsson, J., Sowter, A.,
Sjogersten, S. (2022) Towards a monitoring
approach for understanding permafrost
degradation and linked subsidence in arctic
peatlands. Remote Sensing, 14(3), 444, 19 pp.
Dise, N.B. (2009) Peatland response to global
change. Science, 326(5954), 810–811.
Esch, C., Köhler, J., Gutjahr, K., Schuh, W.D. (2019)
On the analysis of the phase unwrapping process
in a D-InSAR stack with special focus on the
estimation of a motion model. Remote Sensing,
11(19), 2295, 20 pp.
Estonian Environment Agency (2021) Weather,
Observation data: daily data. Online at:
http://www.ilmateenistus.ee/ilm/ilmavaatlused/v
aatlusandmed/oopaevaandmed/?lang=en,
accessed 14 Sep 2021.
Estonian Environment Agency (2023) Climate:
Climate normals: Temperature, Precipitation etc.
Online at: https://www.ilmateenistus.ee/kliima/
kliimanormid/ohutemperatuur/?lang=en,
accessed 29 Apr 2023.
Evans, C.D., Callaghan, N., Jaya, A., Grinham, A.,
Sjogersten, S., Page, S.E., Harrison, M.E., Kusin,
K., Kho, L.K., Ledger, M., Evers, S., Mitchell Z.,
Williamson J., Radbourne A.D., Jovani-Sancho,
A.J. (2021) A novel low-cost, high-resolution
camera system for measuring peat subsidence and
water table dynamics. Frontiers in Environmental
Science, 9, 630752, 18 pp.
Evans, M.G., Burt, T.P., Holden, J., Adamson, J.K.
(1999) Runoff generation and water table
fluctuations in blanket peat: evidence from UK
data spanning the dry summer of 1995. Journal of
Hydrology, 221(3–4), 141–160.
Ferretti, A., Prati, C., Rocca, F. (2001) Permanent
scatterers in SAR interferometry. IEEE
Transactions on Geoscience and Remote Sensing,
39(1), 8–20.
Ferretti, A., Monti-Guarnieri, A., Prati, C., Rocca, F.,
Massonnet, D. (2007) InSAR Principles:
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
22
Guidelines for SAR Interferometry Processing
and Interpretation. Fletcher, K. (ed.), ESA
Publications, Noordwijk, The Netherlands, 48 pp.
Fiaschi, S., Holohan, E.P., Sheehy, M., Floris, M.
(2019) PS-InSAR analysis of Sentinel-1 data for
detecting ground motion in temperate oceanic
climate zones: a case study in the Republic of
Ireland. Remote Sensing, 11(3), 348, 30 pp.
Foster, J., Brooks, B., Cherubini, T., Shacat, C.,
Businger, S., Werner, C.L. (2006) Mitigating
atmospheric noise for InSAR using a high
resolution weather model. Geophysical Research
Letters, 33(16), L16304, 5 pp.
Fritz, C. (2006) Surface Oscillation in Peatlands:
How Variable and Important Is It? Master’s
Thesis, University of Waikato, Hamilton, New
Zealand, 119 pp.
Fritz, C., Campbell, D.I., Schipper, L.A. (2008)
Oscillating peat surface levels in a restiad
peatland, New Zealand—magnitude and
spatiotemporal variability. Hydrological
Processes, 22(17), 3264–3274.
Fuhrmann, T., Garthwaite, M.C. (2019) Resolving
three-dimensional surface motion with InSAR:
Constraints from multi-geometry data fusion.
Remote Sensing, 11(3), 241, 21 pp.
Glaser, P.H., Chanton, J.P., Morin, P., Rosenberry,
D.O., Siegel, D.I., Ruud, O., Chasar, L.I., Reeve,
A.S. (2004) Surface deformations as indicators of
deep ebullition fluxes in a large northern peatland.
Global Biogeochemical Cycles, 18(1), GB1003,
15 pp.
Goldstein, R.M., Werner, C.L. (1998) Radar
interferogram filtering for geophysical
applications. Geophysical Research Letters,
25(21), 4035–4038.
Gorham, E. (1991) Northern peatlands: role in the
carbon cycle and probable responses to climatic
warming. Ecological Applications, 1(2), 182–195.
Heikurainen, L., Päivänen, J., Seppälä, K. (1964)
Ground water table and water content in peat soil.
Acta Forestalia Fennica, 77, 1–18.
Heuff, F.M., Hanssen, R.F. (2020) InSAR phase
reduction using the Remove-Compute-Restore
method. In: IGARSS 2020, 2020 IEEE
International Geoscience and Remote Sensing
Symposium, IEEE, 786–789.
Holden, J., Chapman, P.J., Labadz, J.C. (2004)
Artificial drainage of peatlands: hydrological and
hydrochemical process and wetland restoration.
Progress in Physical Geography, 28(1), 95–123.
Howie, S.A., Hebda, R.J. (2018) Bog surface
oscillation (mire breathing): A useful measure in
raised bog restoration. Hydrological Processes,
32(11), 1518–1530.
Hoyt, A.M., Chaussard, E., Seppalainen, S.S.,
Harvey, C.F. (2020) Widespread subsidence and
carbon emissions across Southeast Asian
peatlands. Nature Geoscience, 13(6), 435–440.
Islam, M.T., Bradley, A.V., Sowter, A., Andersen,
R., Marshall, C., Long, M., Bourke, M.C.,
Connolly, J., Large, D.J. (2022) Potential use of
APSIS-InSAR measures of the range of vertical
surface motion to improve hazard assessment of
peat landslides. Mires and Peat, 28, 21, 19 pp.
Iwahana, G., Busey, R.C., Saito, K. (2021) Seasonal
and interannual ground-surface displacement in
intact and disturbed tundra along the Dalton
Highway on the North Slope, Alaska. Land, 10(1),
22, 18 pp.
Jiang, J., Lohman, R.B. (2021) Coherence-guided
InSAR deformation analysis in the presence of
ongoing land surface changes in the Imperial
Valley, California. Remote Sensing of
Environment, 253, 112160, 19 pp.
Jiang, M., Li, Z.W., Ding, X.L., Zhu, J.J., Feng, G.C.
(2011) Modeling minimum and maximum
detectable deformation gradients of
interferometric SAR measurements. International
Journal of Applied Earth Observation and
Geoinformation, 13(5), 766–777.
Kellner, E., Halldin, S. (2002) Water budget and
surface‐layer water storage in a Sphagnum bog in
central Sweden. Hydrological Processes, 16(1),
87–103.
Kull, A. (2016) Soode ökoloogilise funktsionaalsuse
tagamiseks vajalike puhvertsoonide
määratlemine pikaajaliste häiringute leviku
piiramiseks või leevendamiseks, II etapp
(Defining the Buffer Zones Necessary to Ensure
the Ecological Functionality of Peatlands to Limit
or Mitigate the Spread of Long-term
Disturbances, Stage II). Report of the Targeted
Financing Agreement 8286 SFL nr 3-2_15/835-
14/2014 aruanne, 183 pp. (in Estonian). Online at:
https://ainkull.wixsite.com/sood/aruanne,
accessed 29 Apr 2023.
Lees, K.J., Quaife, T., Artz, R.R.E., Khomik, M.,
Clark, J.M. (2018) Potential for using remote
sensing to estimate carbon fluxes across northern
peatlands–A review. Science of the Total
Environment, 615, 857–874.
Leifeld, J., Menichetti, L. (2018) The
underappreciated potential of peatlands in global
climate change mitigation strategies. Nature
Communications, 9(1), 1071, 7 pp.
Lhosmot, A., Collin, L., Magnon, G., Steinmann, M.,
Bertrand, C., Stefani, V., Toussaint, M.L.,
Bertrand, G. (2021) Restoration and
meteorological variability highlight nested water
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
23
supplies in middle altitude/latitude peatlands:
Towards a hydrological conceptual model of the
Frasne peatland, Jura Mountains, France.
Ecohydrology, 14(6), e2315, 16 pp.
Lidberg, M., Johansson, J.M., Scherneck, H.G.,
Milne, G.A. (2010) Recent results based on
continuous GPS observations of the GIA process
in Fennoscandia from BIFROST. Journal of
Geodynamics, 50(1), 8–18.
Lindsay, R. (2010). Peatbogs and Carbon: a Critical
Synthesis to Inform Policy Development in
Oceanic Peat Bog Conservation and Restoration
in the Context of Climate Change. Environmental
Research Group, University of East London,
London, UK, 315 pp.
Liu, L., Zhang, T., Wahr, J. (2010) InSAR
measurements of surface deformation over
permafrost on the North Slope of Alaska. Journal
of Geophysical Research: Earth Surface, 115,
F03023, 14 pp.
Marshall, C., Sterk, H.P., Gilbert, P.J., Andersen, R.,
Bradley, A.V., Sowter, A., Marsh, S., Large, D.J.
(2022) Multiscale variability and the comparison
of ground and satellite radar based measures of
peatland surface motion for peatland monitoring.
Remote Sensing, 14(2), 336, 22 pp.
Massonnet, D., Feigl, K.L. (1998) Radar
interferometry and its application to changes in
the Earth's surface. Reviews of Geophysics, 36(4),
441–500.
Mohammadimanesh, F., Salehi, B., Mahdianpari, M.,
Brisco, B., Motagh, M. (2018) Multi-temporal,
multi-frequency, and multi-polarization
coherence and SAR backscatter analysis of
wetlands. ISPRS Journal of Photogrammetry and
Remote Sensing, 142, 78–93.
Morishita, Y., Hanssen, R.F. (2015) Deformation
parameter estimation in low coherence areas using
a multisatellite InSAR approach. IEEE
Transactions on Geoscience and Remote Sensing,
53(8), 4275–4283.
Morrison, K. (2013) Mapping subsurface
archaeology with SAR. Archaeological
Prospection, 20(2), 149–160.
Morton, P.A., Heinemeyer, A. (2019) Bog breathing:
the extent of peat shrinkage and expansion on
blanket bogs in relation to water table, heather
management and dominant vegetation and its
implications for carbon stock assessments.
Wetlands Ecology and Management, 27(4), 467–
482.
NASA (2021) Quick Facts. NASA-ISRO SAR
Mission (NISAR) website, NASA Jet Propulsion
Laboratory (JPL), Pasadena CA, USA. Online at:
https://nisar.jpl.nasa.gov/mission/quick-facts/,
accessed 09 Nov 2021.
Nichols, J.E., Peteet, D.M. (2019) Rapid expansion
of northern peatlands and doubled estimate of
carbon storage. Nature Geoscience, 12(11), 917–
921.
Novellino, A., Cigna, F., Brahmi, M., Sowter, A.,
Bateson, L., Marsh, S. (2017) Assessing the
feasibility of a national InSAR ground
deformation map of Great Britain with Sentinel-1.
Geosciences, 7(2), 19, 14 pp.
Ojanen, P., Minkkinen, K., Alm, J., Penttilä, T.
(2010) Soil–atmosphere CO2, CH4 and N2O fluxes
in boreal forestry-drained peatlands. Forest
Ecology and Management, 260(3), 411–421.
Osmanoğlu, B., Sunar, F., Wdowinski, S., Cabral-
Cano, E. (2016) Time series analysis of InSAR
data: Methods and trends. ISPRS Journal of
Photogrammetry and Remote Sensing, 115, 90–
102.
Paal, J., Jürjendal, I., Suija, A., Kull, A. (2016)
Impact of drainage on vegetation of transitional
mires in Estonia. Mires and Peat, 18, 02, 19 pp.
Perissin, D. (2021) SARPROZ©: The SAR
PROcessing tool by periZ. Online at:
https://www.sarproz.com/, accessed 14 Sep 2021.
Python Software Foundation (2021) Python website.
Online at: https://www.python.org/, accessed
15 Sep 2021.
Rosen, P.A., Hensley, S., Joughin, I.R., Li, F.K.,
Madsen, S.N., Rodriguez, E., Goldstein, R.M.
(2000) Synthetic aperture radar interferometry.
Proceedings of the IEEE, 88(3), 333–382.
Roulet, N.T. (1991) Surface level and water table
fluctuations in a subarctic fen. Arctic and Alpine
Research, 23(3), 303–310.
Shapiro, S.S., Wilk, M.B. (1965) An analysis of
variance test for normality (complete samples).
Biometrika, 52(3/4), 591–611.
Spearman, C. (1904) The proof and measurement of
association between two things. American
Journal of Psychology, 15(1), 72–101.
Strack, M., Kellner, E., Waddington, J.M. (2006)
Effect of entrapped gas on peatland surface level
fluctuations. Hydrological Processes, 20(17),
3611–3622.
Tampuu, T. (2022) Synthetic Aperture Radar
Interferometry as a Tool for Monitoring the
Dynamics of Peatland Surface. PhD thesis,
University of Tartu, Tartu, Estonia, 166 pp.
Tampuu, T., Praks, J., Uiboupin, R., Kull, A. (2020)
Long term interferometric temporal coherence
and DInSAR phase in Northern Peatlands. Remote
Sensing, 12(10), 1566, 22 pp.
Tampuu, T., Praks, J., De Zan, F., Kohv, M., Kull, A.
(2021a) Towards assessment of bog breathing
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
24
with Sentinel-1 Differential InSAR. In: IPC2021:
16th International Peatland Congress, Oral
Presentations, Publicon PCO, Tartu, Estonia,
472–478. Online at: IPC 2021 proceedings_oral
presentations_f.pdf (publicon.ee), accessed 12
May 2023.
Tampuu, T., Praks, J., Kull, A., Uiboupin, R., Tamm,
T., Voormansik, K. (2021b) Detecting peat
extraction related activity with multi-temporal
Sentinel-1 InSAR coherence time series.
International Journal of Applied Earth
Observation and Geoinformation, 98, 102309,
11 pp.
Tampuu, T., De Zan, F., Shau, R., Praks, J., Kohv,
M., Kull, A. (2022) Can bog breathing be
measured by synthetic aperture radar
interferometry. In: IGARSS 2022, IEEE
International Geoscience and Remote Sensing
Symposium, IEEE, 16–19.
Touzi, R., Lopes, A., Bruniquel, J., Vachon, P.W.
(1999) Coherence estimation for SAR imagery.
IEEE Transactions on Geoscience and Remote
Sensing, 37(1), 135–149.
Tukey, J.W. (1977) Exploratory Data Analysis.
Addison-Wesley Publishing Company, Reading
MA, USA, 688 pp.
Umarhadi, D.A., Avtar, R., Widyatmanti, W.,
Johnson, B.A., Yunus, A.P., Khedher, K.M.,
Singh, G. (2021) Use of multifrequency (C‐band
and L‐band) SAR data to monitor peat subsidence
based on time‐series SBAS InSAR technique.
Land Degradation & Development, 32(16), 4779–
4794.
Van Asselen, S., Erkens, G., de Graaf, F. (2020)
Monitoring shallow subsidence in cultivated
peatlands. Proceedings of the International
Association of Hydrological Sciences, 382, 189–
194.
Webley, P.W., Wadge, G., James, I.N. (2004)
Determining radio wave delay by non-hydrostatic
atmospheric modelling of water vapour over
mountains. Physics and Chemistry of the Earth,
Parts A/B/C, 29(2–3), 139–148.
Webster, K.L., Bhatti, J.S., Thompson, D.K., Nelson,
S.A., Shaw, C.H., Bona, K.A., Hayne, S.L., Kurz,
W.A. (2018) Spatially-integrated estimates of net
ecosystem exchange and methane fluxes from
Canadian peatlands. Carbon Balance and
Management, 13(1), 16, 21 pp.
Wei, M., Sandwell, D.T. (2010) Decorrelation of L-
band and C-band interferometry over vegetated
areas in California. IEEE Transactions on
Geoscience and Remote Sensing, 48(7), 2942–
2952.
Weydahl, D.J. (2001) Analysis of ERS tandem SAR
coherence from glaciers, valleys, and fjord ice on
Svalbard. IEEE Transactions on Geoscience and
Remote Sensing, 39(9), 2029–2039.
Yu, Z.C. (2012) Northern peatland carbon stocks and
dynamics: a review. Biogeosciences, 9(10), 4071–
4085.
Zhou, Z. (2013) Chapter 4: Mapping the eroding
extent and severity of peatland surface over the
Monadhliath Mountains, Scotland. In: The
Applications of InSAR Time Series Analysis for
Monitoring Long-term Surface Change in
Peatlands. PhD thesis, University of Glasgow,
UK, 231 pp.
Zhou, Z., Waldron, S., Li, Z. (2010) Integration of
PS-InSAR and GPS for monitoring seasonal and
long-term peatland surface fluctuations. In:
Remote Sensing and the Carbon Cycle,
Proceedings of the Remote Sensing and
Photogrammetry Society Conference, Burlington
House, London, UK (Volume 1), 4 pp.
Submitted 09 Jun 2022, final revision 06 Aug 2023
Editor: Olivia Bragg
_______________________________________________________________________________________
Author for correspondence:
Dr Tauri Tampuu, Department of Geography, Institute of Ecology and Earth Sciences, University of Tartu and
KappaZeta Ltd., Tartu, Estonia. Email: [email protected]
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
25
Appendix
Figure A1. Daily medians of peatland surface height and daily averages of peatland water table (WT) for plots
in Umbusi Bog and Laukasoo Bog where levelling data are available, along with daily precipitation totals,
during the growing season 15 Apr–31 Oct 2016. Plots U2 (a) and L2 (b) are haplotelmic; U4 (c) and L4 (d)
are a lawn nanotope in Umbusi and and a hollow nanotope in Laukasoo, respectively, that are significantly
affected by drainage. The levelling data are shown relative to the maximum surface height of the period, and
the WT data relative to the peatland surface at the time of measurement. The WT record from Plot L3 in
Laukasoo Bog is used to represent the WT at L4.
(a)
(b)
(c)
(d)
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
26
Figure A2. Correlations (rs; p-value < 0.001) between the daily median peatland surface height and daily
average of peatland water table (WT) during the growing season 15 Apr–31 Oct 2016. The number of days
correlated is 123. The levelling data are shown relative to the maximum surface height of the period and WT
is relative to the peatland surface at the time of the measurement. The letters “U” and “L” (together with the
plot identifier) denote measurement plots along the drainage gradient in Umbusi Bog and Laukasoo Bog, and
levelling measurements from hollow and hummock nanotopes are denoted by the subscripts “hol” and “hum”,
respectively. WT data from Plots U5, L3 and L7 are used to represent the WT at U6, L4 and L7, respectively.
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
27
Figure A3. Laukasoo Bog: InSAR coherence γ (a) and phase (b) images and radar line of sight deformation
along the virtual transects (c) in radar coordinates for the image pair 18–30 Aug 2016. Stable reference points
used in DInSAR processing and the plots (L4 and L6; ground levelling data available) used in validation of
DInSAR deformation estimates are indicated. The geographic coordinates of the scene are shown in Figure 1.
(a)
(b)
(c)
T. Tampuu et al. RELATING GROUND BASED AND SATELLITE ESTIMATES OF BOG BREATHING
Mires and Peat, Volume 29 (2023), Article 17, 28 pp., http://www.mires-and-peat.net/, ISSN 1819-754X International Mire Conservation Group and International Peatland Society, DOI: 10.19189/MaP.2022.OMB.Sc.1999815
28
Figure A4. Laukasoo Bog: InSAR coherence γ (a) and phase (b) images and radar line of sight deformation
along the virtual transects (c) in radar coordinates for the image pair 30 Aug–11 Sept 2016. Stable reference
points used in DInSAR processing and the plots (L4 and L6; ground levelling data available) used in validation
of DInSAR deformation estimates are indicated. The geographic coordinates of the scene are shown in
Figure 1.
(a)
(b)
(c)
RMK projekti „Ammendatud turbamaardlate vee-režiimi taastamise kompleksuuringu
metoodika väljatöötamine ja uuringu läbiviimine“ täitmiseks 2018-2023 a. toimunud
taimestiku seire kokkuvõte
Koostajad: TÜ rakendusökoloogia kaasprofessor Edgar Karofeld ja
TÜ taimeökoloogia kaasprofessor Kai Vellak
Vegetatsiooniperioodi keskel (juuni teisest poolest juuli esimese dekaadini) 2018-2023
aastal tehti taimestiku seireks välitööd Ess-soo, Laiuse, Kildema, Maima ja Kõima jääksoode
tähtedega (A,B, C jne) tähistatud erineva töötlusega aladel kokku 156-l märgistatud 1x1 m
püsiruudul, kus määrati taimestiku üldkatvus ja taimeliikide (soon- ja sammaltaimed) katvu-
sed (%), samblike katvus ning kulu (surnud ja kuivanud taimed) protsent. Sammal- ja soon-
taimedest, mida ei olnud välitingimustes võimalik liigini määrata võeti kaasa väike proov
määrangu kontrollimiseks laboris. Taimkatteanalüüside tulemused on sisestatud Excel tabe-
litesse jääksoode, alade ja aastate kaupa (Lisa 1). Aruandes esitatud tabelites on taimestiku
keskmised katvused ümardatud täisprotsendini. Iga taimeruudu ühe nurga perforeeritud plast-
torus mõõdeti veetaseme sügavus (cm) maapinnast (mõõtetulemused Lisas 1). Eri aastatel on
mõõtmised tehtud küll ligikaudu samal perioodil juuni teises pooles, kuid ühekordsete mõõt-
miste alusel ei saa veel teha järeldusi muutuste dünaamika kohta veetaseme sügavuses, mis
jääksoodes muutub sesoonselt suures ulatuses. Igast püsiruudust tehti koos etiketiga foto,
mida pärast järeltöötlust säilitatakse võrdluseks jääksoode ja alade kaupa kataloogides (Lisa
2). 2023. a kevadel märgistati Ess-soo ja Maima jääksoo turbasamblafragmentidega korras-
tatud alade (Ess-soos L, G, E ja Maimas E, P, C) kõrgemas ja madalamas osas kokku 72 elu-
jõulisemat turbasamblalaiku, mille juures mõõdeti ka veetaseme sügavus. Igast laigust koos
mõõtskaalaga tehti foto, et sügisel oleks võimalik hinnata samblalaigu seisundi ja suuruse
muutust. Samuti paigaldati väike mõõtevai, et vegetatsiooniperioodi lõpus oktoobris mõõta
turbasammalde pikkuskasvu. Selleks kasutati peenikesi traatvaiu, mille alumised, pudeliharja
taolised otsad paigutati aplikaatoriga ca 5-7 cm sügavusele turbasamblavaipa. Vaia sambla-
kattest väljaulatuva osa pikkust mõõdetakse kevadel (aprilli lõpus) ja sügisel (oktoobri lõpus)
ning vaia pikkuse vähenemine osutab sammalde pikkuskasvule (metoodikast täpsemalt
Karofeld et al., 2020). Võrdluseks paigaldati turbasammalde pikkuskasvu mõõtmise vaiad ka
Ess-soo ja Maima raba looduslikele aladele pruuni, punase ja raba turbasambla pikkuskasvu
mõõtmiseks. Varem ei olnud turbasamblafragmentidega korrastatud aladel võimalik selliseid
uuringuid teha, sest alad olid üleujutatud või liiga püdelad ning ei kandnud isegi räätsadega
inimest. Samuti ei ole võimalik mõõta turbasammalde pikkuskasvu kui taimefragmendid on
turbapinnal horisontaalselt või nurga all ja ei ole võimalik ette näha nende kasvu suunda.
Järgnev ülevaade projekti raames uuritud jääksoodel 2018-2023. a taimestikus toimunud
muutustest on antud soode ning alade kaupa erinevate töötlustega.
TULEMUSED ja ARUTELU
KILDEMA jääksoo on projekti raames seiratutest ainuke, kus 2023. aastaks ei ole veel
korrastamist tehtud. Nii A kui B ala asuvad mosaiikse taimestikuga alal.
Kildema jääksoo A ala asub iseloomulikul mahajäetud jääksoo osas, kus hõredate 1-2 m
kõrguste sookaskede (Betula pubescens), männi (Pinus sylvestris) tõusmete ja tupp-villpea
(Eriophorum vaginatum) mätaste vahel katab paljas turbapind 80-90 % (Foto 1). Veetase on
60-70 cm sügavusel, olles eelnenud aastatega võrreldes 20-30 cm sügavamal. Tihedam ja
kõrgem on kaskede ning mõnede mändide rivi piki kraave, milles kasvab konnaosi (Equise-
tum fluviatile), laialehine hundinui (Typha latifolia) ja pilliroog (Phragmites australis).
Taimeruutudes on soontaimedest kõige suurema katvusega tupp-villpea ning sammaltai-
medest raba-karusammal (Polytrichum strictum). Taimestiku üldkatvus on 2018. aastaga
võrreldes suurenenud ligikaudu üle kahe korra, sh soontaimedel veidi alla kahe ning sammal-
taimedel üle kolme korra (Tabel 1, Foto 2). Liike registreeriti kogu seireaja vältel püsiruutu-
dest vähe, vaid 6-7 liiki. Aja jooksul oli kadunud ruutudest harilik jõhvikas ja pugu-kaksik-
hambake (Dicranella cerviculata). Soontaimedest on suurenenud eelkõige tupp-villpea
katvus, mis 2018. aastal oli kuue seireruutude keskmisena 12,5 % ning 2023. aastal 21,4 %
(Joonis 1). Ka sookase, kanarbiku (Calluna vulgaris) katvused on tõusnud pea kaks korda,
kuid endiselt jääb nende katvus marginaalseks ning on vastavalt 5 % ja 6,7 % kuue püsiruudu
keskmisena. Sammaldest on kõigis püsiruutudes raba-karusambla katvus kõige enam suure-
nenud, tõustes 3,7%-lt 14,7 %-ni (Joonis 1). Teiste samblaliikide katvused on endiselt margi-
naalsed, kuigi katvuse tõusu on võimlaik täheldada ka longus pirniku (Pohlia nutans) katvu-
ses. Üldkatvuse suurenemine on osalt tingitud vähesest taimestumisest vaatlusperioodi algu-
ses ning esmase taimestiku ebaühtlases jaotumises, mis mõjutab järgnevat taimestumist.
Jääksoode spontaansel taastaimestumisel ongi niiskustingimuste kõrval aeg seda kõige enam
mõjutavaks teguriks (Triisberg et al., 2011, 2013). Kaugleviga jääksoole jõudnud ja kasvama
hakanud taimed saavad katvust kiiremini suurendada nii lähi- kui ka vegetatiivse levi teel,
kuid samas jääb taimestumine väga ebaühtlaseks ning paljas turbapind on domineerivaks.
Taimestiku üldkatvuse hulka on arvatud ka kulu ehk varis (st surnud taimeosised) osakaal,
mis on samuti aja jooksul pigem suurenenud ning eriti märkimisväärne oli kulu protsent 2019.
aastal tupp-villpea kuivanud mätaste näol (kuue taimeruudu keskmisena kokku 9,7 %). Kahel
viimasel aastal on alalt registreeritud ka samblikke, mille katvus on küll marginaalne, kuid
viitab samuti kuivadele tingimustele alal.
Tabel 1. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Kildema
jääksoo A ja B ala taimeruutudes 2018-2023. aastal.
2018 2019 2021 2022 2023 A üldkatvus 23 20 29 36 48 A kulu katvus 0 10 4 4 7 A soontaimede katvus 18 12 23 24 33 A sammaltaimede katvus 5 6 10 13 16 B üldkatvus 76 70 62 49 66 B kulu katvus 13 18 0 4 8 B soontaimede katvus 48 42 40 12 61 B sammaltaimede katvus 14 14 14 10 9
Joonis 1. Kildema jääksoo taimestiku seireruutudes domineerivate taimeliikide keskmised
katvused 2018-2023 a.
0
5
10
15
20
25
2018 2019 2021 2022 2023
K at vu s %
tupp‐villpea kanarbik longus pirnik raba‐karusammal
Foto 1. Üldvaade korrastamata Kildema jääksoo A alale.
Foto 2. Kildema jääksoo taimeruut A VI 2018. (vasakul) ja 2023. aastal (paremal).
Kildema jääksoo B ala asub turbakaevandamise ettevalmistamise ja kraavide kaevamisega
segipööratud turbapinnaga tervikul. Keskmine veetaseme sügavus 2023. a üle 60 cm, mis on
eelnenud aastatest ligikaudu 20 cm sügavamal. Üldilmelt on tegu kõdusooga, kus kased ja
männid on kuni 5-6 m kõrgused (Foto 3). Kuival turbapinnal domineerivad kanarbik ja raba-
karusammal, kelle katvused olid ainsana kõrgemad kui 10 %. Nendele lisaks esines püsiruu-
tudes vähesel määral ka sookail (Rhododendron tomentosum) ja harilik jõhvikas (Oxycoccus
palustris) paljal turbapinnal. Laiguti kattis turbapaljakuid kaselehtedest varis, palju oli ka
kuivanud kanarbikku. Varise/kulu katvus on küll aastati kõikunud, kuid ajas vähenev. Kulu
katvus võrreldes 2018. aastaga oli 2023. aastaks viie protsendipunkti võrra väiksem. Välja-
kute vahelised kraavid on paisutamata, kraavi põhjas vesi ja kraavide servas võib väikeste
laikudena näha ka raba-turbasammalt (Sphagnum medium), kraavikallastel kasvab raba-karu-
sammal ja kollakas tömptipp (Calliergon stramineum). Võrreldes 2018. aastaga on taimestiku
üldkatvus vähenenud ligikaudu kümne protsendipunkti võrra (75,8>65,8 %), seda eelkõige
kulu protsendi vähenemise arvelt (Tabel 1). Aja jooksul on soontaimede katvust suurenenud,
kuid sammalde katvus on vähenenud. Nii soon- kui ka sammaltaimede katvuste vähenemine
on toimunud peamiselt 2019. aastal, olles eelnevalt olnud küllalt stabiilne. Ajas on kõige
enam tõusnud kanarbiku katvus, mis kattis seireperioodi lõpus keskmiselt pool püsi-ruutude
pindalast (Foto 3). Katvus on pisut tõusnud ka sookailul, küündides siiski ka seireaja lõpuks
vaid 3 %-ni. Enim on katvus vähenenud aga raba-karusamblal, mis suures osas on üle kasva-
mas samblikega, aga tõenäoliselt vähendab tema katvust ka tugev kanarbikuvaris ja selle
suurenev katvus (Joonis 2). Liikide arvukus oli suurim 2022. aastal, mil registreeriti 14 taime-
liiki. Samal aastal mõõdeti ka kõige kõrgem keskmine veetase seireruutudes - 32,2 cm, sellele
eelneval aastal 47, 2 cm ja järgneval aastal oli vesi koguni 67,5 cm sügavusel. Seireaja lõpus
ei leitud ruutudest tõenäoliselt viimaste aastate suure põua tõttu väikeseid niiskuselembeseid
helviksamblaid, nagu näiteks kuu- ja ümaralehine peensammal.
Foto 3. Üldvaade korrastamata Kildema jääksoo B alale (vasakul) ja taimeruut B IV
kanarbiku domineerimise ning varisega kaetud turbapinnaga (paremal).
Joonis 2. Kolme sagedasema taimeliigi keskmised katvused Kildema seireala B ala
taimestiku püsiruutudes 2018-2023 aastal.
0
10
20
30
40
50
60
2018 2019 2021 2022 2023
K at vu s %
kanarbik sookail raba‐karusammal
Korrastamata jääksoos sõltuvad muutused taimestikus peamiselt taimeruutude asukohast
(kaugus taimestunud aladest ja kraavist, veetaseme sügavus, olemasolev taimestik jms) ning
sademete hulgast ja sesoonsest jaotusest ning võib tingimustest sõltuvalt toimuda eri suunas.
Samas jätkub sügava veetaseme tõttu turba lagunemine ja kasvuhoonegaaside (KHG)
emissioon ning alad on pikka aega tuleohtlikud. Mida kauemaks jääb jääksoo korrastamata,
seda raskem on seal luua tingimusi sootekke ja sootaimestiku taastumiseks.
ESS-SOO jääksoos tehti korrastamistööd 2021. a sügisel, sealhulgas osadel aladel turba-
samblafragmentide ja põhuga. K ja M ala asuvad jääksoost lõuna pool puisrabas, kuhu on
kaevatud kitsaid kuivenduskraave ja mõne meetri laiustest kraavidest turvast lõigatud.
Ess-soo K ja M alade I-III taimestiku seireruudud asuvad vanades turbaaukudes ning IV-VI
seireruut nende vahelistel kuivematel tervikutel mändide ja puhmastega (Foto 4). Tervikutel
kasvavad 4-6 m kõrgused männid ja nende all kanarbik, sookail ja teised puhmastaimed,
turbaaukudes peamiselt hõre turbasammal (Sphagnum fallax) ja pudev turbasammal (Sphag-
num cuspidatum). K ja M aladelt on raiutud puid ja sinna on tehtud madalaid turbavalle, kuid
kitsad kuivenduskraavid on paisutamata. Veetase on seire teostamise ajal juunis kõikunud
kümnekonna sentimeetri piires. Turbaaukudes oli see 2023. aastal valdavalt 2-10 cm süga-
vusel, kuid 2022. aastal olid M ala I-III ruudud vee all. Tervikutel oli veetase 13-33 cm
sügavusel. Taimestiku üldkatvus K ja M alal on keskmiselt 63-100 % ning võrreldes 2018.
aastaga on suurem muutus toimunud K ala ruutudes, kus keskmine katvus on langenud 96 %-
lt 63%-ni (Tabel 2, Foto 5). Peamiselt on see tingitud soontaimede (58 >38 %), sh kanarbiku
ja tupp-villpea katvuse vähenemisest. M alal on nii turbaaukudes kui ka tervikutel küllalt
stabiilse üldkatvuse juures (vastavalt 97 ja 87 %) vähenenud soontaimede ja suurenenud
sammaltaimede katvus. Katvuse suurenemist täheldati nii kuivematel tervikutel domineeriva
hariliku palusambla puhul kui ka turbaaukuse kasvava pudeva turbasambla puhul (Joonis 3).
Domineerivateks liikideks (katvus vähemalt 10%) on soontaimedest kanarbik ja sammaldest
palusammal ning pudev turbasammal. Pudeva turbasambla katvus oli viimsel seirel pisut lan-
genud, tõenäoliselt pika kuivaperioodi tulemusena. 2023. a mõõdeti ka madalaim keskmine
veetase 21,2 cm, samas kui 2022. aastal olid M ala turbaaukudes paiknevad püsiruudud üle-
ujutatud. Mõlemad alad oli liikide poolest mitmekesised. Ruutudest registreeriti kokku 29
liiki, kõige vähem leiti liike 2018. aastal - 21 liiki; 2023. aastal 23 liiki.
Foto 4. Taimestiku seireruudud vanas turbaaugus Ess-soo K alal (vasakul) ning M ala tervikul
(paremal).
Tabel 2. Taimestiku keskmine üldkatvus, soon ja sammaltaimede katvus (%) Ess-soo K ja M
alade taimeruutudes (turbaaukudes – a, tervikul – t) 2018-2023 aastal.
2018 2019 2021 2022 2023 K a üldkatvus 100 100 100 100 100 K a soontaimede katvus 27 28 30 25 23 K a sammaltaimede katvus 91 95 96 91 89 K t üldkatvus 96 88 63 60 63 K t soontaimede katvus 58 53 40 42 38 K t sammaltaimede katvus 38 42 43 37 38
M a üldkatvus 100 98 98 89 97
M a soontaimede katvus 47 32 33 17 22
M a sammaltaimede katvus 73 82 82 73 90
M t üldkatvus 88 85 87 87 87
M t soontaimede katvus 52 45 43 45 40
M t sammaltaimede katvus 47 57 60 65 74
Joonis 3. Ess-soo K ja M aladel domineerivate taimeliikide keskmised katvused püsiruutudes
2018-2023 aastal.
0
5
10
15
20
25
30
35
40
45
2018 2019 2021 2022 2023
K at vu s %
kanarbik tupp‐villpea
palusammal pudev turbasammal
Foto 5. Ess-soo K ala taimestiku seireruut K IV 2019. (vasakul) ja 2023 aastal (paremal).
Vanade turbaaukude taimestik on kõrge veetaseme ja piirnevate alade sootaimestiku tõttu juba
seireperioodi alguseks hästi taastunud. Taimestiku üldkatvus on vaid veidi alla 100 %, sh
sammaltaimedest peamiselt turbasammaldel < 90 %, mis on võrreldav looduslike rabadega. K
ja M alade, eriti tervikute, niiskustingimusi oleks saanud sootaimedele soodsamaks muuta
kitsaste kuivenduskraavide paisutamisega ning turbavallidega turbaaukudes.
Ess-soo jääksoo A, B, C, H ja J alad endisel freesturba tootmise alal on võrdluseks jäetud,
kus väheste muutuste tõttu niiskustingimustes toimub taimestumine spontaanselt ja taimestiku
üldkatvus ulatub neil aladel keskmiselt 40 protsendini (Tabel 3) olles asukohast sõltuvalt väga
varieeruv (19-60 %). Aladel on liike vähe ja suuremaid katvused moodustavad üksikud liigid.
Tabel 3. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Ess-soo jääksoo
korrastamata võrdlusalade A, B, C, H ja J taimestiku seireruutudes 2018-2023 a.
2018 2019 2021 2022 2023 A üldkatvus 9 11 9 12 22.4
A soontaimede katvus 9 9 8 12 19.8
A sammaltaimede katvus 0.2 0 0 0 0.2
B üldkatvus 33 37 38 35 60
B soontaimede katvus 21 26 29 23 26
B sammaltaimede katvus 13 12 10 11 20
C üldkatvus 43 43 46 42 49
C soontaimede katvus 25 21 30 33 35
C sammaltaimede katvus 5 20 19 10 12
H üldkatvus 6 8 13 16 22
H soontaimede katvus 7 4 12 15 21
H sammaltaimede katvus 0.5 0.5 0.7 1 1
J üldkatvus 20 23 21 28 41
J soontaimede katvus 8 9 25 16 21
J sammaltaimede katvus 13 16 16 17 20
Ess-soo A alal moodustavad taimestiku üldkatvusest suurema osa tupp-villpea mättad, üksi-
kud kased on 1-3 m kõrgused (Foto 6, vasakul). Veetase keskmiselt 32 cm sügavusel. Välja-
kute pind valdavalt taimestumata, kuivusest lõhenenud turbapinnaga. Väljakute vahelised
kraavid on paisutamata ja vett täis, kuid taimestunud väga vähesel määral. Taimestiku üld-
katvus A alal on suurenenud ligikaudu kaks korda, eelkõige sinikapuhmaste katvuse laiene-
mise tulemusel kahes püsiruudus. Samas näiteks A I taimeruudus on muutused paljal turba-
pinnal minimaalsed, vaid kasevõrse seireruudu nurgas on suuremaks kasvanud (Foto 7).
Sinikas (Vaccinium uliginosum ) ongi peamine taimeliik A-ala taimeruutudes. Kõikide teiste
väheste alalt registreeritud liikide katvused ei ületa ruutudes viite protsentigi. Kokku leitu alalt
vaid seitse taimeliiki, neist üks samblaliik – raba-karusammal, kasvas ainult ühes ruudus
katvusega vaid 1 %. Ess-soo B ala on taimestunud paremini, kuid samuti ebaühtlaselt.
Kumera turbapinnaga väljakutel asuvatest taimeruutudest on peamiselt kanarbiku ja raba-
karusamblaga taimestunud 70-80 % (Foto 6, paremal). Turbamudaga täitunud kraavilohud on
arvatavalt niiskustingimuste suurest kõikumisest (üleujutusest kuni turba pragunemiseni
põuaperioodidel) praktiliselt taimestikuta. Väga vähe on taimestunud ka V-VI taimeruudu
ümbrus, millest ca 80 % katab paljas turbapind. Paarikümne ruutmeetri ulatuses on sammal-
dega paremini taimestunud vaid taimeruut B VI ja selle lähiümbrus (Foto 8). Kuigi see väljak
on teistest madalam ja niiskem, takistavad spontaanset taimestumist just kõikuvad ja seetõttu
ebasoodsad niiskustingimused.
Foto 6. Üldvaade 2023. a. juunis Ess-soo A-alale III taimeruudu juurest (vasakul) ja B-alale II
taimeruudu juurest (paremal).
Foto 7. Palja turbapinna ja väga väikeste muutustega taimestikus seireruut A I Ess-soos 2018.
(vasakul) ja 2023. aastal (paremal).
Foto 8. Ess-soo B ala 2018. aastal praktiliselt taimestikuta B VI taimeruut (vasakul) on
2023.ks aastaks tupp-villpeaga täis kasvanud (paremal).
Ess-soo C alal on kraavid tupp-villpeaga praktiliselt täis kasvanud, kuid väljakud on taimes-
tunud väga ebaühtlaselt (Foto 9, vasakul). Väljakutest suuremat osa katab paljas turbapind
ning kanarbiku ja tupp-villpea mättad katavad vaid kuni 20 %. Kohati on näha hariliku
jõhvika (Oxycoccus palustris) väätide kiiret laienemist paljale turbapinnale. Veidi võivad need
ka niiskustingimusi stabiilsemaks muuta ja soodustada teiste taimeliikide kasvama hakkamist.
H ala väljakud on taimestunud (kanarbik, sinikas, raba-karusammal) vaid kuni 10 %, kuna
enamuse väljakutest katab paljas turbapind üksikute taimedega (Foto 9, paremal). Vaid veidi
paremate niiskustingimustega paisutatud kraavides kasvavad tupp-villpead. J ala taimestu-
mine on sarnaselt H alale vähene. Niiskemad kraaviservad on kohati kaetud valge tupp-villpea
lennukarvadega seemnetega, mis soodsates tingimustes võivad kasvama hakates taimestiku
katvust kiiresti suurendada. Paljas turbapind väljakutel mineraliseerumisest tuhkjas, mis ei ole
taimelevistele kasvama hakkamiseks soodne.
Foto 9. Üldvaade 2023. a juunis Ess-soo C alale III taimeruudu juurest (vasakul) ja H alale II
taimeruudu juurest (paremal).
Seireperioodi 2018-2023 jooksul on kõigi võrdlusalade taimestiku üldkatvus suurenenud
ligikaudu 2-3 korda, ulatudes B alal 60 %-ni. Vaid C alal on suurenemine olnud vaid 43>48
%. Jääksoode spontaanne taimestumine on turbatootmise käigus eemaldatud taimestiku ja
idanemisvõimeliste taimeleviste eemaldamise, kõikuvate ja ebasoodsaate niiskustingimuste ja
mitme teise teguri tõttu aeglane ja fragmentaarne (Triisberg et al. 2011, 2013). Enamasti on
esimeseks taimeliigiks tuulleviv tupp-villpea, mis võib katvust kiiresti suurendada, kuid
veetaseme langedes sügavamale kui ca 40 cm ka ära kuivada. Soodsamates tingimustes
jääksool tärganud ja kasvama hakanud taimed saavad edasi levida nii lähileviga kui ka
vegetatiivselt ja parandada niiskustingimusi ka teisele liikidele ning seetõttu võivad mõned
laigud olla ka küllalt hästi taimestunud.
Jääksoode korrastamisel nende kiiremaks taimestumiseks ning sooökosüsteemi talitluse taas-
tumiseks on aktiivsetest korrastamismeetoditest andnud suurtel pindaladel häid tulemusi
Kanadas välja töötatud „The Moss Layer Transfer Technique“ (MLTT, Quinty, Rochefort,
2003, Karofeld, 2011), mida kasutati ka Ess-soo ja Maima jääksoo osade korrastamisel.
Selleks kooritakse pindmine mineraliseerunud jääkturba kiht, väljakud tasandatakse, korras-
tatavale alale laotatakse doonoralalt kogutud turbasamblafragmente ja kaetakse põhust
multšiga ning täidetakse turbaga või paisutatakse kraavid. Eestis kasutati seda meetodit jääk-
soo korrastamisel suurel pindalal ELF poolt Palasi jääksoo korrastamiseks Lääne-Virumaal
(Karofeld, 2021). MLTT-d järgides tehti turbasamblafragmentide ja põhuga 2021. a sügisel
korrastustööd Ess-soo E, F ja G aladel. Nendel aladel oli taimestiku üldkatvus püsiruutude
keskmiste põhjal korrastamise eelselt vaid 7 - 13 % (Tabel 4). Korrastamise käigus eemaldati
pindmine turbakiht koos taimestikuga ja väljakute pind tasandati. Seega tuleks muutusi
taimestikus vaadata alates 2022. aastast ning 2023. aasta oli alles teine vegetatsiooniperiood
pärast korrastamist ja ei anna veel selget pilti korrastamise edukusest. Taimestiku üldkatvus
püsiruutude keskmiste põhjal oli kõikidel aladel 2023. aasta suvel ligikaudu 15%, kuid põuase
suve tõttu väiksem kui eelmisel suvel (Joonis 5). Jääksoode edukaks korrastamiseks oleks
vaja pikemaajalist seiret, et selgitada, kas taimestiku katvus hakkab ajas suurenema ja millised
tegurid seda kõige rohkem mõjutavad.
Ess-soo E alal katab põhk ja nende all vähesed kuivanud taimefragmendid 30-40 % (Foto 10,
vasakul). Väljakute põhjapoolsed osad on niiskemad ja praktiliselt palja turbapinnaga. Tai-
mestiku keskmine üldkatvus taimeruutudes on 18 % (Foto 10, paremal). Korrastamise eelse
ajaga võrreldes on taimestiku üldkatvus tõusnud üle kahe korra, kuid on jätkuvalt väike, alla
20 % (Tabel 4) ja peamiselt on tegu kuivanud samblafragmentidega. Taimefragmentide
kasvama hakkamine sõltub oluliselt sademete hulgast ja jaotusest. Põua tingimustes võivad
nad palja turbapinna kõrge temperatuuri ja ultraviolettkiirguse tõttu hukkuda, kuid soodsate
niiskus-tingimuste korral veel kasvama hakata. Selle kindlakstegemiseks võeti Ess-soost
kaasa turbasamblafragmente, et selgitada nende kasvama hakkamise võime neile loodud
soodsates niiskustingimustes. F alal katab põhk ja selle all kuivanud taimefragmendid kuni 50
% (Foto 11, vasakul). Kuid MLTT-ga korrastatud aladest selles taimestiku keskmine katvus
püsiruutudes langenud 20 %-lt 15,8 %-ni (Foto 11, paremal). Katvusehinnangud on väikse-
mad just sammalde osas, mida võis mõjutada ka pikk kuivaperiood suve alguses. Turbapind
on kuivusest lõhenenud. Vaid mõnes kohas paisutatud niisketes kraavipõhjades kasvab hõre-
dalt tupp-villpead.
Tabel 4. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Ess-soo jääksoo
turbasamblafragmentide ja põhuga (MLTT-ga) korrastatud E, F ja G ala taimestiku seireruu-
tudes 2018-2023 aastal.
2018 2019 2021 2022 2023 E üldkatvus 7 3 7 16 18 E soontaimede katvus 4 3 3 1 1 E sammaltaimede katvus 2 2 4 16 17 F üldkatvus 13 16 16 20 16 F soontaimede katvus 5 7 6 1 4 F sammaltaimede katvus 7 9 11 20 15 G üldkatvus 7 7 8 17 17 G soontaimede katvus 10 6 8 2 1 G sammaltaimede katvus 0.5 0.3 0.5 17 14
Foto 10. Turbasamblafragmentide ja põhuga korrastatud Ess-soo E ala 2023. a juunis
(vasakul) ja E ala V taimeruut (paremal).
Foto 11. Turbasamblafragmentide ja põhuga korrastatud Ess-soo F ala 2023. a juunis
(vasakul) ja F ala IV taimeruut (paremal).
Ess-soo G alal katab põhk ja selle all kuivanud taimefragmendid küllalt ühtlaselt ligikaudu 50
% (Foto 12, vasakul). Alal on vaid üksikud väikesed taimestunud laigud tupp-villpea ja kuni 1
m kõrguste kaskedega. Väljakute pind on kuivusest tingituna kohati kuni 20 cm sügavuselt
lõhenenud. Taimestiku keskmine üldkatvus on korrastamisele eelnevaga võrreldes tõusnud üle
kahe korra peamiselt sammaltaimede (turbasamblafragmentide) tõttu, kuid võrreldes 2022.
aastaga jäänud samale tasemele (17 %). Valdavalt on aga tegu kuivade taimefragmentidega,
mille kasvama hakkamine sõltub niiskustingimustest (Foto 12, paremal).
Foto 12. Üldvaade turbasamblafragmentide ja põhuga (MLTT-ga) korrastatud Ess-soo G alale
2023. a juunis (vasakul) ja taimeruut G II (paremal).
Ess-soo korrastamisele on järgnenud põuased kevaded ja suved ning veetase ei ole tõusnud
prognoositud tasemeni ja korrastamise edukus on senini väiksem kui varasemalt sama metoo-
dikat rakendades jääksoo korrastamisel (Karofeld et al, 2016; 2020). MLTT-ga korrastatud
alad olid korrastamisele järgnenud kevadel üle ujutatud ning kuna sinna ei olnud tehtud turba-
valle, siis lainetuse tõttu suure veepinnaga alal kanti samblafragmendid ja põhk kõrgematele
aladele (Foto 13). Turbasamblafragmendid taluvad lühiajalist üleujutust, kuid kui need kan-
takse lainetusega kõrgematele aladele, neil pole põhu kaitset kõrge pinnatemperatuuri ja
ultraviolettkiirguse vastu, siis pikema kuivaperioodi jooksul nad hukkuvad ja ei ole veel selge
kui paljud ja mil määral on samblafragmendid veel võimelised kasvama hakkama.
Foto 13. MLTT-ga korrastamise järgselt üle ujutatud Ess-soo D ala 2022. a juunis.
Muutused taimestiku (üld-, soon- ja sammaltaimede katvus) katvuses Ess-soo võrdlusaladel
(alade A, B, C, H, J taimeruutude keskmine) ja turbasamblafragmentide ning põhuga korras-
tatud aladel (E, F, G) on esitatud Joonisel 5. Võrdlusaladel on taimestiku katvus aja jooksul
suurenenud. Korrastatud aladel see samuti suurenes vahetult korrastamise järgselt, kuid kui-
vusest tingituna on seal taimed hakanud kuivama ja 2023. aastal on katvused vähenenud ja on
väiksemad kui võrdlusaladel. Võrdlusaladel on katvused enamasti suuremad ja muutused
kiiremad, kuna kord kasvama hakanud taimed saavad katvust suurendada lähi- ja vegetatiivse
leviga. Korrastatud aladelt aga pindmise turbapinna koorimise ja väljakute tasandamisega
kogu taimestik eemaldati, korrastamisel kasutatud taimefragmendid ei ole aga ebasoodsate
niiskustingimuste tõttu hästi kasvama hakanud.
Joonis 5. Ess-soo jääksoo võrdlus- ja korrastatud alade taimeruutude keskmised katvused (üld,
soon- ja sammaltaimede) 2018-2023. aastal.
2023. a aprillis Ess-soo ja Maima jääksoos MLTT-ga korrastatud aladel fotografeeritud turba-
samblafragmentide laigud (Foto 14) pildistatakse uuesti oktoobris muutuste hindamiseks
nende seisundis (hukkunud, kuivanud, elujõuline) ja mõõtmetes. Samuti mõõdetakse mõõte-
vaiade pikkused turbasammalde pikkuskasvu selgitamiseks nii turbasamblafragmentidega
korrastatud kui ka võrdluseks piirnevates looduslikes rabades ning võrreldakse varasemate
tulemustega Eesti looduslikest rabadest ja korrastatud jääksoodest (Karofeld et al., 2020.
Foto 14. Fotod paremas (vasakul) ja halvemas seisus turbasamblafragmentidest (vasakul) samblafragmentide ja põhuga korrastatud Ess-soo jääksoos nende edasise arengu järgimiseks.
Samuti võeti 2023. a kevadel Ess-soost kaasa korrastamise käigus laotatud turbasambla-
fragmente ja hoitakse neid plastkarbis destilleeritud veega kastes optimaalsetes niiskustingi-
mustes, et selgitada nende kasvama hakkamise võimet.
0
10
20
30
40
2018 2019 2020 2021 2022 2023
Taimestiku keskmine katvus, %
Võrdlus üld Võrdlus soon Võrdlus samm
Korrast. üld Korrast. soon Korrast. samm
MAIMA jääksoos asusid taimestiku seireruudud kolmes eri ilmelises ja töötlusega osas: F ja
G alad vanade kuivenduskraavide ja turbaaukudega puisrabas, A, B, C, D ja H alad jääksool
ning E ala turbasamblafragmentide ja põhuga korrastatud jääksool.
Maima F ala I-III taimeruut asusid tervikul 1-5 m kõrguste mändidega, alusrindes peamiselt
kanarbik, tupp-villpea, küüvits, rabamurakas (Rubus chamaemorus) ja vaevakask (Betula
nana) (Foto 15, vasakul). Veetase keskmiselt 27 cm sügavusel. Võrreldes 2018. aastaga on
tervikul asunud taimeruutudes I-III taimestiku üldkatvus suurenenud 50>66,6 % ja seda pea-
miselt soontaimede (kukemari Empetrum nigrum, tupp-villpea ja rabamurakas) katvuse suure-
nemise tõttu (Tabel 5). IV-VI taimeruut asusid vanas piklikus turbaaugus, kus tupp-villpea
mätaste vahel moodustas kohati lausalise katte pudev turbasammal. Keskmine veetase 18 cm
sügavusel. Kõrgematel kanarbikumätastel kasvas veel küüvits, vaevakask, punane (Sphagnum
rubellum) ja pruun turbasammal (Sphagnum fuscum). Üksikud männid olid 1-3 m kõrgused.
Kõrge ja suhteliselt stabiilse veetaseme tõttu on taimestiku üldkatvus jätkuvalt 100 %, millest
enamuse moodustavad sammaltaimed, peamiselt turbasamblad, kuid ka soontaimede katvus
on suurenenud.
Foto 15. Maima jääksoo F ala tervik taimeruutudega I-III (vasakul) ja vana turbaauk taime-
ruutudega IV-VI (paremal).
Tabel 5. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Maima jääksoo
F ja G alal I-III (tervikul) ning IV-VI taimeruudus (turbaaugus) 2018. ja 2023 aastal.
2018 2023 2018 2023 F I-III üldkatvus 50 67 G I-III üldkatvus 38 62 F I-III soontaimede katvus 43 55 G I-III soontaimede katvus 33 34 F I-III sammaltaimede katvus 5 6 G I-III sammaltaimede katvus 6 3 F IV-VI üldkatvus 100 100 G IV-VI üldkatvus 76 87 F IV-VI soontaimede katvus 11 25 G IV-VI soontaimede katvus 37 16 F IV-VI sammaltaimede katvus 100 98 G IV-VI sammalt. katvus 40 54
Maima G ala sarnaneb taimestiku liigiliselt koosseisult ja üldilmelt F alaga. Ka püsiruutude
paigutus järgib F-ala ruutude paigutust: G-ala esimesed kolm ruutu (I-III) asusid tervikul, kus
hõreda puistu moodustasid 4-5 m kõrguste männid ning 1-1,5 m kõrgused kased. Alusrindes
peamiselt kanarbiku mättad, sookail, rabamurakas, nende vahel paljas kuiv turbapind ning
kuivematel turbatükkidel samblikud (Foto 16, vasakul). Keskmine veetase 61 cm sügavusel.
Taimestiku keskmine üldkatvus on suurenenud 61,6 protsendini (Tabel 5). Suurema katvusega
taimeliikideks tervikul on kanarbik ja rabamurakas. G IV-VI taimeruut asusid vanas piklikus
turbaaugus, mis olid laiemad kui F alal. Keskmine veetase 16 cm sügavusel. Tupp-villpea
mätaste ja kanarbiku-ribade vahel turbasamblad (peamiselt punane turbasammal), liikidest
veel valge nokkhein (Rhynchospora alba), rabamurakas ja männitõusmed (Foto 16, paremal).
Võrreldes 2018. aastaga on taimestiku üldkatvus suurenenud ligikaudu kümme protsendi-
punkti, peamiselt sammaltaimede katvuse suurenemise tulemusel, mis aja jooksul on kasva-
nud 15 protsendipunti võrra (40 % 2018. a ja 54 % 2023. a). Maima F ja G alal ei ole vana-
dele piklikele turbaaukudele turbavalle tehtud, et niiskustingimusi ühtlustada. G-ala tervikutel
on katvus suurenenud aja jooksul kanarbikul ja ligikaudu kolm korda kõrgem oli 2023. aastal
ka samblike üldkatvus võrreldes seire algusega (2018), kuigi kahel viimasel aastal on märgata
samblike katvuse väikest langust (Joonis 6). Katvus on suurenenud ka kukemarjal, mis püsi-
ruutudest registreeriti esimest korda 2019. aastal. Sammalde katvus on olnud kogu seire-
periood väike ja sagedasema liigi – hariliku kaksikhamba – katvus on aja jooksul pigem
vähenenud.
Joonis 6. Maima jääksoo G-ala tervikutel (püsiruudud I-III) esinevate sagedasemate
taimeliikide katvused seireperioodil (2018-2023).
0
10
20
30
40
50
60
2018 2019 2020 2021 2022 2023
K at vu s %
samblike ÜK kanarbik
kukemari harilik kaksikhammas
Foto 16. Maima G ala tervikul asuv taimeruut G II (vasakul) ja vanas turbaaugus asuv
taimeruut G IV (paremal).
Maima H alal katab tupp-villpea ja raba-karusambla mätaste vahel paljas turbapind ligikaudu
60 % (Foto 17, vasakul). Kased ja männid on 3-4 m kõrgused. Kraavid on lahti, nende kallas-
tel tihedam 5-6 m kõrgune kaskede riba. Niiskemates osades esineb alpi-jänesvill (Trichop-
horum alpinum), hallikas tarn (Carex canescens) ja raba-karusammal, laikudena, nende vahel
vähetaimestunud paljas turbapind. Alal esineb väljaspool püsiruute kraavi servas ka väikeseid
turbasamblalaike (Foto 17, paremal), mis soodsamate niiskustingimuste korral võiksid edasi
laieneda ka väljakute keskosa suunas. Võrreldes seireperioodi algusega 2018. aastal on taime-
ruutude keskmine üldkatvus tõusnud ligikaudu kolm korda ning seda peamiselt soontaimede
katvuse suurenemise tulemusel (Tabel 6). Ligikaudu neli korda on suurenenud ka kulu/surnud
taimeosiste katvus püsiruutudes (vastavalt 1,3 % 2018. a ning 8,5 % 2023. a). Sammalde
katvus on püsiruutudes olnud kogu seireperioodi jooksul väga väike, kuid võib täheldada
väike tõusu ka sammalde üldkatvuses, eelkõige pugu-kaksikhambakese katvuse suurenemise
tulemusel. Viimane on kolonistliku iseloomuga lühiealine süstikliik, esinedes peamiselt
häiringualadel, sealhulgas kuivendatud turbaväljakutel (Dieren 2001). Pugu-kaksikham-
bakest registreeriti vaid kahel aastal, 2020. ja 2023. Samas sagedaseim samblaliik - longus
pirnik, mis samuti on laia ökoloogilise amplituudiga vähenõudlik liik ja tavaline inimmõju-
listes kooslustes, oli üsna stabiilselt väga madala katvusega (Joonis 7).
Foto 17. Maima jääksoo H ala üldvaade II taimeruudu juurest (vasakul) ja turbasamblalaik
paljal turbapinnal (paremal).
Joonis 6. Maima jääksoo H-alal esinevate sagedasemate taimeliikide keskmised katvused
2018-2023 aastal.
Maima A alal on tupp-villpea ja kanarbiku mätaste vahel palju paljast turbapinda (Foto 18,
vasakul). Keskmine veetase 9 cm sügavusel. Teest kaugemal on niiskustingimused paremad ja
tupp-villpead ning pilliroogu rohkem. Suur osa alast on üle ujutatud, kased ja kanarbik on
kuivanud, püdelal turbapinnal tupp-villpea mätastega on raske liikuda (Foto 18, paremal).
Väljakute vahelised kraavid on paisutamata. Nende servades kasvab ahtalehine villpea (Erio-
phorum angustifolium) ja 4-5 m kõrgused kuivanud kased. Kraavides kasvab penikeel (Pota-
mogeton sp.). Taimeruudud on taimestunud ebaühtlaselt, üldkatvus varieerub taimeruuduti 7 –
80%-ni. Keskmine taimestiku üldkatvus 2018. aastal oli 50,3% ning see on ajas pigem vähe-
nenud. Siiski on pärast veetaseme stabiliseerumist märgata taas taimestiku üldkatvuse väikest
tõusu (Tabel 7). Taimestiku katvust on kindlasti mõjutanud ka püsiruutude vee alla jäämine
2021. aastal, mil mõnes ruudus mõõdeti vee sügavuseks üle 15 cm. Kõrge veeseisu tulemus-
ena kuivas seireruutudes tõenäoliselt ka osa sookailust ning vähenes märgatavalt ka kanarbiku
katvus (Joonis 7). Samas, pärast veetaseme stabiliseerumist on tupp-villpea katvus suurene-
mas. Valge nokkheina, mis on soodes just väga niiskete kasvukohtade liik, katvus on aga väga
kõikuv sõltuvalt niiskustingimustest. Sammalde katvus püsiruutudes on olnud kogu seire-
perioodi marginaalne, kuid aja jooksul vähenenud veelgi. Viimastel aastatel ei ole leitud,
küllap pikkade põuaperioodide tõttu, enam pisikesi helviksamblaid, nagu laiahõlmaline rikar-
dia (Riccardia latifrons) ja kahetipuline niitsammal (Cephalozia bicuspidata), mis on tavali-
sed liigid niiskel turbal. 2018. aastal esines väikese laiguna ka narmaslehist turbasammalt
(Sphagnum fimbriatum), kui hiljem ühtegi turbasamblaliiki ruutudest ei ole registreeritud.
Joonis 7. Maima jääksoo A-ala taimestiku soontaimede üld- ja peamiste soonatimeliikdie
keskmised katvused 2018-2023. aastal.
Foto 18. Üldvaade Maima jääksoo A alale palja turbapinna ning tupp-villpea ja kanarbiku
mätastega (vasakul) ning osaliselt üle ujutatud ja kuivanud kaskede ning kanarbikuga osale
(paremal).
0
10
20
30
40
50
60
2018 2019 2020 2021 2022 2023
K at vu
s %
soontaimede üldkatvus kanarbik
valge nokkhein tupp‐villpea
Tabel 6. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Maima jääksoo
A, B, C, D, ja H ala taimeruutudes 2018.-2023. aastal.
2018 2019 2020 2021 2022 2023 A üldkatvus 50 47 52 12 14 39 A soontaimede katvus 50 40 40 12 34 39 A sammaltaimede katvus 0,2 0,1 0,3 0 0 0 B üldkatvus 26 26 25 7 20 24 B soontaimede katvus 22 19 19 8 19 24 B sammaltaimede katvus 4 4 4 0,5 0 1 C üldkatvus 8 15 18 vee all 5 16 C soontaimede katvus 4 7 8 vee all 5 13 C sammaltaimede katvus 5 4 3 vee all 0,3 3 D üldkatvus 19 18 17 16 15 23 D soontaimede katvus 13 11 12 16 15 22 D sammaltaimede katvus 0,2 0,1 0,1 0 0 0,1 H üldkatvus 9 12 11 19 19 29 H soontaimede katvus 9 8 8 19 19 23 H sammaltaimede katvus 0,7 0,9 0,7 0 1 1
Maima B ala katab suuremas osas taimestumata, mudane turbapind tupp-villpea mätastega
(Foto 19, vasakul). Taimeruutudes keskmine veetase 6 cm sügavusel. Väljakute vahelised
kraavid on veega täitunud, kuid paisutamata, nende servades kuivanud kased. Niiskemates ja
üleujutatud osades domineerib pilliroog (Foto 19, paremal) ja väike vesihernes (Utricularia
minor). B-ala taimestiku üldkatvus püsiruutudes vähenes märgatavalt 2021. aastal, mil ruudud
olid vee all. Sel aastal registreeriti esmakordselt kahest püsiruudust väikese-vesiherne kogu-
mikud, mis järgmisel aastal ulatuslikeks vaibanditeks olid kasvanud, kuid 2023. aastal, mil
vesi oli langenud, oli vesiherne katvus juba märgatavalt väiksem. Teiste liikide puhul sellist
katvuse fluktueerimist ajas ei ole täheldatud. Pilliroo katvus on järsult tõusnud pärast vee-
pinna langemist, samas raba-karusambla katvus on tõenäoliselt just vahepealse üleujutuse
tagajärjel vähenenud (Joonis 8).
Joonis 8. Maima jääksoo B-ala peamistee taimeliikide keskmised katvused 2018-2023. a.
0
2
4
6
8
10
12
2018 2019 2020 2021 2022 2023
K at vu s %
pilliroog vesihernes raba‐karusammal
Foto 18. Maima B ala üldvaade tupp-villpea mätaste ja palja turbapinnaga (vasakul) ning üle
ujutatud B I taimeruut pillirooga.
Maima C alal domineerib samuti taimestumata mudane turbapind, laiguti pilliroogu ja üle-
ujutatud alasid (Foto 19, vasakul). Veetase taimeruutudes keskmiselt 6 cm sügavusel. Välja-
kute vahelistes kraavides ja nende servas kraav ja hallikas tarn ning laialehine hundinui
(Typha latifolia). Võrreldes 2018. aastaga on keskmine taimestiku üldkatvus suurenenud
ligikaudu kaks ning soontaimede katvus kolm korda, kuid on endiselt väike (Foto 19,
paremal). Üldkatvus on suurenenud samblike ning soontaimedest peamiselt kanarbiku arvel.
Foto 19. Üldvaade Maima jääksoo C ala üleujutatud osale (vasakul) ja taimeruut C IV
(paremal).
Maima D ala on taimestunud väga ebaühtlaselt. Ahtalehise villpea ja tupp-villpea ning alpi-
jänesvilla mätaste vahel on palju taimestumata turbapinda (Foto 20, vasakul ). Keskmine
veetaseme sügavus taimeruutudes oli 2023. aastal 19 cm. Kõrgema veetasemega niiskemad
alad on paremini taimestunud ning kõige madalamad alad on ajuti ka üle ujutatud (Foto 20,
paremal). Kuigi võrreldes seireperioodi algusega on sammaltaimede katvus suurenenud
kümnekonna protsendipunkti võrra, siis taimestiku üldkatvuses on muutused väikesed.
Foto 20. Üldvaade Maima jääksoo D ala palja turbapinnaga kuivemale (vasakul) ning
niiskemale ja paremini taimestunud osale (paremal).
Maima E ala korrastati turbasamblafragmentide ja põhuga 2020. a sügisel. Selleks eemaldati
oksüdeerunud pindmine turbakiht, pind tasandati ja väljakute vahelised kraavid täideti
turbaga. Põuase 2023. aasta kevad-suve järgselt oli turbapind väga kuiv, kuivanud taimefrag-
mendid ja põhk katsid kuni 30 %. Tupp-villpead kasvas vaid kraavi servades (Foto 21,
vasakul). Korrastamise eelselt oli keskmine taimestiku üldkatvus taimeruutudes jõudnud ligi-
kaudu 30 protsendini. Pärast turbapinna koorimist ja korrastamist oli ala üle ujutatud ning
samblafragmendid ja põhk kanti lainetusega kõrgematele aladele ja on ebaühtlaselt jaotunud.
2023. aastaks on taimestiku üldkatvus jõudnud 7,3 %-ni (Tabel 7, Foto 21, paremal) ja taime-
fragmentide kasvama hakkamine sõltub lähiaja ilmastikust ning niiskustingimustest.
Tabel 7. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Maima jääksoo
samblafragmentide ja põhuga korrastatud E ala taimeruutudes 2018.-2023. aastal.
2018 2019 2020 2021 2022 2023 E üldkatvus 26 29 28 0 5 7 E soontaimede katvus 23 23 18 0 3 7 E sammaltaimede katvus 3 4 4 0 2 2
Foto 21. Üldvaade samblafragmentide ja põhuga (MLTT-ga) korrastatud Maima jääksoo E
alale (vasakul) ja palja turbapinnaga ning põhujäänustega taimeruut E I (paremal).
Muutused taimestiku (üld-, soon- ja sammaltaimede katvus) katvuses seireruutudel Maima
jääksoo võrdlusaladel (alade A, B, C, D taimeruutude keskmine) ja turbasamblafragmentide
ning põhuga korrastatud E alal on esitatud Joonisel 9. Taimestiku katvus oli väikseim 2021.
aastal ja on seejärel nii võrdlus- kui ka korrastatud aladel hakanud tõusma. Võrdlusaladel on
muutus kiirem ja katvus suurem, kuna sealt ei ole spontaanselt tekkinud taimestikku erinevalt
korrastatud aladest eemaldatud ning taimed said kiiremini suurendada katvust lähi- ja vege-
tatiivse leviga.
Joonis 9. Maima jääksoo võrdlus- ja korrastatud alade taimeruutude keskmised katvused (üld,
soon- ja sammaltaimede) 2018-2023. aastal.
Ess-soo korrastatud alal on tulemust mõjutanud põuased suved, kuid ka kuivenduskraavide
paisutamata jätmine mõnedel ning üleujutused ja lainetuse mõju teistel aladel.
0
5
10
15
20
25
30
35
2018 2019 2020 2021 2022 2023
Taimestiku keskmine katvus, %
Võrdlus üld Võrdlus soon Võrdlus samm
Korrast. üld Korrast. soon Korrast. samm
LAIUSE JÄÄKSOOS tehti korrastustööd 2019. aastal, mil paisutati väljakute vahelisi kraave
ja ehitati turbast valle vee hoidmiseks korrastataval alal. Jääksoo lõunapoolne osa on püsivalt
üle ujutatud.
Laiuse C alale on tehtud turbavalle, mille vahel on niisked, osalt üle ujutatud alad kraavi- ja
hallika tarna ning nende all kohati lausalise, peamiselt hõreda turbasambla vaibaga (Foto 22,
vasakul). Taimestiku seireruudud asuvad teele lähemal veidi kõrgemal alal. Jääkturba kihi
paksus 1,7-st kuni üle 2 m. Taimestiku üldkatvuse suurenemine võrreldes 2018. aastaga
(59>78 %) on toimunud sammaltaimede ligikaudu kahekordse katvuse suurenemise tõttu
(Tabel 8), mis osutab paremate niiskustingimuste tekkele. Näiteks suurenenud katvusega
tugev vesisirbik (Warnstorfia exannulata) on hüdro-hügrofüüt, mis kasvab just ajuti üleuju-
tatud kasvukohtades ja vees. Kõrgema veetasemega aladel moodustavad turbasamblad (pea-
miselt pudev turbasammal) tarnade all juba lausalise katte (Foto 22, paremal) ning veetaseme
tõusul saavad nad kasvupinda veelgi suurendada. Veel 2022. aastal registreeritud invasiivse
samblaliigi võõr-kõverharjaku (Campylopus introflexus) laik C III taimeruudust on arvatavalt
veetaseme tõusu tõttu praktiliselt kadunud. Eestis ongi seda samblaliiki leitud valdavalt vaid
mahajäetud jääksoodest.
Tabel 8. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Laiuse jääksoo
C, A ja D ala taimeruutudes 2018.-2023. aastal.
2018 2019 2020 2021 2022 2023 C üldkatvus 59 58 48 61 72 78 C soontaimede katvus 31 31 8 20 19 27 C sammaltaimede katvus 33 37 34 46 57 64 A üldkatvus 86 90 91 88 83 81 A soontaimede katvus 43 40 36 37 32 31 A sammaltaimede katvus 49 68 61 60 63 65 D üldkatvus 54 54 26 25 25 53 D soontaimede katvus 41 28 9 14 20 44 D sammaltaimede katvus 14 16 0 15 15 15
Foto 22. Laiuse jääksoo C ala niiskem osa tupp-villpea ja tarnadega (vasakul) ning lausaline
turbasammalde kate tarnade all C ala niiskemates osades.
A ala tervikutel on turbakihi paksus üle 2 m. Seal kasvavad 2-5 m kõrgused männid ja kased,
paljal turbapinnal tupp-villpea, kanarbiku, raba-karusambla, kukemarja ja sinika mättad ning
laigud (Foto 23, vasakul). Kraavid vett on täis ja väikese vesihernega, servades tarnad. Võrrel-
des 2018. aastaga on taimestiku üldkatvus veidi langenud, kuid sammaltaimede (peamiselt
raba-karusambla) katvus ligikaudu kolmandiku võrra suurenenud (49>65 %).
D alal on kitsaste kuivade tervikutega risti tehtud korrastamisel turbavallid, milliste vaheline
ala on osaliselt üle ujutatud ka põuase kevade järel (Foto 23, paremal). Kraavid on vett täis,
kuid taimestunud peamiselt servadest. D alal vaid VI ruut asub tervikul, teised aga enamasti
üleujutatud alal. Suhteliselt soodsatest ja stabiilsetest niiskustingimustest tingituna on ka muu-
tused taimestiku katvuses olnud väikesed. Laiuse jääksoo C ja D alal on korrastamise järgselt
niiskustingimuste muutumisest tingitud uued taimekooslused alles kujunemisel ja taimestiku
üldkatvus pärast 2020. aastat tõusmas, seda peamiselt sammalde üldkatvuse suurenemise
tulemusel. Muutused taimestikus ei toimugi mitte niivõrd katvuses kui liigilises koosseisus.
Erinevate taimeliikide katvused kõiguvad aastati sõltuvalt niiskustingimustest. Aja jooksul on
erinevates seireruutudes tõusnud nii kuivemaid kasvukohti eelistavate samblike ja sookailu
katvus, kui ka vesiseid elupaiku asutava pudeva turbasambla ja pilliroo katvused. Aja jooksul
on Laiuse seirealadelt kadunud ka 2018. aastal leitud liike nii sammalde osas (näit. teravtipp
ja südajas tömptipp) kui ka mõned soontaimeliigid, näiteks roomav tulikas ja sale tarn. Pärast
veetaseme tõstmist 2020. aastal on seireruutudest leitud niiskuselembeseid liike nagu alpi-
jänesvill (Trichophorum alpinum) ja soovildik.
Foto 22. Üldvaade Laiuse jääksoo A (vasakul) ja D alale (paremal) 2023. a. juunis.
Laiuse jääksoo kõrgema veetaseme aladel on juba küllalt palju turbasamblaid ning teisi soo-
taimeliike ja niiskustingimuste jätkuval paranemisel ja stabiilsemaks muutumisel hakkab
nende kasvuaala ja katvus veelgi suurenema ja kiireneb turba teke. Kasuks oleks tulnud
väljakute vaheliste kraavide sulgemine turbapaisudega mõnekümne meetri järel. Laiuse
jääksoo taastaimestumine on korrastamise järgselt läinud küllaltki edukalt ning veetaseme
tõustes ja niiskustingimuste stabiilsemaks muutumisel suureneb soo- ja veetaimede arvukus
ning katvus veelgi ja arenevad tingimused soo talitluse, sh turbatekke taastumiseks.
KÕIMA jääksoos tehti korrastamistööd 2019. aastal, mille käigus ehitati turbavalle ja paisud
mõnele kraavile. B ja D ala asusid looduslähedases lagerabas väheste mändidega. A ala asus
kuivendatud puisrabas ja C ala tiheda kraavitusega turbatootmiseks ettevalmistatud alal.
Kõima jääksoo A ala asub kuivendatud puisrabas (Foto 23). Mändide ja mõne kase kõrgus
on kuni 4-5 m. Taimestiku alusrindes domineerivad kanarbik, tupp-villpea, küüvits, murakas,
turbasamblad (pruun ja punane) ja samblikud. Kraavid on paisutamata, nendes kasvavad tupp-
villpea, raba ja pudev turbasammal. Keskmine veetase 41 cm sügavusel. Taimestiku iseloomu
mõjutab kaugus kraavist ja niiskustingimused, millest tulenev mitmekesisus on näha ka taime-
ruudus A V (Foto 23, paremal). Seireperioodi jooksul on taimestiku üld- ja sammaltaimede
katvuses toimunud vaid väikesed muutused, soontaimede katvus on ligikaudu kolmandiku
võrra suurenenud (Tabel 9). Kõige enam on suurenenud kanarbiku katvus, mis 2018. a oli 15
% ning 2023. aastal 30 %. Suurenenud on ka samblike üldkatvus püsiruutudes, tõustes 14,8
%-lt 2018. aastal 23,3 %-ni 2023. aastal. Nii kanarbik kui ka samblikud on pigem kuivemate
kasvukohtade asukad.
Foto 23. Üldvaade Kõima jääksoo A alale (vasakul) ja taimestiku seireruut A V (paremal).
Tabel 9. Taimestiku keskmine üldkatvus, soon- ja sammaltaimede katvus (%) Kõima jääksoo
A, B, C ja D ala taimeruutudes 2018.-2023. aastal.
2018 2019 2020 2021 2022 2023 A üldkatvus 70 68 66 66 68 65 A soontaimede katvus 28 32 20 34 38 43 A sammaltaimede katvus 44 51 47 43 42 41 B üldkatvus 92 88 89 93 92 92 B soontaimede katvus 40 41 26 24 26 31 B sammaltaimede katvus 86 84 82 81 81 79 C üldkatvus 66 55 52 61 61 63 C soontaimede katvus 44 34 29 34 35 41 C sammaltaimede katvus 27 35 30 35 35 36 D üldkatvus 99 98 98 99 98 99 D soontaimede katvus 18 17 15 25 17 24 D sammaltaimede katvus 99 99 98 94 94 93
Kõima jääksoo C ala asub turbatootmiseks ettevalmistatud kraavitatud alal (Foto 24). I-III
taimeruut asuvad kuival tervikul kuni 3 m kõrguste mändidega, alusrindes peamiselt kuiv
kanarbik, tupp-villpea ja murakas. Taimeruudud IV-VI asuvad paisutamata kraavi põhjas, mis
kohati on vee all või kaetud turbamudaga. Kraavide ääres kasvab tupp-villpead, sookailu ja
vaevakaski (Betula nana), kraavi põhjas peamiselt tupp-villpea ja selle vahel raba ja pudev
turbasammal. Kõigi seireruutude keskmine taimestiku üld- ja soontaimede katvus on seire-
perioodi jooksul olnud küllalt stabiilne, kuid sammaltaimede osas veidi suurenenud (Tabel 9).
Sammaltaimede katvus seireruutudes oli väga varieeruv sõltuvalt asukohast. Võrreldes 2018.
aastaga on see tervikul, seireruutudes I-III, vähenenud (12,3>8 %) ning kraavis (seireruutudes
IV-VI) tõusnud (41,7>63,3 %) (Foto 25). Just kraavides võib täheldada väikest samblakatte
tõusu, seda eelkõige mõne turbasamblaliigi katvuse suurenemise tulemusel kraavis paikne-
vates ruutudes (Joonis 10).
Foto 24. Üldvaade Kõima jääksoo C alale.
Foto 25. Kõima jääksoo C ala III seireruut tervikul (vasakul) ja C VI kraavis (paremal).
Joonis 10. Raba- ja punase turbasambla keskmised katvused Kõima jääksoo C-alal 2018-
2023. aastal.
Kuna kraavitus Kõima jääksoo A ja C alal on toimunud juba mitukümmend aastat tagasi ja
kasvutingimused on praeguseks stabiliseerunud, siis on kujunenud tingimustele kohastunud
taimekooslustes muutused väikesed.
0
2
4
6
8
10
12
14
16
18
20
2018 2019 2020 2021 2022 2023
K at vu
s %
raba‐turbasammal punane turbasammal
Kõima jääksoo B ja D ala asuvad väheste mändidega lagerabas üksikute 0,5 m kaskedega.
Taimestiku alusrindes kanarbiku ja tupp-villpea mättad ning vaevakask, küüvits (Andromeda
polifolia), ümaralehine huulhein (Drosera rotundifolia), valge nokkhein ja sammaldest pea-
miselt pruun ja punane turbasammal (Foto 26). Keskmine veetase B alal 23 cm ja D alal 20
sügavusel. Nii Kõima jääksoo B kui ka D ala on taimestiku ja niiskustingimuste poolest
looduslähedased ja seetõttu on ka muutused taimestikus olnud väga väikesed, vaid mõne
protsendi piires (Tabel 10, Foto 27).
Foto 26. Üldvaade Kõima jääksoo B (vasakul) ja D alale (paremal).
Foto 27. Kõima jääksoo taimestiku seireruut B V (vasakul) ja D II (paremal).
Kõima jääksoo A ja C ala niiskustingimuste paranemisele, sootaimestiku ja soo talituse
taastumisele oleks kaasa aidanud kuivenduskraavide paisutamine turbast paisudega. Veetase
saaks kraavide paisutamise järgselt hakata tõusma ja aja jooksul suureneks sootaimede
osatähtsus ja katvus. Mida kauemaks jäävad kraavid paisutamata ja jääksood korrastamata,
seda kauem võtab aega kujunenud taimestiku asendumine sootaimestikuga ning jätkub KHG
emissioon.
LISA 1 RMK taimestik 2018-2023 KOOND
Viidatud kirjandus
Dieren, K. Distribution, ecological amplitude and phytosociological characterization of
European bryophytes. – Bryophytorum Bibliotheca 56, 3-289.
Karofeld, E. 2011. Tingimuste loomine taassoostumiseks. Kogemusi maailmast. Rmt.: Paal, J.
(koost. ja toim.) Jääksood, nende kasutamine ja korrastamine. Tartu, 111-130.
Karofeld, E., Müür, M., Vellak, K. 2016. Factors affecting re-vegetation dynamics of
experimentally restored extracted peatland in Estonia. – Environmental Science and
Pollution Research, 23, 13706-13717.
Karofeld, E., Kaasik, A., Vellak, K. 2020. Growth characteristics of three Sphagnum species
in restored extracted peatland. Restoration Ecology, 28 (6), 1574-1583.
Quinty, F., Rochefort, L. 2003. Peatland Restoration Guide. Second edition. Canadian
Sphagnum Peat Moss Accociation, New Brunswick Department of Natural Resources and
Energy.
Salm, J.-O., Meriste, M., Karofeld, E., Kohv, M. Kogemused Palasi jääksoo korrastamiselt
turbasamblafragmentide ja põlevkivituha laotamisega. Rmt: Soode taastamine: senised
kogemused teguloo “soode taastamine ja kaitse” näitel. Käsiraamat. Tartu, 2021, 56-62.
Triisberg, T., Karofeld, E., Paal, J. 2011. Re-vegetation of block-cut and milled peatlands: an
Estonian example. – Mires and Peat, 8, 1-14.
Triisberg, T., Karofeld, E., Paal, J. 2013. Factors affecting the re-vegetation of abandoned
extracted peatlands in Estonia: a synthesis from field and greenhouse studies. – Estonian
Journal of Ecology, 62 (3) 192-211.