Modelling he his o ic dis ibu ion and habi a o Ame ican
ches nu (Cas anea den a a) in Geo gia, USA using edaphic and
land o m p edic o s
Joyce M. Klaus1, Na han A. Klaus2
1 Te a-Ignea En e p ises LLC, 4684 GA Hwy 83 S., Culloden, Geo gia, 31016, USA
2 Wildli e Conse a ion Sec ion, Wildli e Resou ces Di ision, Geo gia Depa men o Na u al Resou ces, 116 Rum C eek D ., Fo sy h,
Geo gia, 31029, USA
Co esponding au ho : Joyce M. Klaus (joycema ieklaus@p o on.me)
Edi o Jane F anklin
Recei ed
12 June 2025♦
Accep ed
17 Oc obe 2025♦
Published
24 No embe 2025
F on ie s o Biogeog aphy 18, 2025,
e161937|DOI 10.21425/ ob.18.161937
Copy igh Joyce M. Klaus and Na han A. Klaus. This is an open access a icle dis ibu ed unde he e ms o he C ea i e Commons
A ibu ion License (CC BY 4.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal
au ho and sou ce a e c edi ed.
RESEARCH ARTICLE
FRONTIERS OF
BIOGEOGRAPHY
The scien i ic jou nal o
The In e na ional Biogeog aphy Socie y
Abs ac
The loss o Ame ican ches nu (Cas anea den a a Ma sh.
Bo k.) caused ecological change in many communi y ypes
o he eas e n Uni ed S a es. Res o a ion is challenged by
clima e change and wo na u alised in asi e non-na i e
diseases, ches nu bligh (caused by C yphonec ia pa a-
si ica [Mu ill] M.E. Ba ) and Phy oph ho a oo - o (caused
by Phy oph ho a cinnamomi Rands). Lea ning how o o e -
come hese challenges in he sou he n po ion o he o me
ches nu ange may make ange-wide es o a ion e o s
mo e success ul because hese p essu es a e likely s on-
ges in sou he n po ions o he ange. To es ablish a base-
line o ches nu dis ibu ion and en i onmen al co ela es,
we used ca. 1830 land lo e y maps o documen he his o ic
abundance and dis ibu ion o ches nu in Geo gia. Land lo -
e y su eyo s documen ed 717,901 ees wi hin ou s udy
a ea, iden i ying 15,710 as ches nu . We used hei da a o
c ea e a species dis ibu ion model wi h soil and land o m
p edic o s, ac o s possibly ela ed o disease isk and p e-
dic ed he ela i e habi a sui abili y o ches nu h oughou
much o he S a e. Ou esul s e ealed ha long-held as-
sump ions abou ches nu ange and abundance in Geo gia
a e inco ec . Ches nu was less common in no he n Geo -
gia and mo e common in he Piedmon han p e iously pos-
ula ed, and ches nu anged well in o he Coas al Plain. Soil
and land o m a iables adequa ely p edic ed ches nu dis-
ibu ion and habi a sui abili y models p edic ed ha ches -
nu occupied a wide di e si y o habi a s; habi a p e e enc-
es we e complex and di e ed by physiog aphic p o ince. All
he models ag eed ha ches nu p e e ed highe ele a ion,
mo e slope and land o m cu a u e, bu he ela ionship o
ches nu o o he a iables was di icul o gene alise and
may ha e been con ex dependen . These esul s can be
used o guide e-in oduc ion and maximise success in he
ace o changing clima e and non-na i e disease isk.
Highligh s
• Anecdo al accoun s limi he his o ic ange o Ame i-
can ches nu in Geo gia o he Blue Ridge, Appalachian
Pla eau and Ridge and Valley physiog aphic p o inc-
es; howe e , his o ic land lo e y maps da ing om
he ea ly 1800s documen ed ches nu h oughou he
Geo gia Piedmon and much o he Coas al Plain;
• Ches nu was ound in a wide ange o Geo gia hab-
i a s in e e y physiog aphic p o ince be o e he in-
oduc ion o non-na i e pa hogens, some imes in
high abundance and some imes as less dense com-
ponen s o habi a s wi h o he dominan o co-dom-
inan species;
• Al hough pa e ns o ches nu dis ibu ion we e
la gely d i en by ele a ion and slope, o he en i on-
men al ac o s including aspec , land o m cu a u e,
numbe o os - ee days, pe cen age clay in soil
and soil pH signi ican ly in luenced ches nu dis i-
bu ion in Geo gia. These pa e ns a ied by physio-
g aphic p o ince, sugges ing ches nu habi a was
no mono ypic ac oss Geo gia;
• Res o a ion e o s ha ecognise egional di e -
ences and he ull sui e o di e se habi a s ha once
included ches nu may be mo e likely o succeed.
Ou models and egional habi a sui abili y maps
can be used o guide plan ing si e selec ion, while
also conside ing he la ge a iabili y inhe en in eco-
logical es o a ion o ches nu ac oss mul iple phys-
iog aphic p o inces;
• As hey a e ela i ely ine scale, edaphic and land-
o m ac o s a e use ul in p edic ing sui able habi a
o plan s and may be used in conjunc ion wi h cli-
ma e change p edic ions o guide si e selec ion o
species ein oduc ions ha will be sus ainable in o
he o eseeable u u e.
F on ie s o Biogeog aphy 18, 2025, e161937
Joyce M. Klaus and Na han A. Klaus
2
Keywo ds
Ame ican ches nu , Cas anea den a a, land lo e y map,
habi a sui abili y, his o ic dis ibu ion, land o m, soil,
species dis ibu ion model, wi ness ee
In oduc ion
Ame ican ches nu (Cas anea den a a [Ma shall] Bo khau-
sen, hence o h ches nu ) was once a undamen al com-
ponen o o es ecosys ems ac oss 29 eas e n U.S. S a es
(co e ing > 800,000 km2) and wo eas e n Canadian P o -
inces. The unc ional elimina ion o ches nu om i s en i e
ange has caused his species o be lis ed as ulne able o
c i ically impe illed in 14 S a es and p o inces and caused
a loss o di e si y in nume ous communi y ypes as well as
unknown changes in plan and animal popula ions h ough-
ou i s o me ange (B aun 1950; Hawkins 2006; Ellio
and Swank 2008; W igh e al. 2022; Na u eSe e 2024).
Gene ic, b eeding and es o a ion p og ammes ha e been
es ablished o es o e he species (G i in 2000; Anagnos-
akis 2012; Cla k e al. 2014) and nea ly a cen u y o wo k
has b ough conse a ionis s close o his achie emen
(Wes b ook e al. 2019; Wes b ook e al. 2020).
The demise o ches nu was caused by wo non-na-
i e in asi e diseases; i s , Phy oph ho a oo o (PRR)
(causal o ganism: Phy oph ho a cinnamomi Rands) in he
mid-1800s, ollowed by ches nu bligh (causal o ganism:
C yphonec ia pa asi ica [Mu ill] Ba ) in he ea ly 20 h
cen u y (Anagnos akis 2001). Bo h e en s we e poo ly
documen ed by mode n s anda ds (Paille 2002; Jacobs
2007) and li le is known abou he ull his o ic ches nu
dis ibu ion o ange o ecological communi ies in which
ches nu occu ed be o e PRR (Woods 1953; Kane e al.
2020; Na u eSe e 2024). Remnan oo -sp ou s and an
occasional ma u e ches nu ee emain h oughou he
o me ange, bu hese a e a poo measu e o he o me
ull ex en o he species. Al hough some anecdo al his-
o ic accoun s desc ibe s ands o pu e ches nu , his le el
o abundance could ha e been a misleading elic o p io
imbe ha es and was likely he excep ion (F o hingham
1924). Ul ima ely, he e emains no consensus on he dis-
ibu ion o abundance o ches nu p io o hese disease
ou b eaks o guide es o a ion, pa icula ly in he sou he n
po ions o he ange whe e PRR may ha e been he mos
des uc i e and whe e losses p eda e mos o es and bo-
anical in en o ies (Li le 1977; Faison and Fos e 2014).
Unde s anding ches nu in he sou he n po ions o i s
o me ange and in Geo gia, in pa icula , may be aluable
o ange-wide eco e y e o s. Wi h ele a ions anging
om sea le el o nea he uppe limi s o ches nu ole -
ance (Wang e al. 2013), in o ma ion om Geo gia may
guide eco e y ac oss he sou he n U.S. and a highe la -
i udes (Von Humbol e al. 1807). Due o clima e change,
he p essu e om PRR which un il ecen ly was la gely
limi ed o la i udes below 40 deg ees will likely shi no h-
wa ds (Be go e al. 2004; Gus a son e al. 2022), making
sou he n s a egies o o e come PRR ge mane o eco e y
e o s ange-wide. Clima e change may also make unde -
s anding ches nu habi a in he sou he n limi s o i s o -
me ange aluable o ensu ing his species’ su i al o he
no h as ches nu ange shi s no hwa ds (Adeyemo and
G ainge 2023).
Success ul es o a ion will also depend on a be e un-
de s anding o he habi a needs o his species, as e i-
denced by he mixed success o many es o cha ds and
es o a ion es plan ings. While b eeding p og ammes
show p omise o disease esis ance, imp o ed ches -
nu s ock will no be immune o PRR o ches nu bligh
(Wes b ook e al. 2019; Wes b ook e al. 2020). Finding
he mos sui able habi a s o es o a ion using habi a
modelling could help mi iga e disease p essu es, in-
c ease plan ing success, ensu e ha ches nu is es o ed
o app op ia e communi ies and imp o e use o limi ed
esou ces. Knowing as many o he places whe e ches -
nu once occu ed as possible, a he han only he mos
dominan o elic si es, may also ensu e i s es o a ion o
he ull ange o communi ies and habi a s possible, im-
p o ing he esilience o es o a ion e o s (F o hingham
1924; Lapo e al. 2022).
Geo gia is especially well-sui ed o his esea ch be-
cause o a sys ema ic su ey o ee species composi ion
conduc ed be ween 1804–1832 as pa o he Geo gia land
lo e y. “I shall be he du y o he su eyo s … o ma k, o
cause o be ma ked, plainly and dis inc ly upon ees, i p ac-
icable, o he wise on pos s, all s a ions and all lines which
hey may be equi ed o un o he pu pose o making he
su eys (and) o cause all such lines o be measu ed wi h
all possible exac ness, wi hin a hal chain, con aining hi -
y- h ee ee ” (Geo gia Gene al Assembly 1825). Al hough
he su eys we e conduc ed o acili a e he edis ibu ion
o land ollowing he o ced emo al and genocide o na-
i e peoples by he Geo gia and Uni ed S a es go e nmen s
(F ench 1978; Bleakley and Fe ie 2010), hey inad e en ly
documen ed o es composi ion p io o la ge-scale o es
clea ing, he a i al o ches nu bligh and possibly be o e
PRR was widesp ead (F o hingham 1924; C andall 1945;
Anagnos akis 1995; Anagnos akis 2012) by eco ding ‘wi -
ness ees’ along land lo bounda ies.
O he esea che s ha e used species dis ibu ion mod-
elling echniques o in es iga e en i onmen al ela ion-
ships o Ame ican ches nu (see Tulowiecki (2025) o a
summa y). Species dis ibu ion modelling, also known as
habi a sui abili y modelling, was speci ically de eloped
o ela e geospa ial species occu ence da a o en i on-
men al a iables using a ious echniques ha can ac-
commoda e i ually all ypes o p edic o da a (con inu-
ous and ca ego ical), esponse da a (p esence/absence
o p esence only), ela ionships (simple linea o complex
non-linea ) and app oaches (pa ame ic and non-pa ame -
ic) (Guisan and Zimme mann 2000; Pe e son e al. 2011).
Which app oach wo ks bes is a ma e o scale, he ypes
o da a a ailable, he ela ionship o ches nu o i s en i-
F on ie s o Biogeog aphy 18, 2025, e161937
Ches nu dis ibu ion in Geo gia
3
onmen wi hin he ange explo ed o he s udy and he
in ended use o he esea ch (Tulowiecki 2025). Fo in-
s ance, some esea che s a e in e es ed in ange shi s
due o clima e change so hey may use clima e da a a a
scale ela i e o he species ange (e.g. Ba nes and Del-
bo ne (2019); Adeyemo and G ange (2023)), while o he
esea che s a e in e es ed in species e-in oduc ion so
hey may use edaphic ac o s a a mo e si e-speci ic scale
(e.g. Fei e al. (2007); Zhang e al. (2019)). While ou e -
o lacked clima e da a ha was con empo a y wi h land
lo e y su eys, Geo gia has o he p edic o a iable da a
a ailable on a scale ≤ 100 m2 including a wide ange o
ele a ion, soil ypes and quali ies and physiog aphic p o -
inces wi h a ying geologic o igins ha can be assumed
o ha e changed li le since he 1800s.
Ou objec i es we e o: a) model he his o ical dis i-
bu ion (i.e. c ea e a species dis ibu ion model, SDM) o
ches nu in Geo gia o de e mine which en i onmen al
a iables we e co ela ed wi h ches nu p esence and b)
o c ea e a habi a sui abili y map, based on he bes pe -
o ming species dis ibu ion model o se o models o
he pu pose o assis ing si e selec ion o species e-in-
oduc ion. In addi ion o modelling a he scale o he
whole S a e, we also modelled ches nu dis ibu ion and
habi a in each o Geo gia’s i e dis inc physiog aphic
p o inces (Fenneman and Johnson 1946) indi idually as
p io esea ch demons a ed ha ches nu SDM esul s
a e dependen on scale (Hende son e al. 2023) and phys-
iog aphic p o ince (Black e al. 2002).
Ma e ials and me hods
Ou s udy ook place in he S a e o Geo gia, Uni ed S a es
o Ame ica. A 153,910 km2, Geo gia is a ela i ely la ge
and di e se S a e. I con ains i e dis inc physiog aphic
p o inces, se e al moun ainous, o he s olling o ela-
i ely la . Soils a y widely om sandy o ocky o hea y
clay. Geo gia has he six h highes ee species di e si y
o he 50 Uni ed S a es and he hi d highes ee di e si y
wi hin he o me ange o ches nu . Mode n and his o ic
o es ypes wi hin ou s udy a ea ange om pine domi-
na ed o ha dwood domina ed. Ele a ions ange om sea
le el o 1,458 m, nea he uppe limi s o ches nu ole -
ance (Wang e al. 2013).
Land lo su eyo s loca ed, blazed (ma ked) and iden i-
ied wi ness ees a and nea he bounda y in e sec ions
o all lo s which we e ypically 82 hec a es, bu wi h a mi-
no i y o lo s su eyed o 16, 65, 82, 101 o 198 ha lo sizes;
lo s we e squa e. Wi ness ees we e selec ed o be la ge
enough o blaze and as close as possible o he lo bound-
a y, wi hin 10 me es. T ees we e blazed whe e lo co ne s
me , one ee a each bounda y in e sec ion ( ep esen ing
ou lo co ne s) and a single ee on each bounda y line
ex ending om each in e sec ion in he ou ca dinal di-
ec ions; hus each bounda y in e sec ion whe e ou land
lo s me , a o al o i e ees we e blazed, a anged in he
shape o a + sign wi h one ee in he middle and one ee in
each o he ca dinal di ec ions. T ee species, hei loca ion
and lo numbe s we e compiled on o hand d awn maps
which we accessed a he Geo gia S a e A chi es and he
Geo gia GIS Clea inghouse (Geo gia Spa ial Associa ion
2024). Remains o some blazed ees a e also housed a
he Geo gia S a e A chi es; lo numbe s ca ed in o he
ees we e app oxima ely 15 cm all and 10 cm wide, and
many land lo s we e iden i ied by h ee digi s, hus we
specula ed blazed ees we e likely 30–35 cm diame e a
b eas heigh o la ge .
We downloaded digi ised, geo e e enced Geo gia land
lo e y maps (1804–1832) om he Geo gia GIS Clea ing-
house (Geo gia Spa ial Associa ion 2024). Using A cGIS
P o .3.1.0, we sys ema ically sea ched all 392 maps o
Ame ican ches nu loca ions (indica ed on mos maps by
“Ches” o “ches nu ”, no be con used wi h ches nu oak
which was designa ed by su eyo s as “Moun ain Oak”, “M
Oak”, “Ch Oak” o “Ches O”) and c ea ed a poin in a ec-
o laye o each loca ion. Fo each physiog aphic p o -
ince (Appalachian Pla eau, Blue Ridge, Ridge and Valley,
Piedmon and Coas al Plain as de ined by Fenneman and
Johnson (1946)), we di ided he numbe o ches nu s de-
ec ed in ha p o ince by he o al numbe o ees sam-
pled in ha p o ince o es ima e he pe cen age o ees
ha we e ches nu by physiog aphic p o ince. We did no
include a eas beyond whe e land lo e y su eyo s docu-
men ed ches nu when calcula ing pe cen age ches nu in
he Coas al Plain.
We downloaded he G idded Soil Su ey Geog aphic
(gSSURGO) Da abase (NRCS Da a 2025) o he S a e o
Geo gia, clipped he ex en o ma ch he ches nu da a
ex en (land lo e y da a only co e ed ~ 70% o he S a e,
see Fig. 1) and using A cGIS P o joined “Componen ” and
“Cho izon” ables o he 10 m map uni as e . We isola -
ed he ollowing soil a ibu es om he gSSURGO ables
ha we e p e iously shown o assumed, based on expe
opinion o a ec Ame ican ches nu dis ibu ion (Russell
1987; Fei e al. 2007; Rhoades e al. 2009; Tulowiecki
2020; Hende son e al. 2023): pa icle size class, d ainage
class, pe cen age clay, e osion class, ep esen a i e soil
pH, numbe o os - ee days, pa en ma e ial kind, pa -
en ma e ial o igin, soil axonomic class (o de , subo de ,
g ea g oup and subg oup) and we c ea ed independen
as e laye s o each o hese a iables (see SSURGO
me ada a ables and columns epo [NRCS Da a] and
Keys o Soil Taxonomy Thi een h Edi ion [Soil Su ey
S a 2022] o desc ip ions o a iables). We used 1/3
a c-second (10 m) digi al ele a ion maps om he Na ion-
al Ele a ion Da ase (USGS 2022) o c ea e an ele a ion
as e laye in A cGIS P o. F om he ele a ion as e laye ,
we used A cGIS ools o calcula e and c ea e as e laye s
o ele a ion (m), aspec (deg ees), slope (deg ees) and
cu a u e (amoun o land o m conca i y o con exi y [-1
o +1, espec i ely wi h 0 being la ]). As aspec is a a-
dially dis ibu ed a iable wi h 1° and 359° being nea ly
equal, we con e ed aspec as e da a o wo linea a i-
ables “eas ness” (sin(aspec °(π/180)) and “no hness”
(cos(aspec °(π/180)). We p ocessed all as e laye s o
F on ie s o Biogeog aphy 18, 2025, e161937
Joyce M. Klaus and Na han A. Klaus
4
ma ch he p ojec ion and ex en o he ches nu laye p o-
duced om he land lo e y da a.
Fo species dis ibu ion modelling, we used he sdm
package .1.1-8 (Naimi and A aújo 2016) o R .4.2.1 (R)
ha allows use s o compa e a ious ela ionships and ap-
p oaches o ind he bes i o da ase s and desi ed ou pu s
(Naimi and A aújo 2016). We checked all 18 p edic o a i-
ables o collinea i y; none had collinea i y issues, so we e-
ained hem all o modelling. We classi ied class, d ainage,
g ea g oup, o de , pa icle size class, pa en ma e ial kind,
subg oup and subo de as ca ego ical p edic o s (NRCS
Da a 2025). As we did no ha e independen aining da a,
we used a subse o ches nu occu ence da a o model
aining. As we did no ha e absence da a, (su eyo s did
no eco d i a ches nu ee was no obse ed), we c ea ed
47,130 andom backg ound poin s, h ee imes he numbe
o ches nu loca ions (Ba be -Massin e al. 2012; Naimi and
A aújo 2016; Sille o and Ba bosa 2020; Whi o d e al. 2024),
using he ‘gRandom’ unc ion ( andom in geog aphic space).
To compensa e o limi ed compu ing powe , we con e ed
p edic o s o 100 m esolu ion om a 10 m esolu ion using
A cGIS Spa ial Analys Resample ool.
Using he p ocessed en i onmen al as e da a, he
ches nu occu ence poin s and he backg ound poin s,
we c ea ed a ches nu species dis ibu ion model using
he sdm package o R. We e alua ed model pe o mance
using s anda d model selec ion me ics gene a ed by he
sdm o R package such as a ea unde he cu e (AUC),
co ela ion sco e (COR) and ue skill s a is ic (TSS)
( o an explana ion o hese s a is ics, see San ini e al.
(2021)). We in ended o use he ensemble me hod ha a -
e ages se e al use -selec ed modelling me hods including
boos ed eg ession ee, gene alised addi i e, gene alised
linea , maximum en opy and andom o es , all commonly
used SDM echniques o Ame ican ches nu (Tulowiecki
2025). We eplica ed each model un h ee imes and ied
bo h subsampling and boo s apping da a pa i ioning
me hods wi h 30% o he da a ese ed o es ing models.
The o iginal land su ey me hods and ou sdm me hods
ensu ed no spa ial au oco ela ion be ween aining and
Figu e 1. Loca ions o Ame ican ches nu (black do s) on Geo gia land lo e y maps (1804–1832). Thin ed lines a e GA coun y bound-
a ies, hick ed lines a e ex en o land lo e y maps, blue lines dis inguish physiog aphic p o inces: A. Appalachian Pla eau, B. Ridge
and Valley, C. Blue Ridge, D. Piedmon and E. Coas al Plain.
F on ie s o Biogeog aphy 18, 2025, e161937
Ches nu dis ibu ion in Geo gia
5
es da a (Naimi and A aújo 2016). Random o es wi h
he boo s apping da a pa i ioning me hod consis en ly
p oduced signi ican ly highe AUC sco es and less o e lap
in densi y plo s ha demons a ed be e abili y o di e -
en ia e be ween ches nu occu ences and backg ound
poin s, so we only epo esul s om andom o es . We
gene a ed a iable impo ance ac o s (no malised o sum
o 1) o de e mine which en i onmen al a iables had he
mos in luence on he model and species esponse cu es
o he andom o es model o isualise how ches nu e-
sponded o indi idual a iables ha had signi ican in lu-
ence on he model.
We di ided all he p ocessed as e and ec o da a by
physiog aphic p o ince: Appalachian Pla eau, Ridge and
Valley, Blue Ridge, Piedmon and Coas al Plain (Fenneman
and Johnson 1946), hen epea ed he en i e p ocess om
checking o collinea i y o modelling o gene a ing a i-
able impo ance ac o s and species esponse cu es
o each indi idual physiog aphic p o ince. Once we had
models o he whole S a e (i.e. all po ions co e ed by he
land lo e y maps) and each o he i e p o inces, we an
he ‘p edic ’ unc ion in sdm o o p edic ela i e habi a
sui abili y, based on he models we c ea ed and w o e he
esul s o as e iles ha we mapped in A cGIS P o.
Resul s
A ho ough inspec ion o he land lo e y maps e ealed
ha no all ees we e iden i ied o species, o exam-
ple, ‘pine’ and ‘bay’ and many colloquial names equi ed
some esea ch o con i m species; 66 ecognisable spe-
cies we e documen ed including many in iguing species
such as ‘chinquapin’ (Cas anea pumila [Linnaeus] P. Mill-
e ) and ‘wahoo’ (Ulmus ala a Michaux). Wi hin ou s udy
a ea 717,901 ees we e sampled, o which 15,710 we e
iden i ied as ches nu (2% o ees sampled, Fig. 1). In he
Coas al Plain po ion o ou s udy a ea (wi hin he ange
o ches nu occu ences), 0.25% o ees sampled we e
ches nu , 3.1% in he Piedmon , 2.8% in he Blue Ridge,
1.0% in he Ridge and Valley and 1.0% Appalachian Pla-
eau. Ches nu abundance, howe e , was no dis ibu ed
homogeneously o andomly ac oss he landscape; many
disc e e egions had highe o lowe abundance han
hese alues (Fig. 1).
The anges o en i onmen al a iables di e ed by p o -
ince (Table 1). The e we e no collinea i y issues amongs
a iables o he whole-S a e model o any o he indi-
idual p o ince models, so we e ained all a iables o
u he analyses. Fo he whole S a e and each p o ince,
we cons uc ed ches nu species dis ibu ion models ha
included all a iables ( ull models), bu some a iables
(pe cen age clay, e osion class, pH and pa en ma e ial
o igin) had oo many da a gaps in he gSSURGO da abase
o use hem o build a use ul p edic i e habi a sui abili y
map, so we also cons uc ed species dis ibu ion models
( educed models) and p edic i e habi a sui abili y maps
wi hou hese a iables.
The ull S a e-wide mean model pe o mance had an
AUC sco e o 0.96, COR sco e o 0.74 and TSS sco e o
0.77, which is conside ed good o e y good, bu no o e -
i ed (Allouche e al. 2006; Shabani e al. 2018) (Table 2).
Ele a ion had he g ea es in luence on he model by an o -
de o magni ude (no malised a iable impo ance sco e o
0.66 compa ed o 0.04–0.10 o o he a iables), ollowed
in o de o impo ance by numbe o os - ee days, slope,
pe cen age clay, cu a u e, pH, no hness and eas ness
(Fig. 2). O he a iables (all soil axonomy a iables, pa i-
cle size class, d ainage class, e osion class, pa en ma e i-
al kind and pa en ma e ial o igin) did no con ibu e signi i-
can ly o he model. The species esponse cu es indica ed
ha he p obabili y o ches nu occu ence inc eased wi h
ele a ion, slope, eas ness, pH, pe cen age clay and any
ype o land o m cu a u e, con ex o conca e (Table 2).
The p obabili y o ches nu occu ence dec eased wi h
numbe o os - ee days. The ela ionship o ches nu o
no hness was oo complex o be gene alised (Fig. 3).
The educed whole-S a e mean model pe o mance
was nea ly he same as he ull model (Table 2). Ele a-
ion had he g ea es in luence on he p obabili y o ches -
nu occu ence, ollowed by slope, numbe o os - ee
days, cu a u e, no hness and eas ness (Fig. 4). Species
esponse cu es we e simila o hose o he ull model,
excep ha ches nu was posi i ely ela ed o no hness
and he ela ionship o numbe o os - ee days was oo
complex o be gene alised (Table 2, Fig. 5).
As he ull model p oduced a habi a sui abili y map
wi h nume ous da a gaps (due o da a gaps in some o
he SSURGO soils a iables), we used he educed whole-
S a e model o cons uc a S a e-wide habi a sui abili y
map (Fig. 6). Due o he o e whelming in luence o ele a-
ion on he species dis ibu ion model and he di e ences
in ele a ion amongs he S a e’s physiog aphic p o inces,
he whole-S a e model sugges s he highes concen a-
Table 1. Range o con inuous en i onmen al a iables conside ed in his s udy.
Physiog aphic
p o ince
Cu a u e
(1/m)
Eas ness and No hness
( ans o med deg ees)
Ele a ion
(m)
F os - ee
days pH Slope
(deg ees) Clay (%)
Appalachian Pla eau -1.19–1.07 -1 - 1 194–717 180–217 4.6–7.0 0–77 9–55
Blue Ridge -1.08–0.95 -1 - 1 234–1458 140–235 4.6–7.3 0–74 6–53
Coas al Plain -0.97–0.91 -1 - 1 3–248 180–287 3.2–7.9 0–74 1–70
Piedmon -1.06–1.46 -1 - 1 61–1003 140–275 4.5–7.0 0–77 3–60
Ridge and Valley -0.90–0.86 -1 - 1 171–868 180–241 4.5–7.5 0–69 3–69
Whole S a e -1.19–1.46 -1 - 1 3–1458 140–287 3.2–7.9 0–77 1–70
F on ie s o Biogeog aphy 18, 2025, e161937
Joyce M. Klaus and Na han A. Klaus
6
ion o highly sui able habi a is in he uppe ele a ions o
he S a e, especially he uppe Piedmon and Blue Ridge.
Indi idual physiog aphic p o ince models anged in pe -
o mance om good o e y good (AUC sco e 0.86–0.95).
The ull and educed Coas al Plain models pe o med bes
(Table 2). The same a iables ha signi ican ly in luenced
he ull and educed whole-S a e models signi ican ly in lu-
enced each o he physiog aphic p o ince models, wi h el-
e a ion he mos in luen ial a iable in e e y model, excep
he ull Coas al Plain model whe e slope in luenced p ob-
abili y o ches nu occu ence mo e han ele a ion (Figs 2,
3, Table 2). The ela i e impo ance o ele a ion compa ed
o o he signi ican a iables was less in he moun ainous
egions o he S a e (Ridge and Valley, Appalachian Pla-
eau and Blue Ridge) han in he Piedmon o Coas al Plain,
ue o bo h ull and educed models. P obabili y o ches -
nu occu ence was posi i ely associa ed wi h ele a ion
and slope, excep o he ull models o he Piedmon and
Blue Ridge whe e he ela ionship wi h slope was di icul
o gene alise (Figs 3–5). The ela ionships o eas ness,
no hness, numbe o os - ee days, pH and pe cen -
age clay wi h ches nu occu ence we e no consis en
amongs ull e sus educed models o amongs physio-
g aphic p o inces. In no case was a ela ionship one di-
Table 2. Model pe o mance sco es and gene alised species esponses o en i onmen al a iables o educed models ( op) and ull
models (bo om). + and – symbols indica e posi i e and nega i e ela ionships, espec i ely. Fo cu a u e, he i s symbol indica es
i ches nu had a ela ionship wi h cu a u e (con ex o conca e) and he symbol in pa en hesis indica es p e e ence o con ex (+) o
conca e (-). Ques ion ma ks indica e no clea unidi ec ional pa e n in species esponse. AUC = a ea unde he cu e, COR = co ela ion
sco e, TSS = ue skill s a is ic.
Reduced Model
Model AUC COR TSS Cu a u e Eas ness Ele a ion F os - ee days No hness Slope
Whole S a e 0.95 0.76 0.77 +(+,-) + + ? + +
Coas al Plain 0.95 0.79 0.82 +(-) + + - - +
Piedmon 0.89 0.66 0.61 +(-) ? + + ? +
Ridge and Valley 0.88 0.65 0.57 +(-) ? + ? + +
Appalach. Pla eau 0.86 0.62 0.56 +(+) ? + - - +
Blue Ridge 0.86 0.63 0.54 +(-) + + + - +
Full Model
Model AUC COR TSS Cu a u e Eas ness Ele a ion F os - ee days No hness Slope pH %Clay
Whole S a e 0.96 0.74 0.77 +(+,-) + + - ? + + +
Coas al Plain 0.96 0.77 0.84 +(-) + + - - + ? ?
Piedmon 0.89 0.67 0.60 +(-) ? + + + ? ? ?
Ridge and Valley 0.86 0.62 0.54 +(-) + + - + + ? +
Appalach. Pla eau 0.87 0.64 0.60 +(+) ? + - - + - -
Blue Ridge 0.86 0.61 0.56 +(+) + + + ? ? - ?
Figu e 2. No malised a iable impo ance sco es o ull ches nu species dis ibu ion models.
F on ie s o Biogeog aphy 18, 2025, e161937
Ches nu dis ibu ion in Geo gia
7
Figu e 3. Species esponse cu es o ull species dis ibu ion models o Ame ican ches nu dis ibu ion in he whole S a e o Geo gia
and in each o i s p o inces indi idually.
Figu e 4. No malised a iable impo ance sco es o educed ches nu species dis ibu ion models.
F on ie s o Biogeog aphy 18, 2025, e161937
Joyce M. Klaus and Na han A. Klaus
8
Figu e 5. Species esponse cu es o educed species dis ibu ion models o Ame ican ches nu dis ibu ion in he whole S a e o
Geo gia and in each o i s p o inces indi idually.
Figu e 6. Ame ican ches nu habi a sui abili y map o A. Whole S a e model o Geo gia and indi idual physiog aphic p o ince models;
B. Piedmon ; C. Appalachian Pla eau; D. Ridge and Valley; E. Blue Ridge and F. Coas al Plain. Thin ed lines a e coun y bounda ies; hick
ed lines ma k he ex en o he land lo e y maps. Da ke o ange is highe ches nu sui abili y; ligh e o ange is lowe ches nu sui abili y.
F on ie s o Biogeog aphy 18, 2025, e161937
Ches nu dis ibu ion in Geo gia
9
ec ion in he ull model (e.g. posi i e) and he opposi e
di ec ion in he educed model (e.g. nega i e) o he same
physiog aphic p o ince. I he e was a di e ence be ween
he ull and educed model, i was because he ela ionship
o a a iable in one model had a clea pa e n (posi i e o
nega i e) and, in he o he model, was oo complex o gen-
e alise as posi i e o nega i e (Table 2, Figs 3–5).
Discussion
The me hods employed by land lo e y su eyo s docu-
men ed a he e ogeneous dis ibu ion o ches nu s ac oss
he 1800s landscape. Su eys documen ed a ches nu
abundance a below mos con en ional es ima es o he
Blue Ridge, Ridge and Valley and Appalachian Pla eau and
much highe in he Piedmon and Coas al Plain (F o hing-
ham 1912; B aun 1950; Saucie 1973; Li le 1977; G i in
2000; TACF 2024). Al hough some bias is appa en in he
wi ness ee da a and di e ences in land lo size com-
plica ed ou analysis, he land lo e y su eys emain he
mos comp ehensi e and leas biased da ase a ailable
o Ame ican ches nu dis ibu ion be o e ches nu bligh
and likely be o e la ge-scale e ec s o PRR (Anagnos-
akis 2012). While ou es ima es o ches nu abundance
con adic many ea lie s udies, se e al o he esea che s
ha e also concluded ha accoun s o a ches nu domi-
na ed landscape a e apoc yphal. A e iew o land lo e y
maps in he no h-eas e n Uni ed S a es by Thompson e
al. (2013) also ound ches nu did no comp ise mo e han
3% o ees in he egion and a s udy o land lo e y maps
in he uppe Midwes came o a simila conclusion (Fais-
on and Fos e 2014). O he lines o e idence using pollen
eco ds (Wa s 1980; Delcou and Delcou 1998) also
concluded ha Ame ican ches nu was no a dominan
species in hei espec i e s udy a eas. The widesp ead
accep ance o hese o e es ima es, mos o en gi en as
‘25% o mo e’ likely o igina ed wi h obse a ions o excep-
ional s ands documen ed by E. Lucy B aun (1956). The e
she documen ed se e al s ands o excep ionally high
ches nu abundance, ye aken in whole, he s udy clea ly
documen ed ha ches nu was no a dominan ee on he
landscape (Faison and Fos e 2014). F o hingham (1924)
no ed ha , because o ches nu ’s abili y o e-sp ou p oli -
ically a e clea cu ing, dense nea ly pu e s ands o ches -
nu we e mos likely an h opogenic in na u e. Despi e se -
e al decades o s ong science o he con a y, he my h
o a landscape domina ed by Ame ican ches nu emains
s ubbo nly p esen .
I is wo h no ing se e al ins ances o appa en su ey-
o bias, whe e he equency o ches nu s documen ed by
su eyo s ab up ly changed along no h/sou h o eas /
wes lines. These we e ound a se e al loca ions whe e
wo o mo e land lo e y maps joined and whe e a su ey
c ew was appa en ly mo e o less inclined o use ches -
nu s as wi ness ees. The easons o his a e unknown,
bu , o una ely o ou pu poses, appea o be limi ed o
ew ins ances and a e y small pe cen age o he ees
sampled, wi h he p eponde ance o he land lo e y da a-
se lacking hese su eyo a e ac s. Bias in hese ea ly
su eys has been ex ensi ely s udied (see Cogbill (2023)
o a e iew) and can ake o he sub le o ms. Howe e ,
su eyo biases a e mos p oblema ic when using ela i e-
ly small sample sizes o when esea che s a e a emp ing
o es ima e ee densi y (Cogbill e al. 2018), nei he o
which a e ue o ou wo k.
Soils da a we e p oblema ic in se e al ways: he o igi-
nal o ma (Mic oso Access da abase o s andalone a-
bles ha had o be linked o map uni s) was di icul o use,
he e we e la ge amoun s o missing da a and da a quali y
was some imes ques ionable. The Na ional Coope a i e
Soil Su ey (NRCS NCSS 2024) ha collec s soils da a in
he Uni ed S a es elies on olun ee s om di e en agen-
cies o collec soils da a, allows o di e en collec ion and
es ing me hods and has been collec ing da a o e a long
ime span and wi h di e en echnologies (USDA 2017).
This may explain why only h ee o he 12 soils a iables
we used (pH, os - ee days and % clay) had a signi ican
in luence on he species dis ibu ion models and may
ha e esul ed in some o he complex ches nu esponses
ha we e di icul o in e p e . Ne e heless, ou s udy and
o he s p o ide e idence ha hese en i onmen al ac o s
a e in luen ial o ches nu dis ibu ion and, al hough we
could no include pH, os - ee days and pe cen age clay
in ou p edic i e maps, hey in luenced he ull species dis-
ibu ion models and should be gi en conside a ion when
p io i ising ches nu e-in oduc ion si es.
Despi e he challenges, ou s udy e ealed some e-
sul s o pa icula in e es , no ably, he high abundance o
ches nu in he Geo gia Piedmon , 11% highe han in he
Blue Ridge and o e 200% highe han in he Ridge and
Valley and Appalachian Pla eau, p o inces whe e high
ches nu abundance was be e documen ed and is wide-
ly accep ed. Ou indings con adic anecdo al accoun s
and many ange maps which exclude mos o he Geo gia
Piedmon om he ches nu ange (Li le 1977); i.e. p io
o ches nu bligh and PRR ches nu was widesp ead in
he Geo gia Piedmon and abundan h oughou much
o his physiog aphic p o ince a le els compa able o o
highe han any o he p o ince in Geo gia. C andall e al.
(1945), Woods (1953), Anagnos akis (2001) and o he s
ha e specula ed ha ches nu was i s los in he Geo -
gia Piedmon because o PRR which may ha e h i ed in
he wa me clima e, lowe slopes and hea ie clay soils,
howe e , he loss was poo ly documen ed. PRR ypically
kills ches nu s on poo ly d ained o dis u bed si es, lea -
ing li le e idence o hei o me occupancy, whe eas
ches nu bligh mo e o en op-kills ches nu s, lea ing a
legacy o ches nu sp ou s o ouch o hei o me p es-
ence (Wang e al. 2013). The dis ibu ion o ches nu in he
Coas al Plain is also no able whe e o he maps exclude
i en i ely. Al hough i was appa en ly es ic ed o idges,
hill ops and s eep i e blu s, ches nu was p esen and a
signi ican componen o some Coas al Plain ecosys ems.
In ou s udy, ele a ion was he only a iable o which
ches nu esponded consis en ly ac oss e e y p o ince in
