Con ibu ed Pape
Combining geodi e si y wi h clima e and opog aphy
o accoun o h ea ened species ichness
Helena Tukiainen,∗¶§ Joseph J. Bailey,† § Richa d Field,† Ka ja Kangas,‡ and Jan Hjo ∗
∗Geog aphy Resea ch Uni , Uni e si y o Oulu, P.O. Box 8000, Oulu, FI 90014, Finland
†School o Geog aphy, Uni e si y o No ingham, Uni e si y Pa k, No ingham NG7 2RD, U.K.
‡Na u al Resou ces Ins i u e Finland (Luke), Economics and Socie y, Uni e si y o Oulu, P.O. Box 413, Oulu, FI 90014, Finland
Abs ac : Unde s anding h ea ened species di e si y is impo an o long- e m conse a ion planning.
Geodi e si y— he di e si y o Ea h su ace ma e ials, o ms, and p ocesses—may be a use ul biodi e si y
su oga e o conse a ion and ha e conse a ion alue i sel . Geodi e si y and species ichness ela ionships
ha e been demons a ed; es ablishing whe he geodi e si y ela es o h ea ened species’ di e si y and dis i-
bu ion pa e n is a logical nex s ep o conse a ion. We used 4 geodi e si y a iables ( ock- ype and soil- ype
ichness, geomo phological di e si y, and hyd ological ea u e di e si y) and 4 clima ic and opog aphic
a iables o model h ea ened species di e si y ac oss 31 o Finland’s na ional pa ks. We also analyzed a i y-
weigh ed ichness (a measu e o si e complemen a i y) o h ea ened ascula plan s, ungi, b yophy es, and
all species combined. Ou 1-km2 esolu ion da a se included 271 h ea ened species om 16 majo axa.
We modeled h ea ened species ichness ( aw and a i y weigh ed) wi h boos ed eg ession ees. Clima ic
a iables, especially he annual empe a u e sum abo e 5 °C, domina ed ou models, which is consis en
wi h he c i ical ole o empe a u e in his bo eal en i onmen . Geodi e si y added signi ican explana o y
powe . High geodi e si y alues we e consis en ly associa ed wi h high h ea ened species ichness ac oss
axa. The combined e ec o geodi e si y a iables was e en mo e p onounced in he a i y-weigh ed ichness
analyses (excep o ungi) han in hose o species ichness. Geodi e si y measu es co ela ed mos s ongly
wi h species ichness ( aw and a i y weigh ed) o h ea ened ascula plan s and b yophy es and we e
weakes o molluscs, lichens, and mammals. Al hough simple measu es o opog aphy imp o e biodi e si y
modeling, ou esul s sugges ha geodi e si y da a ela ing o geology, land o ms, and hyd ology a e also
wo h including. This ein o ces ecen a gumen s ha conse ing na u e’s s age is an impo an p inciple in
conse a ion.
Keywo ds: biodi e si y, conse ing na u e’s s age, geology, geomo phology, he e ogenei y, hyd ology
Combinaci´
on de la Geodi e sidad con el Clima y la Topog a ´
ıa pa a Rep esen a la Riqueza de Especies Amenazadas
Resumen: En ende la di e sidad de especies amenazadas es impo an e pa a la planeaci´
on de la con-
se aci´
on a la go plazo. La geodi e sidad – la di e sidad de ma e iales, o mas y p ocesos en la supe icie
e es e – puede se un sus i u o ´
u il de la biodi e sidad pa a la conse aci´
on y puede ene un alo de
conse aci´
on p opio. Las elaciones en e la geodi e sidad y la iqueza de especies han sido demos adas; el
siguien e paso l´
ogico pa a la conse aci´
on es es ablece si la geodi e sidad se elaciona con la di e sidad de
especies amenazadas y los pa ones de dis ibuci´
on. Usamos cua o a iables de la geodi e sidad ( iqueza
de ipo de oca y de ipo de suelo, di e sidad geomo ol´
ogica, ca ac e ´
ıs icas de la di e sidad hid ol´
ogica) y
cua o a iables clim´
a icas y opog ´
a icas pa a modela la di e sidad de especies amenazadas en 31 de los
pa ques nacionales de Finlandia. Tambi´
en analizamos la iqueza ponde ada con la a eza (una medida
de la complemen a iedad de si io) de las plan as ascula es, hongos y b io i as amenazadas y odas las
especies combinadas. Nues o conjun o de da os de esoluci´
on de 1-km2inclu´
ıa 217 especies amenazadas de
¶email [email p o ec ed].
§These au ho s con ibu ed equally o his a icle.
Pape submi ed Ma ch 1, 2016; e ised manusc ip accep ed July 25, 2016.
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.
364
Conse a ion Biology, Volume 31, No. 2, 364–375
C
2016 The Au ho s. Conse a ion Biology published by Wiley Pe iodicals, Inc. on behal o Socie y o Conse a ion Biology
DOI: 10.1111/cobi.12799
Tukiainen e al. 365
16 axones mayo es. Modelamos la iqueza de especies amenazadas (c uda y ponde ada con la a eza) con
´
a boles de eg esi´
on es imulados. Las a iables clim´
a icas, especialmen e la suma de la empe a u a anual
sob e los 5 °C, domina on nues os modelos, lo que es consis en e con el papel c ´
ı ico de la empe a u a en es e
ambien e bo eal. La geodi e sidad a˜
nadi´
o un pode explica i o. Los al os alo es de geodi e sidad es u ie on
asociados cons an emen e con la al a iqueza de especies amenazadas en los axones. El e ec o combinado
de las a iables de la geodi e sidad es u o m´
as p onunciado en los an´
alisis de iqueza sopesados con la
a eza (excep o po los hongos) que en aquellos pa a la iqueza de especies. Las medidas de geodi e sidad
se co elaciona on m´
as ue emen e con la iqueza de especies (sopesada con la a eza y la c udeza) de las
plan as ascula es y las b io i as y ue on m´
as d´
ebiles pa a los moluscos, los l´
ıquenes y los mam´
ı e os. Aunque
las medidas simples de la opog a ´
ıa mejo an el modelado de la biodi e sidad, nues os esul ados sugie en
que los da os de geodi e sidad elacionados con la geolog´
ıa, las o maciones e es es y la hid olog´
ıa ambi´
en
deben se incluidos. Es o e ue za los a gumen os ecien es que dicen que conse a el es ado de la na u aleza
es un p incipio impo an e en la conse aci´
on.
Palab as Cla e: biodi e sidad, conse aci´
on del es ado de la na u aleza, geolog´
ıa, geomo olog´
ıa, he e ogenei-
dad, hid olog´
ıa
In oduc ion
Land-use and clima e change h ea en species globally
(Man yka-P ingle e al. 2015). I is he e o e inc easingly
impo an o unde s and and conse e species’ di e si y
and dis ibu ions. One coa se- il e s a egy in conse a-
ion and p o ec ed-a ea managemen , conse ing na u e’s
s age, cen e s on he physical s uc u es ha unde lie
bio ic p ocesses and ecognizes ha geodi e si y– he di-
e si y o Ea h su ace o ms, ma e ials, and p ocesses
(G ay 2013)–i sel has conse a ion alue and is ela ed
o biodi e si y (Ande son & Fe ee 2010; Lawle e al.
2015). Thus, p oponen s sugges ha geodi e si y be in-
co po a ed in o biodi e si y esea ch and conse a ion
(Lawle e al. 2015). In p ac ical e ms, geodi e si y da a
a e equen ly easie and less expensi e o ob ain han
biodi e si y da a (Hjo e al. 2012) and ad ance scien i ic
unde s anding o he spa ial dis ibu ion o biodi e si y
o long- e m conse a ion planning.
Well-mapped abio ic da a a e commonly used, oge he
wi h da a om ecological communi ies, o model o p e-
dic biodi e si y o conse a ion planning (Albuque que
& Beie 2015a). Nume ous esea che s ha e explo ed
how abio ic ac o s a e ela ed o species’ di e si y and
dis ibu ions (Lawle e al. 2015), and he ela ionship
be ween en i onmen al he e ogenei y and species ich-
ness has been es ablished ac oss mul iple axa and spa ial
scales (S ein e al. 2014). Se e al esea che s ha e shown
ha local, idiosync a ic ea u es, such as ock ype, and
landscape-scale ac o s, such as ene gy- ela ed clima o-
logical a iables, co ela e wi h h ea ened species’ pa -
e ns a di e en spa ial scales and di e among axo-
nomic g oups (Be g e al. 2002; K e & Je z 2007; Lawle
e al. 2015). Vascula plan s a e commonly s udied and
ypically ha e s ong ela ionships wi h clima ic a i-
ables and opog aphic he e ogenei y (Field e al. 2009),
whe eas he e ec o abio ic condi ions is less known
o a e species in o he axonomic g oups (e.g., Vi kkala
e al. 2005; Ande son & Fe ee 2010).
Geodi e si y ep esen s a mo e comple e cha ac e i-
za ion o Ea h-su ace he e ogenei y han opog aphic
he e ogenei y does. We de ine geodi e si y as dis inc
om widely used opog aphic measu es such as ele a-
ion, ange in ele a ion, and slope (he ein e e ed o
as opog aphy). Geodi e si y includes explici geo ea-
u es such as ock ypes, soil ypes, geomo phological
land o ms, and hyd ological ea u es. These can be ex-
plici ly inco po a ed in o analyses o cap u e he aspec s
o local he e ogenei y, such as mic oclima ic e ec s and
mic osi e pa e ns, which a e ecologically impo an bu
a e no cap u ed by clima ic and opog aphic da a (Field
e al. 2009; Dob owski 2011). Such local he e ogenei y,
caused by geology, land o ms, and hyd ology, ela es o
ex ended local- esou ce g adien s, niche space, and habi-
a a ie y (S ein e al. 2014).
Conse a ion p o essionals ha e a ely inco po a ed
geodi e si y in o conse a ion p io i iza ion e o s (Beie
e al. 2015). Howe e , ela ionships be ween geodi e -
si y and biodi e si y a e being demons a ed inc easingly
(Lawle e al. 2015) and may ha e conside able implica-
ions o conse a ion. A signi ican link be ween plan
species ichness (domina ed by common species) and
geodi e si y has been iden i ied (Nichols e al. 1998; Hjo
e al. 2012). Es ablishing whe he such a ela ionship ex-
is s be ween geodi e si y and h ea ened species’ di e -
si y and dis ibu ion pa e ns is a logical nex s ep. Indeed,
quan i ying geodi e si y may p o ide g ea e insigh in o
a landscape’s po en ial o p ese e species di e si y (An-
de son & Fe ee 2010; Lawle e al. 2015).
Howe e , numbe s o species, and e en numbe s o
h ea ened species, may no p o ide op imal measu es
by which o p io i ize si es o conse a ion (Ki kpa ick
1983; Albuque que & Beie 2015a). Fo op imal plan-
ning, manage s o en wan o iden i y g oups o si es ha
collec i ely ep esen mul iple conse a ion a ge s ( yp-
ically h ea ened species) in small a eas (Albuque que &
Beie 2015a). Thus, ins ead o selec ing he si es wi h he
g ea es species ichness o he mos h ea ened species,
Conse a ion Biology
Volume 31, No. 2, 2017
366 Geodi e si y and Th ea ened Species
a se o si es wi h species assemblages ha complemen
each o he and collec i ely cap u e he la ges numbe
o species is chosen (Albuque que & Beie 2015b). Va i-
ous si e-p io i iza ion me hods ha e been de eloped; so -
wa e such as Zona ion (Moilanen e al. 2014) and Ma xan
(A d on e al. 2010) a e qui e commonly used. Howe e ,
a simple al e na i e, a i y-weigh ed ichness (RWR), is
e icien and eliable and allows he iden i ica ion o p i-
o i y si es (Albuque que & Beie 2015b). Albuque que
and Beie (2015a, 2015c) ecen ly demons a ed ha si e
complemen a i y can be highly p edic able om abio ic
cha ac e is ics.
We modeled he ela ionship be ween he physical
en i onmen and bo h h ea ened species ichness and
RWR a 1-km²g ain size ac oss 31 p o ec ed a eas o Fin-
land. We used 2 p edic o ca ego ies: clima e and opog-
aphy (i.e., con en ional p edic o s) and geodi e si y.
Ou main aim was o de e mine he explana o y powe
o he geodi e si y measu es, which in his s udy a e
geo ichness a iables. We expec ed da a on geomo phol-
ogy, geology, and hyd ology o be ele an o landscape-
scale pa e ns o h ea ened species ichness and RWR be-
cause hey should ep esen local geophysical condi ions
ha a e impo an o he es ablishmen and pe sis ence
o h ea ened species (Rich & Weiss 1991; Engle e al.
2004). In heo y, mo e h ea ened species should be able
o pe sis whe e he e is g ea e geological a ie y be-
cause o he b oade a ie y o nu ien s, esou ces, and
pH (as p e iously obse ed o common species [Hjo
e al. 2012]). Ou s udy a ea included only p o ec ed a -
eas, so he le els o human impac we e consis en ly low.
We analyzed da a o h ea ened species o ascula
plan s, ungi, lichens, bee les (Coleop e a), b yophy es,
bu e lies and mo hs (Lepidop e a), molluscs, mammals,
and all o hese combined. We used h ea ened species
ichness and RWR as measu es o h ea ened species di-
e si y and a i y, espec i ely. We es ed he explana o y
powe o en i onmen al a iables in modeling di e si y
and RWR; assessed he consis ency o hese ela ion-
ships ac oss di e en axonomic g oups; and s udied
which geo ichness measu es ( ock- ype ichness, soil-
ype ichness, geomo phological di e si y, and hyd olog-
ical ea u e di e si y) added explana o y powe o ou
models. We add essed he ollowing hypo heses, which
a e no mu ually exclusi e. Th ea ened species di e si y
is s ongly ela ed o clima e, and in high-la i ude en i-
onmen s especially o ene gy- ela ed clima e a iables
(H1) (Hawkins e al. 2003; S ein e al. 2014). Di e en
axonomic g oups show di e en esponses o clima ic,
opog aphic, and geodi e si y p edic o s (H2)(S eine al.
2014). Geodi e si y measu es imp o e models o h ea -
ened species di e si y and RWR (H3) (Bu ne e al. 1998;
Ande son & Fe ee 2010). Th ea ened species di e si y
and RWR can be success ully modeled wi h clima e, o-
pog aphy, and geodi e si y a iables (H4) (Pausas e al.
2003; Hjo e al. 2012; Albuque que & Beie 2015a).
Me hods
S udy A ea
Ou s udy a ea co e ed 31 na ional pa ks (Fig. 1), ex end-
ing om sou he n Finland’s coas al a chipelago o no h-
e n Finland’s glacially ounded hills wi h a c ic-alpine
condi ions. Pa k a ea anged om 6 km²(Pe kelj¨
a i) o
2850 km2(Lemmenjoki); he o al a ea was 8091 km².
Finnish na ional pa ks ollow he de ini ions and man-
agemen objec i es o he In e na ional Union o he
Conse a ion o Na u e (IUCN) and na u al esou ces
p o ec ed a ea managemen ca ego y II (Heinonen 2013).
We supe imposed a egula sys em o 1-km²g id cells and
e ained all cells con aining h ea ened species eco ds
ha had a leas 10% o hei a ea in a na ional pa k. Fo
cells wi hou h ea ened species eco ds, we e ained all
cells loca ed en i ely wi hin na ional pa k bounda ies. We
he e o e selec ed 6571 g id cells, 583 wi h obse a ions
o h ea ened species and 5988 wi hou .
Biogeog aphically, he s udy a ea co e ed hemi-,
sou he n-, middle-, and no he n-bo eal ege a ion zones
(Ah i e al. 1968) and included a g ea a ie y o land-
co e ypes, such as o es s, ell a eas, and we lands.
Mean annual ai empe a u e a ied om −2°Cin he
no h o app oxima ely 6 °C in he sou h (Pi inen e al.
2012), and he leng h o he he mal g owing season
(>5°C daily mean empe a u es) was om >185 days in
he sou h o <125 days in he no h. Mean annual p ecipi-
a ion was mode a e h ough all seasons and anged om
444 o 739 mm (Table 1) (Pi inen e al. 2012). Finland is
pa o he P ecamb ian bed ock block o no he n and
eas e n Eu ope and consis s mainly o c ys alline ocks
(A las o Finland 1990). The soils o Finland o igina e
mainly om du ing o a e he las glacial pe iod and a e
domina ed by g ound mo aine and pea deposi ions.
Th ea ened Species Da a
We conside ed h ea ened species om he ollow-
ing axonomic g oups: ascula plan s, ungi, lichens,
b yophy es, bee les, bu e lies and mo hs, molluscs,
mammals, 2-winged lies, ue bugs, bi ds, hymenop e -
ans, caddis lies, s one lies, amphibians, and spide s. The
ichness o each o he i s 8 o hese (up o and in-
cluding mammals) was modeled, as was ha o all 16
axa combined (‘all’ ca ego y). Th ea ened species we e
hose conside ed c i ically endange ed, endange ed, ul-
ne able, o nea - h ea ened in Finland acco ding o he
IUCN Red Lis (Rassi e al. 2001). We included a ew
da a-de icien species known o be a e ( o de ails, see
Suppo ing In o ma ion).
Geog aphic coo dina es o he eco ds o h ea ened
species we e de i ed om he He a da abase (Finnish
En i onmen Ins i u e 2015). To ensu e hei spa ial accu-
acy, we used occu ences eco ded a e he yea 2000.
Conse a ion Biology
Volume 31, No. 2, 2017
Tukiainen e al. 367
Figu e 1. Loca ions and majo ege a ion zones o he Finnish na ional pa ks included in ou s udy o hei
h ea ened species di e si y and a i y-weigh ed ichness. The pa ks a e spli in o eas e n (unde lined) and
wes e n pa ks.
Nea ly all (99.3%) he coo dina es we e eco ded wi h
GPS wi h 100-m accu acy. The e was no bias in ela-
ion o he numbe o species occu ences and p oximi y
o ec ea ional ou es ( his was examined o ascula
plan s, b yophy es, lichens, and ungi by Siikam¨
aki e al.
[2015]).
As well as analyzing aw h ea ened species ichness,
we modeled RWR (Williams e al. 1996; Albuque que
& Beie 2015b). The a i y alue o each species is he
in e se o he numbe o g id cells in which i occu s.
The RWR alue pe g id cell is he sum o he a i y alues
om each species eco ded. G id cells con aining a e
species he e o e ha e highe RWR. We calcula ed RWR
o each axonomic g oup wi h su icien da a ( ascula
plan s, b yophy es, and ungi) and all species combined.
Fo he RWR modeling, we used he same en i onmen al
a iables as o h ea ened species ichness analyses. We
also used hese a iables o addi ional analyses o he
dis ibu ion o each axon (de ails gi en in Suppo ing
In o ma ion).
En i onmen al Va iables
We compiled 24 en i onmen al (abio ic) a iables o he
6571 1-km2cells as po en ial p edic o s (Table 1). As well
as geodi e si y measu es (numbe o ypes o ock, soil,
land o m, and hyd ological ea u es), hese abio ic en i-
onmen al a iables included widely used clima ic and
opog aphic a iables so we would co e he mos likely
abio ic co ela es o species di e si y a he landscape
scale (Field e al. 2009; S ein e al. 2014).
We de i ed opog aphic a iables om a 25-m-
esolu ion digi al ele a ion model (DEM) (NLS 2000).
Ele a ion and slope angle (mean, SD, and ange) we e
calcula ed pe g id cell wi h A cMap e sion 10.2 (ESRI,
USA). Topog aphy-de i ed mois u e condi ions o he
s udy a ea we e calcula ed using he opog aphic we ness
index (TWI) (Be en & Ki kby 1979).
De ailed, expe -de i ed da a on all he land o ms in
each g id cell we e a ailable only o 2083 cells, so ge-
omo phological ichness o all cells was de e mined by
Conse a ion Biology
Volume 31, No. 2, 2017
368 Geodi e si y and Th ea ened Species
Table 1. De ails o conside ed en i onmen al a iables o use in boos ed eg ession- ee analyses o spa ial ichness and a i y-weigh ed ichness
pa e ns o se e al axa ac oss Finland’s na ional pa ks (n=6571).
En i onmen al a iable Uni Median (min o max) Sou ceaAbb e ia ion
Geodi e si y
ock- ype ichnessbnumbe o ock ypes 1 (1–6) GSF ock ich
soil- ype ichnessbnumbe o soil ypes 2 (1–5) GSF soil ich
geomo phological ichnessbnumbe o
geomo phological ea u e
ypes
6 (0–13) GAM GM ich
hyd ological ea u e
ichnessb
numbe o hyd ological
ea u e ypes
1 (0–6) NLS hyd o ich
Clima ec
mean annual ai empe a u e °C –1.1 (–2 o 5.9) FMId
g owing deg ee days
(>5°C)b
deg ee-days 639 (500.3–1448.3) FMI GDD
mean empe a u e o coldes
mon h
(Janua y) °C –13.4 (–14.5 o –3) FMI
mean empe a u e o
wa mes mon h (July)
°C 13.1 (12.5–17.7) FMI
seasonali y (mean
empe a u e o July–
Janua y) °C 26.4 (20.1–29.3) FMI
mean annual p ecipi a ionbmm 540.7 (443.6–739.3) FMI MP
po en ial
e apo anspi a iond
mm yea −1210.1 (194.6–392.6) FMId
wa e balancedmm yea −1327.8 (221–435.9) FMId
heo e ical sola adia ion
(mean)e
Mj cm−2yea −10.5 (0.3–0.6) DEMe
heo e ical sola adia ion
(SD)e
Mj cm−2yea −10.02 (<0.01–0.2) DEMe
heo e ical sola adia ion
( ange)e
Mj cm−2yea −10.2 (<0.01–0.7) DEMe
Topog aphy
ele a ion (mean) m asl 308.3 (10.2–738.9) DEM
ele a ion (SD) m 11.1 (0–127.5) DEM
ele a ion ( ange)bm 48 (0–414) DEM ER
slope angle (mean) deg ees 2.9 (<0.01–22.7) DEM
slope angle (SD) deg ees 2 (0.02–12.9) DEM
slope angle ( ange) deg ees 12 (0.6–56.7) DEM
opog aphical we ness index
(mean)
– 10.8 (6.4–24.3) DEM
opog aphical we ness index
(SD)
– 4.6 (0.4–7.7) DEM
opog aphical we ness index
( ange)b
– 23.2 (9.5–32.8) DEM TWIR
aAbb e ia ions: DEM, digi al ele a ion model; GAM, gene alized addi i e model; GSF, Geological Su ey o Finland; FMI, Finnish Me eo ological
Ins i u e; NLS, Na ional Land Su ey o Finland.
bVa iables selec ed o species ichness and a i y-weigh ed ichness modeling, based on co ela ion analysis (see Me hods o u he de ails).
cClima e a iables a e o 1981–2010.
dMe hod ollowing Sko and S enning (2004).
eEs ima e o po en ial annual di ec inciden adia ion calcula ed using A cGIS 10.2 (McCune & Keon 2002).
gene alized addi i e modeling (GAM) wi h he a ailable
land o m da a and he 25-m DEM, as ollows (Hjo &
Luo o 2012). Land o m da a we e de i ed om 1:50,000
geomo phological maps and ae ial pho og aphs (app ox-
ima ely 30-cm esolu ion). These we e modeled using
DEM-based and geog aphical a iables (calib a ion wi h
1458 cells and e alua ion wi h 625). The GAM wi h he
inal explana o y a iables was ecalib a ed using all 2083
cells and applied o he 6571 g id cells in he s udy (de ails
gi en in Suppo ing In o ma ion).
O he wise, geodi e si y pe g id cell was calcula ed
ollowing Hjo and Luo o (2012), as ollows. Hyd ologi-
cal ea u e ichness was he sum o di e en hyd ological
ea u e ypes. Fea u es we e mapped om he Na ional
Land Su ey o Finland’s da abase (Table 2) (NLS 2007).
Soil- and ock- ype ichness we e measu ed by summing
he numbe o di e en soil and ock ypes, espec-
i ely. Soil and ock ypes we e de i ed om digi al soil
and bed ock maps, espec i ely, bo h p oduced by he
Geological Su ey o Finland (Table 2) (GSF 2010a,
Conse a ion Biology
Volume 31, No. 2, 2017

Tukiainen e al. 369
Table 2. De ails o he ea u es o classes o geology, soil, and hyd ol-
ogy on which geodi e si y (geo ichness) a iables we e calcula ed o
analyzing spa ial ichness and a i y-weigh ed ichness pa e ns ac oss
Finland’s na ional pa ks.
Geodi e si y
a iable Fea u es o classes
Rock- ype
ichness
ul ama ic in usi e o olcanic ocks
ma ic in usi e o olcanic ocks
in e media e, in usi e olcanic ocks
g ani ic o esic ocks
peli ic sedimen a y ocks
conglome a es
a kosic sedimen a y ocks
black schis s
qua z- ich sedimen a y ocks
sedimen a y ca bona e ocks o
ca bona i es
gneisses and migma i es
i on o e
high-g ade me amo phic ocks
me asoma ic ocks
impac mel ocks
sulphide o e
Soil- ype ichness ock (ba e ock o hin soil co e ;
<1m)
ill (glacigenic deposi s)
s one and block ields
sand and g a el
sil
clay
gy ja (lake and sea sedimen s;
>6% o ganic ma e ial)
pea
Hyd ological
ea u e ichness
lakes (>1ha)
ponds (<1ha)
la ge i e s (>5 m wide)
small i e s (2–5 m wide)
s eams (<2 m wide), sp ings
2010b). Clima e da a o 1981–2010, a 1-km² esolu ion,
we e de i ed om he Finnish Me eo ological Ins i u e
(Pi inen e al. 2012; Table 1).
We emo ed highly co ela ed (Spea man’s ank co -
ela ion coe icien , | s|>0.7) clima ic and opo-
g aphic a iables a e p elimina y analysis o a oid mul i-
collinea i y. Selec ion o he inal a iables was based on
hei mu ual co ela ions and concep ual ele ance and
designed o ob ain he same numbe o geo ichness a i-
ables as nongeo ichness a iables (clima e and opog a-
phy). The e o e, 2 clima e ( ep esen ing ene gy and mois-
u e a ailabili y [Hawkins e al. 2003]) and 2 opog aphic
a iables we e selec ed o ma ch he 4 geo ichness a i-
ables (Table 1). Based on he p e ious s eps, we used
ock- ype ichness, soil- ype ichness, geomo phological
di e si y, hyd ological ea u e di e si y, g owing-deg ee
days, mean annual p ecipi a ion, ele a ional ange, and
ange o he TWI.
Analyses
We used boos ed eg ession ees (BRTs) o analyze
he pa e ns o h ea ened species ichness and RWR.
Boos ed eg ession ees a e an ensemble modeling
me hod in which eg ession ees a e applied ( om he
classi ica ion and eg ession- ee g oup o models) and
hen boos ed o combine a collec ion o models (Eli h
e al. 2008). The BRT models we e i ed in R e sion
3.1.3 (R Co e De elopmen Team 2008) wi h he gbm
package ( e sion 2.1.1) (Ridgeway 2015) and he unc-
ion gbm.s ep, which uses egula iza ion me hods o dis-
cou age o e i ing and balance p edic i e pe o mance
wi h model i (Has ie e al. 2001). We used a ee com-
plexi y o 4, lea ning a e o 0.001, bag ac ion o 0.5,
and a Gaussian e o dis ibu ion. A e expe imen a ion,
all o he a gumen s used de aul alues. Models we e
in e p e ed based on p edic o s’ ela i e in luence (RI)
alues, which can be hough o as model con ibu ions.
Using RI alues allows hese complex ensemble models o
be easily in e p e ed. Rela i e in luence alues a e based
on weigh ing he numbe o imes a p edic o is used
o spli ing a ee acco ding o he imp o emen o he
model as a esul o each spli (F iedman & Meulman
2003).
Al hough BRTs handle da a se s wi h many ze os ea-
sonably well (Eli h e al. 2008), e y la ge p opo ions o
ze os may be p oblema ic. Fu he mo e, some absences
(i.e., g id cells wi h no h ea ened species eco ds) ep-
esen loca ions whe e he condi ions a e oo ha sh o
any o he h ea ened species o exis , which a o s cli-
ma ic a iables in he modeling. This is app op ia e o
b oad-b ush modeling, bu geodi e si y is hypo hesized
o be mos use ul in modeling biodi e si y a ine scales
(Lawle e al. 2015), allowing explana o y powe whe e
he en i onmen is simila in o he espec s (pa icula ly
clima ically). The e o e, o educe he in luence o ab-
sences and cons ain he analyses o clima es likely o
con ain h ea ened species, we e an he species- ichness
analyses wi h da a se s in which he absence cells we e
sampled. We used only cells wi h h ea ened species
eco ds (p esences) and cells immedia ely su ounding
hose p esences. Analyses pe o med wi h he ull da a
se had om 6317 o 6560 absences, depending on he
axonomic g oup, whe eas analyses wi h he sampled
da a had om 34 o 339 absences. This me hod o sam-
pling he absences was in ended o ocus he esul ing
models on dis inguishing cells ha con ained h ea ened
species om o he wise simila cells ha did no (on he
basis ha neighbo ing cells end o be simila because
he en i onmen is spa ially au oco ela ed) o add ess
he hypo hesis ha geodi e si y is impo an o dis in-
guishing o he wise simila en i onmen s.
We an BRT models o he ull se o h ea ened
species (all 16 axonomic g oups combined) and sep-
a a ely o he h ea ened species ichness o each o
Conse a ion Biology
Volume 31, No. 2, 2017
370 Geodi e si y and Th ea ened Species
8 axa wi h su icien numbe s o eco ded h ea ened
species o model indi idually. Models we e un on bo h
he sampled and ull da a se s o ascula plan s, ungi,
bee les, and b yophy es. Due o da a quan i y, models
could no be un on he sampled da a o lichens, bu e -
lies and mo hs, molluscs, and mammals. Analyses wi h
RWR alues only in ol ed he g id cells wi h h ea ened
species eco ds, which dec eased da a quan i y and lim-
i ed analyses o ascula plan s, b yophy es, ungi, and all
species combined.
Sel -s a is ics (SS) we e used o assess in e nal model
i . We hen e alua ed ou models wi h 10- old c oss-
alida ion (CV), which, along wi h SS, is included in
he gbm package (Ridgeway 2015). Sel -s a is ics and CV
ange om 0 o 1; a highe numbe sugges s a be e
model. The CV p ocedu e andomly selec s da a om
he a ea wi hin which he model was calib a ed, ex-
cludes hese da a om he calib a ion, and hen es s he
o iginal model on his held-back po ion o da a. This is
epea ed 10 imes o gi e an a e age co ela ion be ween
he aining and es ing da a. To es whe he model i
e lec ed mo e han spa ial au oco ela ion o he a i-
ables, we eassessed he i s o all he BRT models by
geog aphically sepa a ing calib a ion and e alua ion da a
and calcula ing he oo mean-squa ed e o o p edic ed
and ac ual alues o he e alua ion da a. The di ision o
he g id cells in o e alua ion and calib a ion da a se s was
made a he na ional pa k le el by di iding he pa ks in
an app oxima ely eas –wes di ec ion. The i s da a se
consis ed o 46% o he g id cells om 10 na ional pa ks,
which we e mos ly in he eas , and he second da a se
consis ed o g id cells om he 21 emaining na ional
pa ks (Fig. 1). F om hese 2 da a se s, we chose he one
wi h mo e p esence cells in a gi en axonomic g oup as
he calib a ion da a and he o he as he e alua ion da a.
Resul s
Richness o Th ea ened Species
The bes models pe o med well (SS o CV alue close o
1) o some axa (e.g., all species, molluscs, and ascula
plan s) and poo ly (SS o CV alue close o 0) o o he s
(e.g., lepidop e a and mammals) (Table 3 shows he mod-
els o ull da a se ). The mo e di icul ask o modeling
di e ences in he numbe o h ea ened species be ween
simila (neighbo ing) cells had lowe le els o success
(Table 3, sampled da a se ).
G owing-deg ee days and mean p ecipi a ion we e usu-
ally he dominan p edic o s o h ea ened species ich-
ness; ele a ional ange and TWI ange we e also impo -
an o h ea ened ascula plan s and b yophy es, e-
spec i ely (Table 3). In e ms o de e mining he numbe
o h ea ened species p esen in he sampled da a se ,
geodi e si y a iables con ibu ed ela i ely mo e han
o he ull da a se (Table 3); hei g ea es ela i e con-
ibu ion was 24.7% o ascula plan s. O he geodi e -
si y a iables, geomo phological ichness was he mos
impo an o mos axa, whe eas ock- ype ichness was
he mos impo an o lichens and ascula plan s.
Using he ull da a se , he combined model in luence
(con ibu ion) o nongeodi e si y a iables anged om
86.6% (bu e lies and mo hs) o 99.4% (molluscs). The
combined con ibu ion o geodi e si y a iables he e-
o e anged om 0.6% o 13.4%. A his scale, he numbe
o h ea ened species was hus de e mined p ima ily by
clima e, and mos o hem occu ed in ela i ely wa me
and we e a eas. Geodi e si y a iables we e mos im-
po an in de e mining he numbe o h ea ened species
o Lepidop e a and ascula plan s (CV co ela ion o
ascula plan s was highe han ha o all o he species
g oups) (Table 3). In he sampled models, geodi e si y
a iables had 1.5–2 imes mo e in luence and he impo -
ance o clima e- ele an a iables was lowe han in he
ull model.
Ra i y-Weigh ed Richness o Th ea ened Species
G owing-deg ee days we e s ongly posi i ely ela ed o
RWR o ascula plan s, ungi, and all species combined,
whe eas his ela ionship was nega i e o b yophy es
(Table 4). High le els o RWR o h ea ened species we e
also associa ed wi h high ain all, TWI ange, and hyd o-
logical ea u e ichness o ungi; low ain all, small ele a-
ional ange, and high hyd ological ea u e ichness and
soil- ype ichness o ascula plan s; high soil- ype ich-
ness o b yophy es; and low ain all and high ele a ional
ange and ock ichness o o al h ea ened species ich-
ness. In e nal model i s (SS) we e easonable, whe eas
CV s a is ics we e weake han SS, as would be expec ed.
Compa ed wi h he h ea ened species ichness models,
he combined e ec o geodi e si y a iables was much
g ea e o ascula plan s (28.9% g ea e in luence om
geodi e si y ela i e o he nonsampled ichness model)
and b yophy es (22.5%) in he RWR analyses, whe eas
o ungi and o all species combined, he e was li le
di e ence.
Discussion
Ou esul s om modeling he di e si y o 271 h ea -
ened species ac oss Finnish na ional pa ks e ealed ha
he numbe o g owing-deg ee days and mean annual
p ecipi a ion we e o he g ea es impo ance o h ea -
ened species ichness (bo h aw and weigh ed by a i y).
Ele a ional ange, a widely used opog aphic me ic, was
also impo an . These esul s a e eassu ing gi en he
la ge body o knowledge buil up o e cen u ies wi h
ega d o o e all species ichness (Field e al. 2009; S ein
e al. 2014). The use o geodi e si y– he di e si y o Ea h
Conse a ion Biology
Volume 31, No. 2, 2017
Tukiainen e al. 371
Table 3. C oss- alida ion (CV) co ela ion (co .) and de iance (de .), sel -s a is ics (SS), oo -mean-squa e e o (RMSE) alues, and dominan (g ea es ela i e in luence o model con ibu ion)
geodi e si y (GD) and nongeodi e si y (non-GD) p edic o sao and o al combined geodi e si y ela i e in luence (RI [%]) on ichness o h ea ened species om boos ed eg ession- ee modeling
wi h he ull (F) and sampled (S) da a se s o each axona.
Taxon
Da a used
(n cells) CV co . CV de . SS mean RMSE
Non-GD
p edic o
(highes RI%)
Non-GD
p edic o
(second
highes RI%)
Combined
non-GD model
in luence (%)
GD p edic o
(highes RI%)
GD p edic o
(second
highes RI%)
Combined GD
model
in luence (%)
All F (6571) 0.56 1.39 0.66 1.22 MP (39.4)bGDD (36.9) 93.2 ock ich (3.2) GM ich (2.7) 6.8
S (856) 0.41 8.63 0.59 3.12 MP (41.1)bGDD (28.2) 90.3 ock ich (3.7) GM ich (3.1) 9.7
B yophy es F (6571) 0.40 0.05 0.57 0.14 GDD (32.1)bMP (26.1) 90.2 GM ich (5.6) ock ich (2.8) 9.8
S (191) 0.06 1.20 0.46 0.80 TWIR (27.3)bGDD (24.4) 78.8 GM ich (9.5) hyd o ich (8.4) 21.2
Bee les F (6571) 0.34 0.04 0.51 0.15 GDD (63.2)bMP (20.3) 92.9 GM ich (5.2) ock ich (0.8) 7.1
S (110) 0.14 1.64 0.50 0.96 GDD (35.4)bER (25.5) 87.7 GM ich (7.7) ock ich (2.3) 12.3
Fungi F (6571) 0.38 0.60 0.53 0.71 MP (41.1)bGDD (36.4) 94.5 GM ich (3.2) soil ich (1.1) 5.5
S (593) 0.22 5.11 0.48 1.89 MP (36.3)bGDD (25.5) 85.6 GM ich (8.2) hyd o ich (3.6) 14.4
Lepidop e a F (6571) 0.28 0.01 0.42 0.05 MP (62.48)bGDD (16.4) 86.6 GM ich (9.6) soil ich (2.3) 13.4
Lichens F (6571) 0.33 0.23 0.54 0.50 GDD (53.7)bMP (38.9) 95.7 ock ich (2.7) GM ich (1.3) 4.3
Mammals F (6571) 0.22 <0.01 0.47 0.05 GDD (38.0)bMP (35.6) 93.6 GM ich (3.1) soil ich (2.7) 6.4
Molluscs F (6571) 0.58 <0.01 0.59 0.04 GDD (69.5)bMP (25.3) 99.4 GM ich (0.5) ock ich (0.1) 0.6
Vascula plan s F (6571) 0.59 0.09 0.72 0.17 GDD (29.3)bER (24.3) 88.3 ock ich (4.8) GM ich (3.5) 11.7
S (319) 0.44 1.08 0.66 0.79 ER (30.0)bGDD (19.1) 75.3 hyd o ich (11.6) GM ich (9.1) 24.7
aAll p edic o s in he able showed a posi i e co ela ion be ween he a iable and p edic ed species ichness in he models. See Table 1 o de ini ions o a iable abb e ia ions.
bG ea es con ibu ing p edic o pe axon.
Conse a ion Biology
Volume 31, No. 2, 2017
372 Geodi e si y and Th ea ened Species
Table 4. Resul s o analysis o a i y-weigh ed ichness (RWR) in ela ion o he geodi e si y, clima e, and opog aphy. Sel -s a is ics (SS) and c oss- alida ion s a is ics (CV) om he boos ed
eg ession- ee modeling o RWR a e also shown.
Abio ic p edic o ela i e in luence (%)a,b
Taxon Sel -s a is ic
C oss-
alida ion
s a is ic GDD MP ER TWIR CNGDc ock ich soil ich GM ich hyd o ich CGDd
All 0.58 0.33 27.07 –26.08 27.54 7.93(−) 88.62 6.58 0.55(−) 3.16(−) 1.09 11.38
B yophy es 0.55 0.41 49.21(−) –5.55 6.02 6.95 67.73 0.04(−) 30.02 1.84 0.37(−) 32.27
Fungi 0.59 0.28 44.33 18.57 7.58 22.78 93.26 0.07 0.10 0.49 6.08 6.74
Vascula plan s 0.62 0.19 27.39 10.92(−)13.98(−) 7.09 59.38 8.13 12.85 6.38 13.26 40.62
a(−) indica es nega i e ela ionship be ween p edic o and a i y-weigh ed ichness.
bP edic o s and abb e ia ions a e mo e ully desc ibed in Tables 1 and 2.
cCNGD, combined nongeodi e si y p edic o s’ absolu e model in luence.
dCGD, combined geodi e si y p edic o s’ absolu e model in luence.
su ace ma e ials, o ms, and p ocesses–imp o ed model
p edic ions, especially when we sampled he da a o in-
clude only he ange o en i onmen s in which h ea -
ened species a e known o occu . This sugges s ha
clima ic g adien s may de e mine he egional species
pools, and geophysical ac o s and local he e ogenei y
ha e a g ea e in luence a ine scales. To assess he
e ec o he sampling s a egy, we also an BRT models
using a andom subse o absence cells; he esul s we e
e y simila o hose epo ed abo e. Ou esul s a e o
some ex en con ingen on he iden i ies and ela i e
numbe s o a iables used in he analysis, al hough ou
explo a o y da a analyses and model checking indica ed
ha ou conclusions we e obus o hese issues.
Ou indings a e consis en wi h hypo hesis H1and
wi h p e ious s udies ha indica e he mal condi ions
and ene gy a ailabili y a e among he majo limi ing ac-
o s o species pa e ns in high la i udes, especially a
la ge geog aphic ex en s (Hawkins e al. 2003; Field e al.
2009). This should apply pa icula ly o Finland, which
is a long and na ow coun y ha ex ends o e 1000 km
no h o sou h and hus has a s ong la i udinal g adien
in clima e. Sampling he da a con olled clima e o some
ex en because i emo ed om conside a ion he cells
ha we e a om places wi h h ea ened species. E en
so, he cells spanned nea ly he en i e leng h o Finland,
and clima e emained ai ly dominan in he models.
Al hough he s ong modeled e ec o g owing-deg ee
days was consis en and always posi i e, h ea ened
species ichness o di e en axonomic g oups had
unique ela ionships wi h clima e, opog aphy, and geo-
di e si y a iables, in line wi h he habi a he e ogenei y
hypo hesis (H2). This was mos p ominen in analyses
o he sampled da a, whe e absences we e es ic ed o
cells neighbo ing p esences. We ound suppo speci -
ically o he hypo hesis ha geodi e si y a iables im-
p o e models o e and abo e clima e and opog aphy
(H3). In he models, geodi e si y a iables gene ally had
ela i ely small (consis en wi h Ande son e al. 2015)
bu consis en addi i e e ec s and imp o ed he di e -
si y models’ p edic i e abili y o bee les, b yophy es,
and ascula plan s. Geodi e si y p edic o s we e mos
impo an when using he sampled da a, a he han he
ull da a se . Thus, al hough clima e is a key d i ing o ce,
explici ly p io i izing he di e si y o geophysical se ings
may help conse e abio ic and bio ic di e si y in a dy-
namic clima e.
O he geodi e si y a iables, ock- ype ichness was
ela i ely impo an o he ichness ( aw and a i y
weigh ed) o h ea ened ascula plan species and o
h ea ened lichen species ichness. Elsewhe e, ock ich-
ness has been i mly linked o biodi e si y o lichens
(Spi ale & Nascimbene 2012) and plan s (e.g., Pausas
e al. 2003; Kougioumou zis & Tiniakou 2014). Geomo -
phological ichness was consis en ly signi ican (i no
always s ong) in ou models, especially o aw ichness
Conse a ion Biology
Volume 31, No. 2, 2017