This is a sel -a chi ed e sion o an o iginal a icle. This e sion
may di e om he o iginal in pagina ion and ypog aphic de ails.
Au ho (s):
Ti le:
Yea :
Ve sion:
Copy igh :
Righ s:
Righ s u l:
Please ci e he o iginal e sion:
CC BY 4.0
h ps://c ea i ecommons.o g/licenses/by/4.0/
Changes in sa elli e‐de i ed spec al a iables and hei linkages wi h ege a ion
changes a e pea land es o a ion
© 2024 The Au ho (s). Res o a ion Ecology published by Wiley Pe iodicals LLC on behal o Socie y o Ecological Res o a ion
Published e sion
Räsänen, Aleksi; Jan unen, Aapo; Isoaho, Aleksi; Ikkala, Lau i; Rana, Pa ez;
Ma ila, Hannu; Elo, Me ja
Räsänen, A., Jan unen, A., Isoaho, A., Ikkala, L., Rana, P., Ma ila, H., & Elo, M. (2024). Changes
in sa elli e‐de i ed spec al a iables and hei linkages wi h ege a ion changes a e pea land
es o a ion. Res o a ion Ecology, Ea ly View. h ps://doi.o g/10.1111/ ec.14338
2024
RESEARCH ARTICLE
Changes in sa elli e-de i ed spec al a iables and hei
linkages wi h ege a ion changes
a e pea land es o a ion
Aleksi Räsänen1,2,3 , Aapo Jan unen4, Aleksi Isoaho1,5, Lau i Ikkala5,6, Pa ez Rana1,
Hannu Ma ila5, Me ja Elo4,7,8
Remo e sensing (RS) can be an e ficien moni o ing me hod o assess he ecological impac s o es o a ion. Ye , i has been used
ela i ely li le o moni o pos - es o a ion changes in bo eal o es y-d ained pea lands, and pa icula ly he linkages be ween
changes in RS and plan species emain ague. To unde s and his gap, we u ilize da a om he Finnish pea land es o a ion
moni o ing ne wo k spanning 150 si es and a 10-yea pos - es o a ion moni o ing pe iod. We employ Bayesian join species
dis ibu ion models (Hie a chical Modeling o Species Communi ies) o s udy (1) he changes in op ical Sen inel-2 and Landsa
sa elli e spec al signa u es, (2) whe he he RS a iables imp o e p edic ions o ascula plan and moss species and unc ional
ype occu ence and co e , and (3) wha kinds o associa ions exis be ween RS a iables and plan species o unc ional ypes.
Ou esul s show ha pea land es o a ion inc eases he eflec ance o ed and nea -in a ed (NIR) bands in spa sely eed pine
mi e o es s and open mi es bu no in densely eed sp uce mi e o es s. Impac s on o he es ed RS a iables consis ing o
mois u e and g eenness indices a e less clea . Addi ionally, RS a iables inc ease species- o unc ional ype-specific p edic i e
powe only modes ly, and he e a e ew clea links be ween he changes in RS a iables and species o unc ional- ype occu -
ence and co e . We sugges ha ed and NIR eflec ance can be used as sa elli e-based indica o s o pea land es o a ion suc-
cess and u he s udies a e equi ed o de elop usable me hods o de ec ing species-specific changes wi h RS.
Key wo ds: b yophy es, join species dis ibu ion models, plan unc ional ypes, emo e sensing, sa elli e image y, ascula
plan s
Implica ions o P ac ice
•Sa elli e emo e sensing is sui able o moni o ing pos -
es o a ion changes in g ound ege a ion, land co e ,
and we ness in pea lands wi h ew o no ees, as ees
hampe isibili y o he g ound.
•High spa ial and empo al esolu ion emo e sensing com-
plemen s field wo k, and i can be used o scale field-
based knowledge o la ge a ea ex en s o o o he si es.
•I should be u he es ed whe he changes in eflec ance
can be used in ope a ional pea land es o a ion moni o -
ing and o which kind o changes he eflec ance changes
a e a ibu able.
•The e is a need o c oss- e iliza ion o esea che s’and
p ac i ione s’knowledge o de elop es o a ion ou come
indica o s ha a e ecologically meaning ul, ope a ionally
implemen able, and de ec able wi h emo e sensing.
In oduc ion
Many o he pea lands in no he n la i udes ha e been d ained o
acili a e o es g ow h and imbe p oduc ion o he o es y
indus y (Vasande e al. 2003). Howe e , his d ainage has
caused widesp ead and ha m ul en i onmen al impac s,
including loss o pea land species and habi a s, g eenhouse gas
emissions, and de e io a ion o wa e quali y in ecipien wa e
bodies (Chapman e al. 2003;U

ak e al. 2017; Nieminen
e al. 2018).
To e e se pea land deg ada ion, ecological es o a ion has
been conduc ed du ing he pas ew decades (Ande sen
Au ho con ibu ions: AR, ME concei ed and designed he esea ch; AJ, ME, AI
conduc ed he analyses; AR w o e he fi s d a o he manusc ip ; all au ho s
con ibu ed o he w i ing and discussed he wo k h ough all o i phases.
1
Na u al Resou ces Ins i u e Finland (Luke), Paa o Ha aksen ie 3, 90570, Oulu,
Finland
2
Geog aphy Resea ch Uni , Uni e si y o Oulu, PO Box 8000, FI-90014, Oulu, Finland
3
Add ess co espondence o A. Räsänen, email aleksi. asanen@oulu.fi
4
Depa men o Biological and En i onmen al Science, Uni e si y o Jy äskylä, PO
Box 35, Jy äskylä 40014, Finland
5
Wa e , Ene gy and En i onmen al Enginee ing Resea ch Uni , Uni e si y o Oulu, PO
Box 4300, FI-90014, Oulu, Finland
6
Geological Su ey o Finland, Teknologiaka u 7, 67101, Kokkola, Finland
7
Finnish En i onmen Ins i u e (Syke), Su on ie 9A, 40500, Jy äskylä, Finland
8
School o Resou ce Wisdom, Uni e si y o Jy äskylä, PO Box 35, 40014, Jy äskylä,
Finland
© 2024 The Au ho (s). Res o a ion Ecology published by Wiley Pe iodicals LLC on
behal o Socie y o Ecological Res o a ion.
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion
License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he
o iginal wo k is p ope ly ci ed.
doi: 10.1111/ ec.14338
Suppo ing in o ma ion a :
h p://onlinelib a y.wiley.com/doi/10.1111/ ec.14338/suppin o
Res o a ion Ecology 1o 15
e al. 2017). In he u u e, pea land es o a ion ac i i ies a e u -
he p ojec ed o inc ease globally. In he Eu opean Union (EU)
alone, he es o a ion law a ge s o es o e 90% o he deg aded
a ea o he ecosys ems, including pea lands be o e 2050
(Regula ion [EU] 2024/1991).
In o es y-d ained pea land si es, es o a ion includes filling
o damming o di ches and emo al o ees ha ha e g own
a e he d ainage (Haapaleh o e al. 2011). The isible changes
a e es o a ion include, e.g. (1) dec ease in ee co e ,
(2) eplacemen o di ches wi h flow-blocking s uc u es,
(3) inc ease in we ness and local wa e able le el, and
(4) changes in g ound ege a ion composi ion (Haapaleh o
e al. 2011). O hese changes, he fi s h ee occu almos imme-
dia ely a e es o a ion, while he changes in g ound ege a ion
a e slowe (Haapaleh o e al. 2011,2017; Menbe u e al. 2016).
The mos ecen s udies ha e indica ed ha du ing he fi s
10 yea s a e es o a ion, he e a e changes in he ege a ion:
pa icula ly he mo e common species o p is ine mi es s a o
colonize he es o ed si es, bu ege a ion in he es o ed si es
does no esemble ha o p is ine coun e pa s (Elo e al. 2024).
Typical ecological a ge s o es o a ion a e ela ed o he
e u n o o iginal pea land communi y s uc u e and unc ioning.
The success can be measu ed, e.g. h ough in en o ies o di e -
en axa, such as plan s (e.g. Haapaleh o e al. 2011,2017).
O e all, moni o ing es o a ion success is impo an o alida -
ing es o a ion me hods and ou comes bu also o imp o ing
ou unde s anding o pea land ecosys em changes and p o-
cesses. Ne e heless, adi ional field-based moni o ing
me hods equi e a conside able amoun o labo and o he
esou ces, and hey a e es ic ed o a limi ed numbe o poin s
ha a e no necessa ily ep esen a i e o he whole pea land in
ques ion. The e o e, cos -e ec i e and spa ially ex ensi e mon-
i o ing me hods a e equi ed, pa icula ly because o he inc eas-
ing amoun o es o a ion ac i i ies.
A po en ial solu ion o de ec ing changes o e la ge a eas
cos -e ec i ely is he u iliza ion o sa elli e emo e sensing
(RS), as i can p o ide high spa ial and empo al esolu ion
obse a ions o global land co e . The s udies so a ha e indi-
ca ed ha pa icula ly op ical sa elli e da a a e usable o ack-
ing changes in pea land we ness (Räsänen e al. 2022; Bu dun
e al. 2023; Isoaho e al. 2024) and land co e and ege a ion
such as habi a ypes and plan communi y s uc u e (Kola i
e al. 2022; Ball e al. 2023). The key s eng h o op ical sa elli e
image y is i s empo al a ailabili y: seamless and c oss-
compa able high- esolu ion da a has been a ailable since he
1980s (Wulde e al. 2022; Radelo e al. 2024).
In pea lands, pos - es o a ion RS assessmen s ha e mainly
ocused on acking changes in we ness (Räsänen e al. 2022;Bu -
dun e al. 2023; Isoaho e al. 2024), while changes in spec al sig-
na u es and ege a ion ha e gained less a en ion. The ew s udies
include he wo k by Ball e al. (2023) analyzing whe he he spec-
al signa u es o es o ed si es s a o esemble hose o p is ine
pea land a eas assessing he possibili y o using RS o de ec ing
changes in g ound ege a ion flo is ic g adien s ela ed o we ness
and p oduc i i y.
The lack o ocus on ege a ion changes has been e iden
o e all in RS s udies in pea lands, no jus hose ela ed o
es o a ion. This is su p ising gi en ha changes in ege a ion
composi ion and abundances o indi idual species a e consid-
e ed key indica o s o pea land es o a ion success (Haapaleh o
e al. 2011; Elo e al. 2024; Ky kjeeide e al. 2024). E en hough
b oad-scale pa e ns in habi a ype changes ha e been moni-
o ed (Kola i e al. 2022; S een oo den e al. 2022), mo e
de ailed analyses o empo al ege a ion changes ha e no been
conduc ed.
Despi e he lack o assessmen s abou empo al changes in eg-
e a ion, he e ha e been mul iple s udies mapping he spa ial pa -
e ns o ege a ion a a specific ime poin . Examples o
moni o ed ege a ion cha ac e is ics include plan communi ies
and flo is ic g adien s (Ha is e al. 2015; Räsänen e al. 2020b),
plan unc ional ypes (PFTs), such as sh ubs, o bs, g aminoids,
and mosses (Räsänen e al. 2020b;Pange al.2024), unc ional
ai s, such as lea -a ea index and plan nu ien con en
(Kalacska e al. 2015; Räsänen e al. 2020a), and he occu ence
and co e o single species (Kalacska e al. 2013;Pang
e al. 2024; Simpson e al. 2024). I can be hypo hesized ha he
empo al changes a e es o a ion in hese cha ac e is ics can be
moni o ed i he e a e sys ema ically collec ed long- e m moni o -
ing da a and i he scale o he changes is de ec able. Fu he mo e,
o e ealing he pos - es o a ion ege a ion succession, he RS
app oaches should simul aneously accoun o se e al changes
in land co e , including in ee co e and we ness.
We u ilize globally unique 10-yea be o e-a e con ol-
impac (c. . Ch is ie e al. 2020) Finnish pea land es o a ion
moni o ing ini ia i e da a spanning 150 si es ha belong o six
di e en pea land ypes (Elo e al. 2024). We use Bayesian join
species dis ibu ion models (O askainen e al. 2017; O askai-
nen & Ab ego 2020) ha can be used o assess plan communi y
change and he associa ions be ween di e en plan species,
PFTs, and RS a iables. Ou objec i e is o s udy how he
pos - es o a ion land co e changes in pea lands a e linked wi h
spec al signa u e changes and wha kinds o associa ions he e
a e be ween spec al and ege a ion changes in di e en pea -
land ypes and ea men s (p is ine, d ained, and es o ed). Ou
b oade objec i e is o con ibu e o he wo k de eloping RS-
based ecological es o a ion success indica o s (c. . Skidmo e
e al. 2021) ha can be used o au oma ic es o a ion success
analysis.
Ou specific esea ch ques ions a e as ollows:
(1) Wha is he e ec o pea land es o a ion on spec al signa-
u es in di e en pea land ypes?
(2) Do sa elli e image y a iables imp o e p edic ions o plan
species and PFT occu ences and co e s in es o ed,
d ained, and p is ine si es?
(3) Wha kinds o associa ions exis be ween RS a iables and
plan species and PFT occu ence/co e ?
Me hods
S udy Si es and Field Da a
We used da a om 150 si es belonging o he Finnish Me sähal-
li us Pa ks & Wildli e pea land es o a ion moni o ing ne wo k
(Fig. 1; desc ip ion in Elo e al. 2024). The si es in he ne wo k
Res o a ion Ecology2o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
a e loca ed h oughou Finland (60–68N, 21–31E; Fig. 1; Elo
e al. 2024) ac oss ele a ion and clima ic g adien s (Fig. S1).
They a e di ided in o six di e en pea land ypes based on hei
ege a ion: ich and poo sp uce mi e o es s, ich and poo pine
mi e o es s, and ich and poo open mi es. Fo each ype, he e
a e da a o 10 es o ed si es and 10 nea by loca ed p is ine coun-
e pa s, wi h he excep ions being 9 +9 si es o poo open
mi es and 11 +11 si es o ich open mi es. Addi ionally, he e
a e 30 d ained con ol si es (4–6 si es pe ype). The es o ed
si es ha e been d ained o o es y be ween he 1960s and
1970s and subsequen ly es o ed be ween 2007 and 2014, while
p is ine si es ha e no been d ained, and d ained si es ha e been
d ained app oxima ely concu en ly wi h he es o ed si es bu
ha e no been es o ed.
Sp uce mi e o es s a e densely eed by Picea abies in oligo-
ophic poo si es, while in meso-eu ophic ich si es, he e a e
also some deciduous ees (esp. Be ula pubescens). The g ound
ege a ion consis s o o bs, g aminoids, and Sphagnum and
ea he mosses. Pine mi e o es s a e spa sely eed by low-
g ow h Pinus syl es is, accompanied by B. pubescens in ich
si es. Pine mi e o es s a e in gene al mo e nu ien -poo han
sp uce mi e o es s, wi h poo si es being omb o ophic and ich
si es oligo-meso ophic. G ound ege a ion consis s ypically o
a ious e e g een and deciduous sh ubs (e.g. Rhododend on
omen osum and Vaccinium uliginosum) and Sphagnum mosses.
Open mi es a e mos ly eeless si es, wi h he ew ees being
P. syl es is in he omb o ophic poo si es and deciduous ees
(e.g. B. pubescens) in oligo-meso ophic ich si es. The g ound
ege a ion in poo si es consis s o Sphagnum mosses and
sh ubs, while in he ich si es, he co e o sedges, o bs,
and we b own mosses inc eases.
Res o a ion aims o aise he wa e able and o e u n he can-
opy s uc u e as simila as possible o he p e-d ained s a e o an
und ained e e ence si e. Typical es o a ion measu es in each
ype consis o filling in and damming he di ches as well as ell-
ing o ees a a ious ex en s, depending on he pea land ype. In
sp uce mi e o es s, a ela i ely dense ee co e has usually
been le a e es o a ion, while in pine mi e o es s, only some
ees ha e been le , and in open mi es, p ac ically all ees ha e
been cu .
In each si e, ege a ion has been moni o ed in 10 one-squa e-
me e squa ed plo s. These plo s a e a anged in wo pa allel
lines, wi h each line con aining fi e plo s spaced ou me e s
apa om each o he (see Fig. 1). The lines a e loca ed o ep e-
sen ypical ege a ion o each si e, and he minimum dis ance o
he nea es di ch is 10 m. The exac loca ion o he fi s plo has
Figu e 1. Finnish pea land moni o ing ne wo k, wi h loca ions o he moni o ing si es (A), numbe o moni o ing si es o each pea land ype and p oduc i i y
(B), and sampling o ege a ion in en o y a each si e (C).
Res o a ion Ecology 3o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
been andomized wi hin he c i e ia defined abo e. The ege a-
ion sampling has been conduc ed be o e es o a ion (yea 0) and
2, 5, and 10 yea s a e es o a ion. In p is ine and d ained si es,
a simila in en o y in e al has been u ilized. Du ing each
in en o y, he %-co e o each ascula plan and moss species
(Table S1) has been isually es ima ed. Fo each si e, we ha e
calcula ed a si e-le el communi y by a e aging he co e o e
plo s o each si e. While he di e ences be ween si es a e la ge
in sp uce mi e o es s han in pine mi e o es s and open mi es
(Elo e al. 2016), wi hin-si e a iabili y is app oxima ely equal
be ween he pea land ypes (Fig. S1).
Fo PFT-le el analyses, we di ided he plan species in o he
ollowing PFTs ha ha e been widely used in pea land esea ch
be o e (e.g. Räsänen e al. 2020a,2020b): deciduous sh ubs,
e e g een sh ubs, o bs, g aminoids, Equise um,P e idophy-
ina,Sphagnum, and o he mosses (Table S1). We u he
di ided he sh ub, o b, and g aminoid PFTs by hei p ima y
habi a equi emen s in o mi e and o he g oups, while Sphag-
num and o he mosses we e di ided in o hummock, lawn, and
hollow species (Eu ola e al. 1995; Finnish Biodi e si y In o
Facili y 2024). This was done because he habi a equi emen s
and po en ial es o a ion impac a e no uni o m wi hin a PFT,
bu species wi hin a single PFT can eac di e en ly o
es o a ion.
Remo e Sensing Da a
We used fi e di e en op ical RS a iables: ed eflec ance,
nea -in a ed (NIR) eflec ance, sho wa e in a ed ans o med
eflec ance (STR; Sadeghi e al. 2015), soil-adjus ed ege a ion
index (SAVI; Hue e 1988), and no malized di e ence mois u e
index (NDMI; Gao 1996). We selec ed a iables ha do no
s ongly co ela e wi h each o he and ha ha e been shown o
be use ul in pea land s udies ela ed o land co e , ege a ion,
and we ness.
O he isible and NIR wa eleng h bands, we chose ed and
NIR due o hei capabili y o ack changes in pea land ege a-
ion, habi a s (Kola i e al. 2022), and we ness (Isoaho
e al. 2023,2024). STR is a ans o ma ion o sho wa e in a ed
(SWIR) eflec ance, and i has been shown o unc ion well in
we ness p edic ion (Isoaho e al. 2024; Jussila e al. 2024). O
di e en ege a ion g eenness indices, we included SAVI due
o i s ela i ely good pe o mance in p edic ing changes in p o-
duc i i y g adien in open and spa sely eed pea lands. We
complemen ed he lis wi h NIR-SWIR index NDMI ha has
co ela ed wi h pea land soil mois u e, wa e able, and we a ea
(Meingas e al. 2014; Ludwig e al. 2019). O e all, a e sa ile
se o a iables has been ecommended due o si e-specific di -
e ences in he mos impo an a iables (Räsänen e al. 2022).
We calcula ed he a iables om he bo om-o -a mosphe e
eflec ance p oduc s o 10–20 m spa ial esolu ion Eu opean
Space Agency Cope nicus Sen inel-2 and 30 m spa ial esolu-
ion Na ional Ae onau ics and Space Adminis a ion/Uni ed
S a es Geological Su ey Landsa 5-9 da ase s ha we ha mo-
nized o Landsa 8-9 eflec ance (Roy e al. 2016; Zhang
e al. 2018). Fo each a iable, we calcula ed ea ly summe
(ES; May 1–June 15) and midsumme (MS; July 1–Augus 15)
annual median image y, om which we calcula ed median
image y o each moni o ing pe iod (1–5 yea s be o e es o a-
ion; 1–3 yea s a e es o a ion, 4–6 yea s a e es o a ion,
and 9–11 yea s a e es o a ion) o bo h seasons. We used
wo seasons as mul i empo al analysis has been shown o boos
model pe o mance in a ious s udies (Räsänen e al. 2020b;
Pang e al. 2022; Wu e al. 2023) and as hese seasons ha e s ik-
ingly di e en hyd ological and phenological condi ions
(Sallinen e al. 2023; Isoaho e al. 2024. We calcula ed median
image y o fil e ou noise p esen in single images and o con-
s uc ep esen a i e da ase s o he selec ed phenological
s ages. Du ing he ES season, he snow has mel , ege a ion
s a s o eme ge, and he wa e able is a i s highes . Du ing
he MS season, ege a ion peaks and he wa e able is ypically
a i s lowes . We did no include image y du ing la e summe o
au umn due o pe sis en cloud co e age du ing ha season. We
u ilized only images wi h a maximum o 30% cloud co e and
masked ou emaining clouds, haze, snow, and shadow wi h
Scene Landco e Classifica ion (Sen inel-2) and Quali y
Assessmen pixel classifica ion (Landsa ).
Fo each a iable, we calcula ed mean alues o a 15-m-
adius bu e a ea ha con ained all ege a ion plo s in he si es.
Fo da es wi h mul iple Sen inel-2 o Landsa sa elli e image
obse a ions, we calcula ed he mean alues o e he obse a-
ions. We conduc ed all sa elli e image p ocessing in Google
Ea h Engine (Go elick e al. 2017).
S a is ical Analysis
We applied a ype o Bayesian join species dis ibu ion model-
ing: Hie a chical Modeling o Species Communi ies (HMSC;
O askainen & Ab ego 2020; O askainen e al. 2017). We con-
duc ed wo di e en se s o HMSC analyses: (1) plan species-
le el and (2) PFT-le el analyses. In bo h analyses, RS a iables
we e included. In Sec ion 3, we mos ly epo species-le el anal-
ysis esul s bu complemen he in o ma ion wi h PFT analysis
esul s.
HMSCs can be used o examine he species- o-species associ-
a ions (he e also RS a iable-species and RS a iable-PFT asso-
cia ions) when con olling o o he co a ia es, as well as
changes in plan communi ies in di e en managemen ypes.
Fo each pea land ype sepa a ely, we modeled he occupancy
(p esence/absence) o he species o PFT ha ing g ea e han
20 occupancies by a p obi model, and condi ionally on he
occu ence, we modeled he co e (log- ans o med, no malized
o ze o mean and uni a iance wi hin each species) o he same
species o PFT wi h a no mal model. We included RS a iables
(no malized o ze o mean and uni a iance) as esponse a i-
ables in he same model o in e hei associa ions wi h species.
As andom e ec s, we included si e, modeled as a spa ially
explici andom e ec and sampling yea . As explana o y a i-
ables, we included ea men (a ac o wi h h ee le els:
es o ed/d ained/p is ine), ime (a con inuous a iable; 0, 2,
5, and 10 since es o a ion o co esponding pe iod), and i s
second-o de polynomial o allow o unimodal esponses, as
well as he in e ac ion o ea men and ime squa ed.
Res o a ion Ecology4o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License

We an he models using R package Hmsc 3.0 (Tikhono
e al. 2020). The package uses he Bayesian amewo k wi h
Gibbs Ma ko chain Mon e Ca lo (MCMC) sampling. We
assumed he de aul p io dis ibu ions, wi h he excep ion o
a1 and a2 pa ame e s o he andom e ec si e, which we e
se bo h o 100 o inc ease he sh inkage and hus a oid model-
ing noise. We sampled he pos e io dis ibu ion wi h ou
chains, each o 250 samples wi h hinning o 10,000, using a
ansien phase o 500,000 and adap a ion ( he numbe o
MCMC s eps a which he adap a ion o he numbe o la en
ac o s is conduc ed) o 400,000. We e alua ed he chain mixing
by assessing he e ec i e size o he pos e io sample wi h a
po en ial scale educ ion ac o (Fig. S2) and assessed he
explana o y powe by Tju ’s
2
(occupancy) and
2
(co e )
(Fig. S3).
Based on he fi ed models, we p edic ed he alues o each
RS a iable in ime o di e en ea men s. F om hese p edic-
ions, we calcula ed he ollowing h ee measu es in o ming
abou di e en aspec s o he e ec s o es o a ion on he RS
a iables.
Fi s , we calcula ed whe he es o a ion a ec ed RS a i-
able as
Resp ¼ R
10  R
0

 D
10  D
0

whe e R
10 and R
0a e alues o RS a iables in es o ed si es in
he yea s 10 and 0, espec i ely, and D
10 and D
0 ep esen co e-
sponding alues in d ained si es. Resp akes posi i e alues i
he change is posi i e in ela ion o change in d ained si es and
nega i e alues i he change is nega i e in ela ion o change
in d ained si es.
Second, alues o RS in d ained and in es o ed si es may di -
e as hey we e no andomly selec ed. To assess he eliabili y
o in e ences o how RS a iables espond o es o a ion, we cal-
cula ed whe he hey di e ed a he beginning o he expe imen
be ween he d ained and es o ed si es:
Di 1 ¼ R
0 D
0
Thi d, we calcula ed whe he he di e ence be ween es o ed
and p is ine con ol si es g ew smalle (o la ge ) du ing he
s udy pe iod:
Di 2 ¼abs R
10  P
10

abs R
0 P
0

Fo all h ee measu es, we calcula ed he median as well as he
pos e io p obabili y o he median being la ge han ze o. We
conside ed he measu e o ha e high suppo o he median
being posi i e/nega i e i he pos e io p obabili y is g ea e
han 95% and mode a e suppo i he pos e io p obabili y is
g ea e han 80%. We calcula ed he same measu es o abun-
dance o each species, o PFT (p obabili y o
occu ence co e gi en occu ence). As he species-specific
esponses o es o a ion me ely ein o ce he p e ious findings
(Elo e al. 2024), we p esen hem only in Fig. S4 oge he wi h
he PFT-le el in o ma ion.
To answe whe he RS a iables imp o e p edic i e powe o
he species- o PFT-specific models, we fi s calcula ed wo old
c oss- alida ion. Then, we pe o med condi ional c oss-
alida ion, whe e we used da a om each RS a iable, one a a
ime, and i ’s es ima ed associa ions wi h he species o PFTs
o calcula e he p edic ions. Finally, we compa ed whe he
including in o ma ion on he RS a iable yielded an imp o e-
men in p edic i e powe by sub ac ing he c oss- alida ed p e-
dic i e powe om he condi ionally c oss- alida ed p edic i e
powe (CCV). We did he c oss- alida ions wi h pa ame e
alues based on hin =10 due o he high compu a ional
demand o he calcula ions and because p edic i e powe s end
o con e ge wi h a ela i ely low numbe o hinning. Fu he -
mo e, a iance pa i ioning o he explana o y a iables
emained simila when hinning o 10 o 10,000 was used.
Finally, we calcula ed associa ion ma ices, which ep esen
he esidual associa ions o RS a iables and species, o PFTs,
a e con olling o he ea men , ime, and hei in e ac ion.
Resul s
E ec o Res o a ion on Spec al Signa u es
Almos all RS a iables we e a ec ed by es o a ion (Fig. 2).
Especially, bo h ES and MS ed and NIR eflec ance inc eased
a e es o a ion in mos pea land ypes. Mo eo e , SAVI MS
inc eased, whe eas o SAVI ES, he esponse had low s a is ical
suppo (pos e io p obabili y <95%). STR and NDMI
dec eased o showed no highly suppo ed esponse o es o a-
ion, wi h STR showing highly suppo ed esponse in mo e pea -
land ypes han NDMI. The only pea land ype whe e no high
suppo was seen in any o he RS a iables was ich sp uce mi e
o es s, whe eas he clea es e ec s we e seen in pine mi e o -
es s and open mi es. In pine mi e o es s and open mi es, he
ed and NIR eflec ance o es o ed si es had simila alues han
d ained si es be o e es o a ion and app oached hose o p is ine
si es a e es o a ion (Figs. 3&4). Fo o he a iables and pea -
land ypes, he empo al ends in es o ed, p is ine, and d ained
si es we e less clea , and he spec al signa u es in es o ed si es
did no clea ly mo e close o he signa u es in p is ine
si es (Figs. 3–5).
Imp o emen in P edic i e Powe F om RS Va iables
Fo mos species o PFTs, a leas one o he RS
a iables imp o ed he p edic i e powe and esul ed in a p e-
dic i e powe highe han 0 (Figs. 6&S5). The e we e di e -
ences be ween species and pea land ypes, which RS a iables
imp o ed he p edic i e powe , and none o he a iables was
clea ly be e han he o he s (Fig. 7). The esul ing p edic i e
powe s we e gene ally ela i ely modes bo h o species and
PFTs. The mean was ypically ci ca 0.1–0.2, bu o some spe-
cies, CCV was e y high (up o 0.65; Table S2). The same
applied o he imp o emen in p edic i e powe when including
he bes RS a iable: ypically, he imp o emen was small
(<0.1), bu o some species, i was e y high (up o 0.52;
Table S2).
Res o a ion Ecology 5o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
Associa ions Be ween Species o PFTs and Remo e Sensing
Va iables
When concen a ing on hose RS-species o RS-PFT linkages in
which (1) es o a ion had an e ec on bo h he RS a iable and
species o PFT and (2) inco po a ing RS a iable inc eased p e-
dic i e powe (Figs. 8&S6), he imp o emen in p edic i e
powe was ypically small (<0.05 in imp o emen in c oss-
alida ed p edic i e powe [CCV-CV]). The excep ions we e
mainly he nega i e associa ions o g aminoids wi h NDMI
and ed eflec ance (Ca ex cho do hiza in poo pine mi e o -
es s and ich open mi es, and also C. lasioca pa co e in ich
open mi es; Fig. 8) and g aminoid PFTs in ich open mi es
(Fig. S6). Addi ionally, a somewha clea imp o emen (>0.05
in CCV-CV) was seen o Vaccinium uliginosum co e associ-
a ed nega i ely wi h NDMI ES in ich pine mi e o es s. The e
we e also o he associa ions, bo h nega i e and posi i e, bu in
hese cases, RS a iables inc eased he p edic i e powe li le
(<0.05 in CCV-CV).
Discussion
Ou esul s show ha (1) pea land es o a ion a ec s sa elli e-
de i ed spec al signa u es, (2) sa elli e image a iables
inc ease modes ly species- o PFT-specific p edic i e powe
in join species dis ibu ion models, and (3) he e a e ew
clea links be ween he changes in RS a iables and he
changes in pos - es o a ion species o PFT occu ence and
co e .
Res o a ion E ec s on Spec al Signa u es
O e ime, he spec al signa u es o es o ed si es mo ed close
o hose o p is ine si es. As p is ine-like ecosys em s uc u e
and unc ioning is he goal o es o a ion, he esul sugges s
ha es o a ion can be success ully moni o ed wi h RS da a.
The end owa d p is ine was pa icula ly e iden in ed and
NIR eflec ance o spa sely eed pine mi e o es s and open
mi es, whe eas o o he es ed a iables and especially o
densely eed sp uce mi e o es s, he ends we e no as clea .
These findings align wi h Ball e al. (2023), who obse ed he
con e gence be ween es o ed and p is ine si es wi h op ical
Sen inel-2 and syn he ic ape u e ada Sen inel-1 da a. In hei
analysis, he simila i y inc eased ela i ely s ongly du ing he
fi s 10–15 yea s a e which he signa u es be ween es o ed
and p is ine si es we e close o each o he . We could no e i y
his finding due o ou 10-yea pos - es o a ion moni o ing
pe iod bu ins ead showed ha du ing he fi s 10 yea s, he ha -
moniza ion in a iable alues be ween es o ed and p is ine si es
was e iden only o ce ain a iables and pea land ypes. Ball
e al. (2023) did no analyze he ends in di e en bands and
indices bu ocused on o e all spec al simila i y using Mahala-
nobis dis ance and limi ed analysis o 1-yea sampling o pea -
lands es o ed du ing di e en yea s. The e o e, ou analysis
complemen s he wo k by Ball e al. (2023) by showing (1) ha
he e a e di e ences be ween pea land ypes and RS a iables
and (2) wha kind o end is seen a e es o a ion.
Ou esul s indica e ha eflec ance o he ed, NIR, and
SWIR (STR is ans o med SWIR eflec ance and nega i ely
co ela ed wi h i ) inc eases a e es o a ion, pa icula ly in pine
Figu e 2. The esponse o es o a ion o emo e sensing a iables in di e en pea land ypes. No e ha he alues a e based on he o iginal alues o each emo e
sensing a iable; he e o e, he ange o sho wa e in a ed ans o med eflec ance (STR) is much la ge han o non- ans o med bands (STR is ci ca 50 and
1 o sho wa e in a ed eflec ance o 1 and 26%, espec i ely). In he figu e, NDMI e e s o no malized di e ence mois u e index, SAVI o soil-adjus ed
ege a ion index, ES o ea ly summe , and MS o midsumme . The s a is ical suppo is “Posi i e 95%”i he pos e io p obabili y o he median being la ge han
ze o is g ea e han 95%; “Nega i e 95%”i he pos e io p obabili y o he median being smalle han ze o is g ea e han 95%; and “Weak”i he pos e io
p obabili y o he median being la ge /smalle han ze o is less han 95%.
Res o a ion Ecology6o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
mi e o es s and open mi es. This is p obably la gely a ibu ed
o he elling o ees and inc eased openness in he landscape
in hese pea land ypes, as especially ed eflec ance and RS-
measu ed albedo is nega i ely associa ed wi h woody canopy
co e (Yang & P ince 1997; Kuusinen e al. 2016). Felling o
ees is conduc ed du ing es o a ion o make oom o
Figu e 3. Changes in emo e sensing a iables o e ime in d ained, p is ine, and es o ed open mi es. The plo s a e d awn only o hose changes in emo e
sensing a iables ha esponded ei he posi i ely o nega i ely o es o a ion wi h a high suppo (a pos e io p obabili y o he median being la ge /smalle han
ze o g ea e han 95%; Fig. 2). In he figu e, NDMI e e s o no malized di e ence mois u e index, SAVI o soil-adjus ed ege a ion index, STR o sho wa e
in a ed ans o med eflec ance, ES o ea ly summe , and MS o midsumme .
Res o a ion Ecology 7o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
exca a o s along he di ches, b ing back p is ine-like canopy
s uc u e, and dec ease e apo anspi a ion by ees. Al hough
he e ec s o es o a ion occu sho - e m and a e e iden in
2 yea s a e es o a ion da a, ou modeling app oach, whe e
ime is used as a con inuous a iable, ends o ex end his e ec .
Howe e , he elling o ees du ing he ime o es o a ion does
Figu e 4. Changes in emo e sensing a iables o e ime in d ained, p is ine, and es o ed pine mi e o es s. The plo s a e d awn only o hose changes in emo e
sensing a iables ha esponded ei he posi i ely o nega i ely o es o a ion wi h a high suppo (a pos e io p obabili y o he median being la ge /smalle han
ze o g ea e han 95%; Fig. 2). In he figu e, NDMI e e s o no malized di e ence mois u e index, SAVI o soil-adjus ed ege a ion index, STR o sho wa e
in a ed ans o med eflec ance, ES o ea ly summe , and MS o midsumme .
Res o a ion Ecology8o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License
Tikhono G, Opedal ØH, Ab ego N, Lehikoinen A, de Jonge MM, Oksanen J,
O askainen O (2020) Join species dis ibu ion modelling wi h he
R-package Hmsc. Me hods in Ecology and E olu ion 11:442–447.
h ps://doi.o g/10.1111/2041-210X.13345
U 
ak I, Ha el T, Gallé R, Balog A (2017) Wo ldwide pea land deg ada ions and
he ela ed ca bon dioxide emissions: he impo ance o policy egula ions.
En i onmen al Science & Policy 69:57–64. h ps://doi.o g/10.1016/j.
en sci.2016.12.012
Vasande H, Tui ila E-S, Lode E, Lundin L, Ilome s M, Sallan aus T, Heikkilä R,
Pi känen M-L, Laine J (2003) S a us and es o a ion o pea lands in no h-
e n Eu ope. We lands Ecology and Managemen 11:51–63. h ps://doi.o g/
10.1023/A:1022061622602
Wu S, Te zla D, Daempfling H, Soulsby C (2023) Imp o ed unde s anding o
ege a ion dynamics and we land ecohyd ology ia mon hly UAV-based
classifica ion. Hyd ological P ocesses 37:e14988. h ps://doi.o g/10.
1002/hyp.14988
Wulde MA, Roy DP, Radelo VC, Lo eland TR, Ande son MC,
Johnson DM, e al. (2022) Fi y yea s o Landsa science and impac s.
Remo e Sensing o En i onmen 280:113195. h ps://doi.o g/10.1016/j.
se.2022.113195
Yang J, P ince SD (1997) A heo e ical assessmen o he ela ion be ween woody
canopy co e and ed eflec ance. Remo e Sensing o En i onmen 59:428–
439. h ps://doi.o g/10.1016/S0034-4257(96)00111-3
Zhang HK, Roy DP, Yan L, Li Z, Huang H, Ve mo e E, Skakun S, Roge JC
(2018) Cha ac e iza ion o sen inel-2A and Landsa -8 op o a mosphe e,
su ace, and nadi BRDF adjus ed eflec ance and NDVI di e ences.
Remo e Sensing o En i onmen 215:482–494. h ps://doi.o g/10.1016/j.
se.2018.04.031
Suppo ing In o ma ion
The ollowing in o ma ion may be ound in he online e sion o his a icle:
Figu e S1. Box and whiske s plo s o ele a ion (m a.s.l.), mean empe a u e (C),
annual p ecipi a ion (mm), numbe o plan species, and plan species be a di e si y
g adien s o di e en ea men s and pea land ypes.
Figu e S2. Con e gence o be a and omega pa ame e s o ich open mi es, wi h hin-
ning o 10,000.
Figu e S3. Va iance pa i ioning o explana o y a iables o ich open mi es, wi h
hinning o 10,000.
Figu e S4. Species and plan unc ional ype median esponse o es o a ion, sepa-
a ely o each pea land ype.
Figu e S5. Analysis whe he he emo e sensing a iables inc eased he condi ionally
c oss- alida ed p edic i e powe o di e en plan uc ional ypes.
Figu e S6. The associa ions be ween he plan unc ional ypes and he emo e sensing
a iables.
Table S1. Species abb e ia ions, ull scien ific names and plan unc ional ypes o
each species.
Table S2. Condi ionally c oss- alida ed p edic i e powe and he imp o emen in
c oss- alida ed p edic i e powe when including he bes pe o ming emo e sensing
a iable.
Coo dina ing Edi o : Raja Hussain Recei ed: 23 Augus , 2024; Fi s decision: 3 Oc obe , 2024; Re ised: 25
Oc obe , 2024; Accep ed: 28 Oc obe , 2024
Res o a ion Ecology 15 o 15
Remo e sensing o pea land ege a ion
1526100x, 0, Downloaded om h ps://onlinelib a y.wiley.com/doi/10.1111/ ec.14338 by Uni e si y O Jy äskylä Lib a y, Wiley Online Lib a y on [10/11/2024]. See he Te ms and Condi ions (h ps://onlinelib a y.wiley.com/ e ms-and-condi ions) on Wiley Online Lib a y o ules o use; OA a icles a e go e ned by he applicable C ea i e Commons License