Over-the-horizon extinction risk assessment reveals rapidly shifting geographic and taxonomic priorities for conservation
preprint
OA: closed
CC-BY-NC-ND-4.0
AI-generated deep summary
The paper develops an “over-the-horizon” forecasting framework that integrates ensemble species distribution modeling with machine-learning automated threat assessment, using projected changes in climate-driven range dynamics, invasive species distributions, land use, and human population density to predict future IUCN Red List threat status for 1,914 Australian terrestrial vertebrate species to 2100. Under a high-emissions scenario, up to 109 species are projected to lose all climatically accessible habitat by 2100, and the number of threatened species increases, while a moderate emissions scenario (SSP1.26) shows threatened-species numbers remaining relatively stable with up to 19 species losing all climatically accessible habitat. Spatially, threatened-species richness becomes increasingly concentrated in southeastern Australia, partly because extinctions and range contractions disproportionately occur outside protected areas. This paper does not explicitly discuss endometriosis or adenomyosis; it was included in the corpus via a keyword match in the upstream search index.
Read from the paper's body, not the abstract. Not a substitute for reading the paper. No clinical advice. How this works
Abstract
A central challenge in conservation is understanding how climate change interacts with other global change drivers to shape future species extinction risk, threatened species hotspots, and the effectiveness of protected areas. Here, we use an integrated over the horizon forecasting framework to jointly model changing species' range dynamics and shifts in extinction risk for 1,914 Australian terrestrial vertebrates to 2100. Our approach links ensemble species distribution models with machine learning-based automated threat assessment, incorporating species traits, changing distributions of invasive species, and projections of land use and human population density. Under a high emissions scenario, up to 109 species are projected to lose all climatically accessible habitat by 2100 and the number of threatened species is predicted to increase, while under a moderate emissions scenario (SSP1.26) the number of threatened species remains relatively stable, and up to 19 lose all climatically accessible habitat. Spatially, threatened species richness becomes increasingly concentrated in southeastern Australia. These shifts elevate the representation of threatened species within existing protected areas, largely because extinctions and range contractions occur disproportionately outside protected areas. Our results highlight that the identity of at risk species and the occurrence of threatened species hotspots will change dramatically, underscoring the need for forward looking conservation strategies that anticipate future biodiversity patterns.
Full text
103,873 characters
· extracted from
oa-pdf
· 6 sections
· click to expand
Keywords
automat ed a sse ssmen t, bi od iversi ty, c limate c hang e, ex t i ncti on ri sk, ma chine learni ng 19
20
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
INTR ODUCTION 21
Biodi ver s i t y lo s s in th e Anth r o poce ne i s dr i ven by multiple int erac ting thre at s 1,2 . Y et most f orec a st s 22
of fu tur e ex t i ncti on ri sk foc u s on cli mat e- driven shi ft s in sp ecie s di s trib ution s . A s w ell a s c lima te , 23
ex t i nction ri sk a l s o chang e s in r e s p on s e to c hang es in la nd use, human p o pulati on de ns i ty , and ot her 24
thre ate ning proce s s e s 3,4 , a nd th e ef f ec ts of th e se proc e s s e s ar e medi at ed by s p ec i es biol og ica l 25
trai ts 5 . Alth oug h recen t d evelop m ent s ha ve a dvanc ed the m od elling of speci e s re s pons e s to c lima te 26
ch ange, sub stan tial unce rtai nty r em ain s about how cli mat e-d riven ran ge shi ft s in terac t with oth er 27
glo bal-cha ng e driv er s a nd s p e cie s’ in trin s ic biologic al t rai ts to inf lue nce futu re t hr eat sta tu s 5,6 . 28
Addre s s ing thi s gap r equi re s an int egrat e d unde rst anding of how multip le th rea t s and ri sk fa ct or s 29
inter ac t t o d et ermine how s p ecie s ’ g eogr aphic range s an d extinc tion ri sk ar e li kely to cha nge 30
through tim e. 31
One o f the ma jor cha llenge s c r e at ed by chan ging pa ttern s o f ex tinctio n ri sk i s th at c onse r v ation 32
prioriti e s— bo th spec ie s -level and s pati al —are unlik el y to remai n stabl e . For ex am ple , pro t e ct ed 33
area s ( PA s ) ar e a c ent ral pil lar o f co n s e rva t i on, wi t h loc atio n s w ith high conc en tra t i on s of thre a ten ed 34
or ende mic spec ie s p riori t i s ed for th e e s t abl is hm e nt o f new P A s 7,8 . How ever, g iven c limate -driv en 35
dist r i buti onal c hange a nd s hif ting thr eat s , the id en tit y o f thr ea tene d sp ec ie s and t he locatio n of 36
futu re biodi ve r s ity hot s po t s may chan ge sub s t an t i a lly t hroug h tim e 9 . Pa s t s tudi e s h ave rea che d 37
mix ed c onclusion s abou t w heth er P A ne t wo r k s w ill l ose or g ain e ffe ctiven e ss und er c limate c hang e. 38
Earli er re sea rc h con cluded th at PA s w oul d de cline in e ff ec tivene s s b eca u s e t h ey los e more s pec i es 39
than th ey gai n under c limat e -drive n ra nge shi ft s 9,10 . H ow ever, mo re r e c ent w or k h as r epor t e d that 40
P A ef fe ct i v e nes s has r e c e nt l y i n cr e a s e d 11–13 or is pre dicted t o incre a se und e r mod elled f ut u re 41
scen a r io s 14–18 . T he s e mixe d finding s hi ghl igh t tha t con s ervati on d eci sion s ba s ed o n p r e sen t-d ay 42
patt ern s r i sk being mi s a ligned with futu r e biodi ver s i t y ne ed s—a ch all eng e th at ca n only be 43
addre s sed by forec as t i ng th e joi nt dyna mi cs of s p eci e s dis tribu tion s , t h r e a te ning proc e s s e s, and 44
ex t i nction ri sk. 45
Here we pr e s en t a n int egrat ed appr oach t o fo r e ca s ting fut ure ex ti nction ri sk th at c ombines 46
ens emble s pe ci e s dis tr i bu t i on mod ell ing w ith informa tion on s p ec ie s biolo gy a nd projected cha ng e s 47
in threa ten i ng proce sse s ( FigureSČqF1 ) . O u r a ppr oa ch u se s a ma chin e -l ear ning– ba se d automa ted 48
as s e ssme nt framew or k to cla s s ify spec ie s by thei r predi c ted I UC N Red L ist sta tu s to the end o f thi s 49
ce ntur y . Thi s au toma te d a s s e s s m en t met hod wa s devel o ped to predic t t h e thre at s t atu s o f 50
unc las s i fie d r eptil e spec ie s 6 bu t h a s not p r e viou sly be en app li ed to for ec a s ting fut ur e chan ge s in 51
thre at st a t u s. We u se ou r in t e gr at ed app roac h to inv e stig ate ch a nge s in the di stri bution a nd 52
ex t i nction ri sk in te r r es tr i a l Au st r ali an ve r teb rat e s, a global ly uni que and diver se c ontin en tal 53
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
as s e mbl age o f >2,20 0 s p ecie s. We qu anti fy tempor al chang e s in th e numb er an d propor tion o f 54
thre ate ned spec ie s ac r o s s t he four ve rte brat e c las se s and a sse s s h ow s pati al p at te r n s of 55
t h r ea t en e d-s p e c ies r i ch ness ar e e x p ec t ed t o s h ift . We t h e n ev a l u at e h o w ef fec t i v e ly t h e cu r r e nt PA 56
netwo r k i s pr e dic ted to r ep re sen t indi vid ual thr e at ened spec i e s and th rea ten ed- ri c hness ho t spo ts 57
under f utur e scen ari o s , con side ring c lima t e -driv en r ange s hi f t s and proje cted c ha nge s i n s pe ci es 58
thre at st a t u s simultan eo u s ly . Fina lly, we a nalys e c hange s in cro ss- taxon cong r ue n ce i n hot s pot –P A 59
ove r l ap to ev aluat e w heth er f uture prio ri t y area s ar e l ikely to bec om e more spa tia ll y c oncentra ted 60
or disp er s e d. 61
62
F i g u re 1 . Wo rk f lo w fo r o u r o ver - t he - ho ri zo n f ra me w or k to p re di c t e x t i n c t ion ri s k . C urat e d occurre nce 63
r e c o r d s a r e co up le d w ith e nv i ron m e nt a l la y er s ( bi o cl im a t e , so i l , e t c . ) to t ra in s p e ci e s d is t r ibu ti on m o de l s 64
(SD M s ) , b ot h for a sse ss ed n a ti v e s peci es and in v a si ve s pecies. T hes e are use d t o gen erate fea t ures for 65
m a ch i ne le a rn ing -ba s ed au to m a te d as s es sme n t m ode l s , c on s i st ing o f sp e ci e s t ra it s ( ma ss a n d r a ng e s i z e , th e 66
l a tte r p r oj e ct e d f r om the f i tte d S DM s) and ov e r l a p w ith th r ea t eni ng p ro ce sse s (S D M p r o j ec t i on s o f in t r odu ce d 67
speci es ran ges a n d s o urc ed l aye rs o f h u m a n-pop ulati o n d ens i t y an d m o d ifi e d h ab i ta t). Th ese f e a tures , c oupl ed 68
wit h c urrent I U CN thre a t as s e s s men ts, are us ed t o trai n a n a ut omat ed a sse ss me n t m o de l u s in g nest ed cross -69
vali d at io n . The n, f ut ure p r oj ecti o n s o f cli mat e lay e rs u nd e r differen t So c ioeco n omic Pat hw ays a r e us ed t o 70
pro j ec t t he SDM s an d g enera t e f ut ure ra nge m a ps for t he nati ve s peci e s an d i n v as i v e sp ecie s un de r diff erent 71
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
dis p e rs a l sc e n ari o s . Fut ure project io ns of ra n g e siz e an d o ve r la p w it h thre a t e n ing process es are use d t o make 72
ne w p re dicti ons of f u ture IU CN thre a t as ses s me nts u s i ng t h e aut om a t ed a s ses sm e n t m o d el. 73
74
Results
75
Predicted changes in species threat status: taxonomic patterns 76
We impl emen ted a novel , int egrat ed w o rkflow f o r fo rec a sting fut ur e cha nge s in s peci e s t hr ea t 77
sta t u s, by comb ining i ndep ende n t l y proj ec t e d cha nge s in g e ographic rang e s ize o f na t i ve a nd 78
inv as i ve spec ie s (a s a th r e a tening proc e s s ) from S D M s ba sed o n 5 .6 mill ion oc curre nce r ec o rd s , with 79
projec t e d cha nge s in hum an popula t i on den sity a nd land u s e , and bio logic al an d geog r a phic 80
predic to rs of ex t i nction ri sk. We appli ed an XGBoo st automa ted a s s e s sm ent mod e l 6 on thi s 81
integ r a ted da ta s et f or 1,914 Au s tr a lia n t erre s t rial ve rteb ra te speci e s, achie ving high c las s i fica tion 82
ac curac y (0.91 4) when di stin gui shing thr eat ene d ( VU, E N / CR ) ver su s non -threa t e ned sp eci e s ( NT, L C) 83
and N ea r Threa ten ed v s. Lea st C onc ern spe cie s (0.93 7 ), bu t lowe r acc uracy (0 .604 ) w hen 84
d i s ti n g u i s hi n g b e t w ee n t he T h r e at e ne d c a t e go r ies ( Ta b le S1 ) . T he m o s t i mp o r ta nt pr e di c t or s o f 85
cu r ren t thr eat s t a tu s w er e geo grap hic r a nge size and b ody ma ss , foll ow ed by hab itat -r elat ed 86
fea tur es and ov erla p w ith inva s i ve sp eci e s ( F igu re S 1, Tab le s S2 –S 3) . 87
Under th e more p e ssimis t i c S SP5 .85 emi s s ion s scena r i o, t he numbe r of s pec i e s in ea ch of t h e f our 88
ve r teb rat e c la sse s (ma mmal s, bir ds, r e p ti les , amphibi an s) t ha t ar e cl assi fied in a T hr e a tene d ( VU, 89
EN/C R) ca teg or y i s predic ted to increa s e throughou t th e centu ry (F igure 2A , T able S 4). Thi s i s al s o 90
the c a se und er th e op t im i stic SSP 1.26 sc ena r i o when n o disp er s a l i s allow ed, but w ith limited or 91
limi t l e ss dis per s a l , the num b er o f t h rea te ned speci e s i s pr edi ct ed to r emai n rel ati ve ly stabl e. Ou r 92
model s sugg e st tha t s tepwi se p rogre s s i o n through th e Re d Li st lev el s i s common: t he highe r a 93
speci e s ’ c urre nt threa t e n ed lev el i s, t he mo r e likel y it is to be t hre a ten ed ( or go e x tinct) in t h e 94
sub s e quen t time ste p (Figu re 2B, Tab l e S5). H ow eve r, the cha n ging tax onomic pat tern s o f extinc t i on 95
risk ar e a l s o dr i ven by s pe cie s mov ing thr ough t he leve l s o f t h e Red L i s t i n variou s direc tion s, i.e. , 96
speci e s mo ving f rom u nt h rea t e n ed to t h reat ened , t hr ea t e ne d to un thre ate ned, a nd from 97
t h r ea t en e d o r unt h re at en e d t o e xt in c t ( Fi gu r e S 2 ) . C ha n g es i n ra n ge s i ze l ed t o s pe ci es p r ed i ct e d t o 98
bec ome le ss thre a ten ed, on ave rag e (Fig ur e S3) . How ever, c hang e s in most oth er predi cto rs ten de d 99
to c au s e s pe cie s to bec ome mo re t h rea t ene d, pa rticu la rly tho se p redic t ors wi t h high im por tan ce 100
(Fig ure S1 ). P at t e r n s di f fer ed be t w een t a x a: for ex ample , overlap with rabbit s wa s a s t ronger d rive r 101
of hig her thr e at pr edictio ns for rep tile s t han othe r c la s s e s, w her ea s overl a p with c ats wa s a st r on ger 102
drive r o f high er th rea t predic tion s fo r a m phibia n s than o the r cla ss e s ( F igur e S 3) . 103
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
104
F i g u re 2 . F u t u re t h re a t le vel s o f Au s t ra li a n te rre s t ri al v e r te b ra te s. A) Pre d ic ted n um be r s o f s p ecies in 105
diff e ren t threat ene d cat e g ories ( V ul ne ra ble, En dan ge r ed/Critical l y End ang e r ed) un der tw o d iff erent S h are d 106
So c ioeco n omic Pat h w ays (SSP 1 . 2 6, SSP5.8 5) a n d t h ree dis p ersa l s ce na ri os (N one, Li mite d, L im i tle ss) in 2 0-yea r 107
tim e s t e p s fr om 20 20-2 1 0 0. E x ti n c t s pe c i es ar e c olo ure d bl a c k a n d inc l u d e s p eci es th a t w ent ext inct i n pre v i o us 108
tim e s t e p s. B) P re dicte d pr ob abilit y o f assi g nme nt t o ea c h IUC N t hre a t c ate gory a s a func t i o n o f t he t h r e at 109
cate g or y in t he pre vio u s t i m est ep. M ean predicte d p r o babil ities an d 95% C I are es ti mate d fr o m predicti on s 110
across all S h are d S ocio e c ono mi c Pat hw a y s a n d dis p e rs a l sc enarios .111
The model s al so pred ic t an in cre a se in th e number o f ex t i ncti on s (Fig ure 2A), whi c h are inf erred 112
whe n a sp ec ie s’ modell e d cl imatic suit ab i lity from on e t im e pe riod t o the n ext re s ults i n n o 113
cl imatic ally s ui tab le ac c es s i bl e ar ea s . Un der S SP1 .26 , our mod el s predic t 2–7 spe c ies o f amphib ian s, 114
2–4 s p ec ie s of bi rd s, 6–7 s pec i es o f mam ma ls and 9– 17 sp ecie s of rep til e s c ould g o extin ct by the 115
end of t h e ce n t ury. U n d er SS P5. 85 all gro ups ar e pr edic t e d to exp er i ence a sha rp i nc r ea s e in 116
ex t i nction s la te i n th e ce ntu r y , in the p eri od 2080– 2100 (F igure 1A ), w ith 11 –12 spe cie s of 117
amph ibian s, 17– 19 speci e s o f b ir d s , 22 –2 3 speci e s o f mammal s, a nd 45–5 5 s p e cie s o f rep tile s 118
poten tial ly ex tinct by th e end o f t h e cent ury. U nde r b ot h s c ena rio s, all bu t one o f the se pr edi ct ed 119
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
7
ex t i nction s ar e s pec ie s en demic to Au str alia , in a ddi tion to the l o ss o f the sh rill w his t le frog 120
( Au s t roc ha perina gr a cilip e s ) from the A u st ralian mainl a nd ( the speci e s is a l so p r e sen t in N ew 121
Guinea , whic h is no t inc luded i n our mod els ). 122
The s pa t i al di st r i buti on of p redic t e d exti nc tion s r e flec ts the pa tt ern s o f r a nge c o ntr a cti on. A l t hou gh 123
ex t i nction s ar e pre dic ted in almo s t every bi ome by 2100, t he biome s o f s outh ern and so uthe a st ern 124
Aus t ral i a are e xpec t e d to ca rry a dispr op or tion ately hi gh shar e of sp eci e s lo sse s ( Figure 3A, Fig ure 125
S4). U nd e r SSP5 . 85, th re e ex t i nction h o ts pot s emerg e (F igur e S4 ) : t he dry mon soo nal G ul f Coun tr y of 126
north ern Au strali a , th e Vic to r ia n an d Mu rray M a lle e, and Ta s ma nia , w hich is p red i cted lo se 2 –27 127
ende mic sp ec ie s, d epend ing on the mod elle d di spe rsal s c enari o. 128
129
Figur e 3. Distri buti o n of Au s tr alian ter r e s tri al v e r t e b rat e sp e c i e s pr e dicte d t o g o ex t i nc t by th e end o f th e 130
centur y. A) Ba r pl ot s repres en ti ng f or s pe c ies pre d ic t e d t o s ur v i ve (l e ft) or go ext inct (ri g h t ) t he pr o port ion of 131
s p e c ie s cu r r e n tl y o cc u r r i ng in e a ch b iom e . E x t i n c t sp e c i e s a r e c on s i d e r e d a c r o ss a l l mod e lle d s ce n a r io s o f 132
clima t e c han ge (tw o SSPs a n d t hree dis pe rs a l sc e n ari o s ), t hu s re p r e s en ti n g th e m ost pess imis tic s ce n ari o. 133
Nu mbers nex t t o e ac h b i o me in t he ce n t re m a p repres ent t he t ot al n u m b er o f s pecies wit h th e maj orit y of 134
th ei r ex te nt dis tri but io n i n each bio me . B ) B o x pl ot s how in g t he dist rib uti o n s of curre nt ran ge s i z e (i n k m
2
; 135
log 10 -tr a n sfor med) of s pe c i e s pre d ic ted t o s u rv i ve or go e x ti n c t un der a ll m o de l led sc e n ari o s of climate ch a n g e . 136
Shifting species distributions under climate change 137
Shi fts in t he cen troid s o f s pecie s ’ g e ogra phic range s (F igur e S5 ) ar e driven by le ad ing-ed ge ex pan si on 138
of th e rang e bo unda r y on on e s id e , t r a ili ng-e dge c on tr a c tion o f the bo undary o n anoth er, or b oth . In 139
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
8
north ern Au strali a , ex panding ar ea s o f cl imatic s ui tab ili ty to th e so uth ea st mea n tha t on av era ge, 140
speci e s rang e s ar e ex pect ed to e xpand (F igure S 6). In s outh ern Au s t ral i a, hard l imi ts to r a ng e 141
ex pansio n at t he e dge s o f th e c ontin ent mean tha t on av erag e, spec ie s r ange s ar e e xpec t e d to 142
co nt rac t (Figu re S6). 143
The pr e dict ed r a nge c ha nge s dif fe r be t w een tax a an d vary over time. T he p at tern of nor thern 144
ex pansio n of r ang e s i s s tr on g es t e a rlier i n the cen tury (20 20–204 0 ), whil e th e pat tern o f sout hern 145
range cont ractio n b eco me s str onger a s t he c entu r y progre s s e s (Fig ure S6 ). Ra ng e c hange s a r e 146
relativ e ly e venly split b etwe e n cont rac t i ons a n d ex pan sion s in th e fir st h al f of t h e 21 st ce nt u r y , bu t 147
co nt rac tion s becom e mor e preva l ent a n d o f grea ter magn i t u de follow ing 20 60 (F ig ure S7). 148
Changing spatial patterns of threatened species richness 149
Curr ent thr e at ho t s pot s (top 10 th pe r c e nt ile of threa t e n ed specie s ric h ne ss) for t h e four c la ss e s are 150
loc ated p rimaril y in the re gion s o f high e st tot al s pec ie s r i chn e ss along Au s t ralia ’ s ea s te rn seab oa r d, 151
wi t h large ho tsp ot s for ma mmal s and r e ptil e s a l s o found in n o r the rn Au s tr a lia , a n d for mammal s in 152
sou thwe ste r n Au strali a (Figur e S 8A) . Und er all e mi ssion s and di s p e rsal sc ena rio s th ese hot s p ot s ar e 153
predic te d to shi ft tow ard s t h e south , t o become conc en tra ted in th e s outh ea s t er n ex t remi t y o f the 154
Aus t ral i an mai nland (a mphibian s , ma mmals & r e p t il e s), and t he ea ste rn/ s ou the a stern co a stal s t rip 155
(bird s), by the en d of t he c en tury (Fig ure S8B– D). T hi s gene ral s hif t pa tte rn ac ro ss the f our c la sse s i s 156
refl ec t e d in the h o tspo ts f o r c ombin a t io ns o f o ne, two, t h ree a nd fou r c la sse s, w hi ch a r e pr e dicted 157
to c ont r ac t f r om nor t he r n A u st ralia and t he c entra l- ea st co as t regio n t ow ard s t he s ou the a ste rn 158
e xt r e m it y of A us tr a l ia , pa r t i c ula r l y u n de r S S P 5. 8 5 ( F i gu r e 4 A) . 159
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
9
160
Figur e 4. C han ges in c o n gr u en c e o f thr e a t h otsp ots. A ) M a ps r ep re s e nt ing ho tspo ts o f t hr e a te n ed sp e c i e s 161
richnes s f or the f o u r c l as s e s of t errestrial ve r t ebrates . C ol ours r epres e nt d eg r ee o f c on gr uence, i.e. , ho w man y 162
c l a s s e s h a v e h ot sp ot s in tha t c el l . L e ft p ane l sho w s cu rr e nt ho tspo t s b a sed on a ut om a t ed a s se ss me n t o f 1 ,9 1 4 163
speci es, a nd ri gh t p anels s ho w pro j ect ed hot s pots a t 21 0 0 u nd e r t w o diffe r ent Sh ared S o c i oec on o m ic 164
Pa t hw a y s (c ol u m ns) an d t hre e differe nt dis p ersal s cenario s (r ows). B ) Box pl ots s h owi ng t he d egree of 165
c o ng r u en c e i n th r e a ten e d spe c i e s r i c h n e s s und e r two d i f f e re nt S ha r e d S o ci o e c ono m i c P a th wa y s ( us in g the 166
Limit e d di s p ers a l scen a r i o) over 20-y e ar ti m e ste ps ( 202 0 r e pre se n t in g pre se nt co nditi o n s ) , c a lcul ated usi ng 167
spa ti a ll y c orrect e d Pearso n’ s correlati o n c oef ficient s. Co lo u re d l i nes represe n t tren ds i n di ffere n t pairs of taxa 168
( e.g. , t h e oran ge li ne s h ows i ncrea s in g c ongr ue n c e be tw een bi r d s a nd re pti l es). 169
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
10
Some more idio s yn cra tic p at t e rns o f ho t spo t shi ft s are a l so e vident . I n bir ds and mamma ls, t he 170
co ncentr ation o f th rea ten e d s p ecie s on the isl and of T a s ma ni a in t e n s i fi es to ho ts pot lev el und er 171
SS P1.26 (F igu re S 8B –D ). Thi s hot s p o t int e ns i fi cati on i s no t see n under SS P 5.85, h o w eve r , due t o a 172
predic te d large wav e o f ex t i ncti on s of e n de mic T as ma nian thr eat ene d s p e cie s le a ving most ly non-173
thre ate ned spec ie s remainin g. H ot s po t a rea s f or ma mmal s and rep tile s i n no rt h e rn A u strali a are 174
ex pected to contr act a nd di sap pea r a s sp ec ie s expa nd th eir di st r ib ut i o ns sou thwa r ds an d r e duc e 175
their thre at sta t u s (Fi gu re S8B -D ). In amp hibia n s a nd, t o a limited ex tent , in ma mmals, the W et 176
Tropic s in th e f ar nor th -ea s t of Au s tr alia emerge a s a t hr e a ten ed spec ie s ho tspo t (Fi gure S8B – D). 177
Bec au se cl a sse s a re pr edic te d to r esp ond diff eren tly to clima te chan ge, the se c ha nges le ad to an 178
inc r e a s e in spa tial congru en ce o f thr eat e ne d richne s s ho tspot s o f mamma l s and r eptile s w ith bi rd s , 179
and a reduc t io n in spati al congruen c e be t w een a mphibi an s and mammal s (F igur e 4B; T able S6) . 180
The area o f th rea ten e d s pe cie s ric hn e ss hot spot s, a s a perc entag e of Au s t ral i a’ s l and a r e a, i s 181
predic te d to d ec r e a se by 210 0 for s i ngle - c las s ho ts pot s ( fr om 9.1 3% to 3.4 3–7 .42 % , summed ac ro ss 182
all four cla ss es and dep e nding o n SSP an d di s pe r sal scen a r i o) , and f or hot sp ot s sh ared by two c la s s e s 183
(4.42% to 1.24 – 2.94% ) ( Figure 4 A). H ow e ver, a s thr eat ene d s p e cie s rich ne ss in ge neral becom e s 184
more c once n t ra ted in s o uthe a stern Au st r a lia, h o tspot s s ha red by thr ee c la s s e s (1. 06% to 0.67 –185
1.58 %) and all f our c la s s e s (0 .21 % to 0–2 . 95%) a r e p r e dict ed to incr ea se in s ize un de r mo s t 186
scen a r io s. 187
Overlap between threatened species and Protected Areas 188
Threat en ed sp ecie s ric hne ss ho t spot s a re pr e dict ed to con trac t an d bec om e con centr ate d in 189
sou thea s t e r n A u st ralia, e s pe cial ly ma inla nd upland a re a s and th e isl an d o f Tasma nia , are a s wh ich 190
hav e rela tiv ely h igh PA c overag e . Thi s le a ds to a predic ted inc re a s e in ove r l ap be t we en hot spo ts an d 191
cu r ren t prot ect ed are a s by 20 40, and throug hout th e c entu r y (Fig ur e 5 A) . For spe cie s p redi ct ed to 192
be thr eat en ed , we predic t in crea s es in th e mea n ove r la p of speci e s rang e s with P As, b ut no i ncr ea se 193
for pr edic te d non -th r e a ten ed sp ec ie s (Fi gure 5 B). There a re al so p redict ed incre a se s in t he 194
propor tion o f thr ea tene d s p ec ie s th at ov e r la p s u b s t anti al ly w ith the cu r r ent PA n e t w ork, wh ile for 195
non- thre at ened s pec i e s the ov e r l ap inc re a s e i s only pr e dic ted un d er a ze ro - d i s pe rsal mo d el (Fi gur e 196
S9). U nd e r ze ro -di sper s a l model s, th e nu m ber of s p e cie s predic ted to re ac h 50% overlap o f thei r 197
range s w ith P As i s g r e at er th an the num b er of s p e cie s wh o s e level o f overl a p will decrea s e t o below 198
50% (Fig ur e 6, uppe r p a nel s ). Howeve r , b eca u s e t h e no -di sper sal model do e s n o t a llow r a nge 199
ex pansio n, thi s e s sentia lly m ean s tha t ra nge c ontrac tion s ar e predi c ted t o be more exten siv e ou ts i de 200
PA s tha n wi t h in th em. Th is p at t e r n i s mo re variabl e and i ncon si sten t und er th e model s tha t a llow 201
limi t e d or limitl e s s disp ersal (Figure 6 , mi ddle and lo wer p ane l s ) , si nc e unde r the s e scen a r i o s r ang e s 202
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
11
ca n also ex pand o ut o f P A s , red uc ing s p e c ies’ co ve rage within P As, or c urren tly un-re pre se nt e d 203
speci e s c an e xpand t h e ir r a ng es in to P A s, inc r ea s in g thei r co verage . For amph ibi a ns and rep t il e s , 204
more s pec ie s ar e ex pect ed to l o se t h e 50% level of overla p with P As th an g ain th e m, r e ducing the 205
ove r a ll pro tecti on fo r th es e tw o t axa. Co nvers ely, many mor e bird speci e s a r e expec ted to re ac h the 206
50% overla p lev el than l o se it . 207
208
F i g u re 5 . De g ree of ov er l ap wi th p ro te c t ed ar ea s . A) Box p lots s how in g t he degree of over l ap betw ee n 209
thre a te n e d s pecie s hots p ots a n d pr otect ed a rea s u n de r a l l c on si de r ed cli m a te c han ge sc enari o s o v e r 2 0-y e a r 210
tim e s t e p s, calc ulate d as t he pr o port ion of ho tsp o t a r ea overlap pin g w it h a ny p r ot e c ted a r e a. T he horiz o n t a l 211
blue d as he d li nes re prese nt c u r rent lev el s of o v erlap wit h p r o t ect ed a reas b a s ed on aut o ma t e d as se ss ment of 212
1,9 1 4 s p eci es. B) M ean l evels of o v erlap wi th PAs pre di c te d t hro u gh time, f a c e t ed by diffe re nt S o c ioeco no m ic 213
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
12
Pa t hw a y s a n d dis persa l s ce narios. Soli d li n e s repre s en t pre d ict io ns f or no n-t hreate ned sp ecie s , a nd das hed 214
lines re pre s ent pre d ict i o ns f or t h r e a te n e d sp e c i es. 215
216
Figur e 6. C han ges in l e vels o f pr o t ec ti o n f or thre a t en e d sp ec i es. Ba r plo ts s how i n g t h e p r oport io n of 217
thre a te n e d s pecie s in ea c h terrestria l ve r t ebrate c la ss t h a t ei t her dr o p be l ow t he 50 % ove rla p wit h P A level (i n 218
red) or gai n t hat pr o t e c ti on level (i n gold). R e sul ts are f acete d b y diffe r ent Soc ioeco n o m ic Pa t hw a y s a n d 219
d i spe r sa l s ce na r i os. 220
221
Discussion
222
Metho ds for fore ca sting spec i e s extinc t i on ris k over d e cada l time scale s a r e beco m ing inc reasingly 223
importan t a s c on s e r v a tion shif ts to b e ing more proac tive 4,6,19,20 . Robust extinc t i on -r i sk fo rec a s t in g 224
requir e s method s that con sid er t he man y fa ctor s th at a r e like ly to in flu enc e a s pe cie s ’ c ha ng ing 225
thre at st a t u s: thi s inc lude s not only clim ate cha nge d r iv en di s tr i bu tion shi fts, bu t a ls o proje cted 226
ch anges in thre at suc h a s human p opul at ion, l and u se, a nd inva s iv e sp ecie s , and t he w ay that 227
speci e s ’ i ntrin sic biolog y medi ate s re spo nse s t o t h rea ts. W e ha ve pre s ent ed the fi r s t (to our 228
kno wledg e) int egra ted forec as t i ng meth odolo gy that c ombin e s the s e fac t o rs t o si multaneou sly 229
ex amine s hi ft s in speci e s d istribu tion a n d cha nges in t hrea t sta tu s . Our model s pr edic t pro found 230
ch anges in the con tine n t -w id e tax onomi c and spa tial pa tte r n s of ex tinction r isk in Australi a n 231
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
13
terre s t rial v erte bra t e s to th e end o f thi s cen t ury . The s e c hange s re sul t in thr ea te ned t e r re s tr ia l 232
ve r teb rat e s bec oming m or e conc e ntra te d within t h e curr ent n etw or k o f pro tect e d a r e a s . 233
Our model s indic a te a s ha rp ri se in extinc t i on s a f t e r 2 080 unde r the wor s t -c a s e cli ma te s c ena rio 234
(SS P5 .85) . Th is incr ea se i s drive n by sub s t a nti al w arming l ate r t hi s c en tury ( from 2.4SČqF° C i n 2041–235
2060 to 4.4SČqF° C in 2081–2 100 unde r S SP5 .8 5 21 ) a nd a ma r k ed dec lin e in rain fal l ac ros s s ou thern 236
A u st r a lia 22 . O v erall , proje ct ed ex tinc tion s by 2100 rang e from 19–3 5 s pe cie s und e r S SP1.2 6 237
(dep en ding on di s pe rs al a s s umption s ) to 95–10 9 s p eci e s unde r SS P5. 85 (Tabl e S7 ) . T his may be an 238
undere s timat e g iven the incr ea sing f r e qu ency and sev e r i t y o f ex t rem e , cli mate -lin ke d s t oc ha st i c 239
ev ent s such a s t h e Au s tr alia n 2019/202 0 Blac k Su mmer mega f i re s, w hich could r e s ult in t h e 240
ex t i r pa ti on o f thr eat ened sp ec ie s long b e fore cha nge s in mea n cl imatic condi tion s mak e area s 241
inhospi tabl e 23 . Furt h er m ore, many mor e extinc tion s ar e lik ely a cro ss th e fou r g r ou p s w e ana ly sed 242
bec au se th e 577 s p e cie s th at w e were u nable t o mod el ar e a non -rand om sampl e of Au s t ral i a’ s 243
ve r teb rat e s, wi t h s ma ller ge og r a phic r a n ge size s a nd high er curr ent thre at s t atu s, on a verage (Figu re 244
S10 ). Of c our s e , our pr edictio ns are a sso ci ated wi t h con sid erabl e unce rtain ty ( e. g . , mod elling 245
as s u mp t i on s , di s pe rsal c apa bili tie s, a da p t i ve c apac ity) in a d dition to tha t c apt ure d by t he range o f 246
scen a r io s w e model l ed. N one thel e ss , the s e r e s ul ts hav e two i mporta nt implic a tio ns. Fi r s t, the 247
predic te d su r g e in e xtinc tion s la te t h i s c e ntury unde rsco re s th e imp ortanc e o f a lo ng-te rm outlo ok 248
for con se rva t i on pl anni ng to mini mize bi odiv ersity lo ss. Se con d, t h e se re sul t s sug ges t t ha t doze n s o f 249
Aus t ral i an ve r t ebra te sp ec ie s coul d be sa ve d fr o m extinc ti on by global -sca l e ac tio n to av oid the 250
fo s s il- fu el led de vel opme n t s c enario imp li ed by SSP 5.85. 251
Our model s a s s ume that in ge n eral , s pec i es t ha t s u f fer the g r e a t e st lo s s of s ui t a bl e c limate a re a wil l 252
be m ost lik ely t o expe r i e nce a s e ve re c on tr a cti on in t he i r di s trib ution . The ex te nt t o whic h this i s 253
true for any give n s pe cie s w ill dep end on spec ie s- spec i fic f a ctor s such a s habi ta t s peci ficity, t he 254
av ailab ility of sui tabl e habit at in to whic h a spec i e s can migrat e, or c hangin g spati a l ov erlap w ith 255
i n t e ra ct in g sp e c ies s u ch as c o m p e t it or s, p r e d at o rs o r pa t h o ge ns ( e. g. , 24 ). Our foc us in t hi s st udy is 256
on a ggrega te t a xonomi c an d s pa t i al pa tt ern s of extinc tion ri sk, w hich a re like ly to be r e la tivel y 257
robu st to s uc h vari a t i on in sp eci e s - sp ec ifi c resp on se s t o clima t e cha nge . Non eth el e ss, o ur model s 258
also rev eal pa r tic ula r s pecie s o f conc er n , whic h are c onsi sten tly predict ed t o b eco me extinc t unde r 259
diff er ent c limat e-c h ange sc ena rio s (Tabl e S7). S om e are na r row - ran g ed ha bit at spe cial i s t s th at ar e 260
alrea dy th rea ten ed ( e. g. , a lpin e bog - s k in k [ P seu de m oia c r y odroma ], Howa r d Rive r toadl e t [ Upe rol eia 261
dav ie sa e ] o r nor ther n hairy- no sed w omb at [ L a s i orhi n u s k ref ft i i ]). O t her s a re no t c ur ren tly 262
recog nized a s thre ate ned ( e. g. , we s t ern f als e pip i s tr elle [ Fl as is t r e l lu s m ac k en z i ei ], Karri fr og 263
[ Ge ocrin ia ro s e a ] or s i lk y mo us e [ P s e u do mys apo de m oide s ]) , and henc e unlike ly t o be consi der ed a 264
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
14
hig h priority f or con serva tio n, ye t our m ode l s s ug ge st th ey cou ld bec ome ex tinct , sugge s ting t h e s e 265
speci e s c ould be prio ritized fo r re s earc h or early int er v en tion. 266
While t h e IUC N t hrea t c la s s ific a tion sche me all ows for fut u re t h rea t s to be i nclud e d in a ss e ssment s 267
under c rite r i o n D2 25 , t he r e i s no curr ent provision for h ow alrea dy ex i s tin g th r e a t ening pr oc e s s e s 268
mig ht s hi ft in ext ent an d magni t ude , and ho w this may chan ge a s se ssmen t outc o mes. O ur 269
integ r a ted fr a m ework repr e sen t s t h e fir st att empt to mode l pr ojected clim a te -dr i ven c hange s i n 270
speci e s di stribu tion s and p roje ct ed c han ges in s pec ie s thre at s t atu s simult aneou s ly and c an ide nt i f y 271
whi ch c hange s i n th r e a tening proc e sse s drive future pre dict ed thr ea t le vel s on a s pe ci e s-by - speci e s 272
basi s (Figu r e S3 ). Whil e t h i s i s no r e place ment fo r empiric al , ex per t- l ed a s s e ssme nts, whic h requir e 273
periodi c upda t i ng a s new thre a ts em er g e and s p ec ie s’ c ir c um st a nc es c h ange, our approac h pr ovide s 274
a robu st fir s t s tep f or over -t he -horizo n cons ervati on planni ng. 275
Early work f o cus ed on pr edi cte d c hange s in s pec ie s di s tr ib u t io n s i n re s p on se to c li mate chang e 276
sugge st ed tha t th e e f f e ctiven e ss o f P A n etw or k s m ay de crea s e unde r c limate cha ng e 9 . Howev er, ou r 277
re sult s sugge s t t hat the r epre s ent ation e ff ec tiven e ss o f th e P A ne t w o rk is pr edic te d to incre a s e for 278
thre ate ned spec ie s (ei ther cu rren tly or p r e dict ed to b ec ome t hr ea t e ne d), bu t not for unt hre ate ned 279
speci e s . T hi s can be ex plain ed by the alig nme nt o f se veral fa ct ors. F ir st, th e prev ai ling pressu re o f 280
cl imate c hange in Au s t ral i a is ex pected t o d r iv e shi ft s i n s pe ci e s dis tr i bu t i on s t ow ard s the s ou th a nd 281
ea st o f the c on tine nt, so t hat the seve re or c omple te lo s s o f spec i e s ’ suita bl e cl imate en vel ope s i s 282
ex pected to be mo st pre vale n t clos e to t he ha r d boun dary of t he s ou th -ea stern c o ast (F igur e s S4 & 283
S6). Sec on d, th e ea s t an d south -e a s te rn co astal r e gi on s are pa r t s o f Au s t rali a wit h h igh 284
co ncentr ation s o f nar row- range en demi c speci e s, whic h are mor e li kely to bec om e thre at ened tha n 285
the mor e b r oa dly di s t ribut ed speci e s o f i nland a nd nor the rn Au s t ralia , eve n if the y are curren tly not 286
thre ate ned. Thir d, south -e a s t ern Au s tr a li a ha s a high d en sit y of PA s compa re d to other are as o f t he 287
co nt i nen t, and many o f t he se a r e in upl a nd region s tha t ha ve t raditi ona lly be en le s s like ly to be 288
dev eloped f o r a gr icultur e. T hu s , a t ende nc y for rang e exp an s i on s a cro ss inl and, w es t ern, a nd 289
north ern Au strali a , co mbined w ith a ten de ncy f o r ra ng e c ont r ac tion s an d eme r g ence of n ew 290
thre ate ned spec ie s in the sou thea s t, lea ds t o the con trac tion and c once ntr ation o f t h rea t e n ed 291
speci e s ric hne s s ho ts pot s in th e P A- rich s outhe as t. Si mila r pat te rn s underlyin g ob se rved or pr edi cte d 292
inc r e a s e s i n P A rep re s e nt a t i on o f s pecie s ha ve be en de scrib ed in ot her c oun t rie s i n whic h area s of 293
co oler cli mat e are mo r e likel y to b e t he re cipie nts o f l ea ding -e dge ex pan sion o f s pec ie s r a nge s from 294
wa r me r are a s 11–13,17,26–28 . 295
Our int egrat ed f ramew ork pr edic ts w ide sprea d cha nge s in e xtinc tion r i s k acro s s Aus tr a li an t e rre s t ria l 296
ve r teb rat e s, inc luding an inc re a sed r epre s e n ta tion o f thr eat ened s p ec ie s within t h e curren t PA 297
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
15
netwo r k . Som e auth or s ha ve in te r p ret ed simila r finding s in a po sitive l ight , rep re se nting an incr ea se 298
in the e ff ectiv en e s s o f r e serve s und er c li mate -drive n r ange shif t s. Howeve r , in ou r re sults, the 299
inc r e a s e d r epr e s e n t a t i o n of thre at ened s pec ie s in PA s in t he sout hea s t is mir ror e d by sharp 300
inc r e a s e s i n extinc tion s in region s w ith fe wer PA s ac r o ss s o u t he r n and nor the rn A us t rali a (Fig ur e S4 ) 301
and r a nge c on tr a cti on s and shif t s into ar ea s with higher de nsi tie s o f P A s along Au s t ralia ’ s s ou the a st 302
(Fig ure s S 5 -S6) . T he s e re sult s sugge st th a t pat tern s o f extinc t io n ri s k will be hig hly dy namic , so tha t 303
the c urren t P A s ys tem mu st al so c on s id e r f u t u re ri sk. Au stra lian governme nt p olic y is to expa nd t he 304
co verage of the co un t ry ’ s P A sy st em fr o m the c ur ren t ~ 24 .5% t o 30% of t h e land a r e a by 2030. T hi s 305
ex pansio n sh ould c on s i de r dynam ic thre at pa tte rn s and stage d plannin g o f new P A es tabli shm ent in 306
area s t hat ma y not ye t be pri oriti e s but a re predic ted to bec ome p riori ty a r e a s. Thus, futu r e -pro of 307
co ns e rva tion p olic y c an be achi eved by c are fully incorpor ating pr i ncipl e s of ove r- the -h or i zon thre at 308
as s e ssme nt 4 . Mo re br oadly , our r e s ul ts h ig hlight the i mpor t a nce o f con se rvation a ction s not 309
co nnected t o t h e e sta bli s h men t of r e s e rv es , incl uding g lobal a c tion on clim at e ch a nge a nd pro activ e 310
loc al in t e r v e ntion s s uc h a s inv a s iv e pred ator c on tr ol o r cons ervat ion fe ncing to se cure speci es 29 . 311
312
Methods
313
Data Collec tion – S pecies 314
We dow nload ed globa l s h ap efil e s of the di s t ribut ion s of n on-ma rine mammal s a n d amphib ian s from 315
the IU CN R ed L ist 30 , of bi r d s from Bi rdLif e Int erna tional 31 , an d no n-ma rine reptil e s f rom the G lob al 316
Ass es s m e nt o f Rep tile Di str i b ution s ( GA R D) ini tiativ e v1. 7 32 . The se sha pe file s w er e then c r opp e d and 317
filt ered i n R v4. 5.0 33 t o o nly i nclude polyg ons t ha t ove rlap Au strali a in thei r nativ e range s, a nd , for 318
birds , in thei r re sid ent, b ree ding or non - breed ing r ange s . We r e moved 8 2 s pe ci e s of s eabi rds tha t 319
are no t k nown t o br e ed in ma inland Au s t ralia or Ta smania . Thu s, our da ta se t did not i nclude 320
inv as i ve, v agran t, or mi gra tor y s pe cie s , but on ly sp eci e s t ha t have e stabl i s he d nati ve population s in 321
Aus t ral i a (n = 2,156 ). 322
We re trieve d occ ur renc e da ta fo r s p ec ie s fr o m th e Atl as o f Liv ing Aus trali a (AL A 34 ) using t h e gal ah 323
pac kag e v2. 1.1 35 . We fil te red oc cur rence s ba s ed on the ALA d at a qual ity pro fil e (r emov ing dup lica t e 324
record s, r eco r d s wi t h s pa tia l or taxon omi c i s s ue s, fo ssil a nd ab se nce reco rd s, and po ssibl e o ut l i er s ) , 325
wi t h year of rec ord > 199 0 to en sure we i nclude only r e la tivel y rece n t r e co rd s wit h hig h s p ati al a nd 326
taxo nomic cert ainty. As an additio nal qu ality control , w e then p rune d o ut all occ u r renc e s t ha t fell 327
out side a 100 km buf fer applie d to eac h speci e s’ rang e p oly gon to remov e oc cu rre nce s tha t we r e far 328
out side kn own spec ie s ra nge s and a re t h e r e f o re likel y to be e r ron eo u s (n = 2,132 s pe ci e s wi t h 329
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
16
oc currence re cord s f o llowin g dat a c lean i ng). We repe a te d the sam e proc e ss o f do w nloadin g 330
oc currence d ata (but wi t h o ut f i lte r i ng wi t hi n buf fe red r ange polyg on s ) for fiv e inv as i ve spec i e s t ha t 331
are k nown to have la rge ne ga tive impac ts on Au s tr a li a’ s te rre strial v er t e b r a te f au na 36 , to be u s ed a s 332
predic to r v ariabl e s for thr eat a s se ssm ent : ca ne to ad s ( R h in el la m a r i na ), cat s ( Fe li s ca t u s ) , r ed f o x es 333
( Vulpe s v ul pe s), c am e ls ( Ca melu s dr ome d ariu s ), a nd Europe a n rabbi t s ( O ryc t o lag u s cun iculu s ). 334
We ac quir ed c urr e n t I UCN R ed L is t a sse s s me n ts fo r all s p e cie s from th e IU CN Re d Li s t u sing t he 335
taxi z e p ack age v 0.10.0 37,38 . We manuall y upda t e d the sta t u s o f seve n t ur tl e spec ie s ( C helo dina 336
long icol li s , C. st e ind ac hneri , El s ey a de nta ta, E m y d ur a ma c q uar i i , E. su bglo s sa , E. v ictoriae , and 337
Myuc hely s l ati s ter num ) th a t d o no t hav e listi ngs on the I UC N websi te bu t w ere a s s e sse d by t h e 338
Turtle T axonom y Working G r oup 39 . For s pec ie s lis t e d un der Au s t rali a’ s Environme nt Pr o tec tion and 339
Biodi ver s i t y C on ser v a tion Ac t 1999 (E PB C Act ), we opt ed to u s e thei r EPB C listin g s 340
( ht tp s : // w ww .environm en t.g ov.au / c gi -bi n/ s pra t/p ublic / pub lic thr eat ened li st.pl ) in l i e u o f t he i r g l o b al 341
I U CN as sess m en t , as t hes e a re b as e d o n t h e s am e s et o f c r it er i a a s t he IU C N R ed Li s t , b ut m o r e 342
ac curate ly refl ect th e s ta tu s of Au s tr alia n popula t i on s o f wider -r anging sp ecie s ( e.g. , th e c ur l ew 343
sandpi pe r, C alidri s f err u gi nea , i s g lobal l y list ed a s N ear Thr eat ene d [N T ], but n atio nal ly a s C ritic ally 344
Enda ngered [ CR] ). 345
Fi nally, we c ollec ted d ata on b ody s ize of Aus tr a li an te rre s t ria l ver teb rat e speci e s f rom p ubli sh ed 346
sou r c e s 40–56 for th e autom ate d a s s e s s me nt pr oc edu re ( s ee b e low) bec au s e s ize h as be en i den ti fied 347
as a k ey p red ictor of ex t i nc tion ri sk in ver tebr ate s 57 . W e u sed th e nat ur a l log a rit h m of mean adul t 348
body mas s (g ) for al l sp eci e s t o main tain c omparabi lity be tw een our f oc al tax a. 349
Data Collec tion – E nvironment 350
For the b ase line c ur r e n t c lima t i c c onditi o ns, whic h we u s e d t o model s pec ie s r ang es ( see sec tion: 351
Spe cie s Di s t ribu tion Mod elling – tr aining ), we used hi s t oric al c limat e da ta fo r 1970–2 000 from 352
World Clim v 2.1 58 . To e ncapsul a te t h e me an, variati on, and e xt r e me s in cl imatic va riation , we c hose 353
the f oll owing W orld Cl im va r ia bl es : BI O1 ( annual mean tem pe ratu r e ; °C), B IO 4 ( te mp er a tu r e 354
se as onali ty; SD * 10 0 ), BI O 1 0 (mea n tem p era tur e o f warme s t quar ter ; ° C) , BI O 1 1 (mean 355
temper atu re o f c olde st qu ar ter ; °C ) , B IO12 (annual pr ecipi t a t i on ; mm), BI O15 (pr ec ipita tion 356
se as onali ty; coef fic ie nt o f va r i ati on) , BIO 16 (prec ipi tati on of w e tte st qua rte r; mm), a nd B IO17 357
(preci pi ta tion o f drie s t quar ter ; mm). Th es e da ta were downlo a ded a t a 2.5 -minu te re s olution . 358
Animal s p e cie s di st r i butio ns are no t de te rmined sole ly by cli mate, bu t ma y also b e limited b y 359
phy s ic al la nd s c ap e fe atu r e s th at i n f l uenc e dis t ribu tion s o f pl ant s p e cie s and vege ta t i on typ es . We 360
ther ef ore inc lud ed top og r a phic and ge ol ogic al da t a in our S D M s ( se e sec tion: S p e cie s Dis tr i bu tion 361
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
17
Modell ing – train ing ). W e obtai ned a Dig ital El evatio n M o del (DEM ) at 2 .5-mi nu t e re s ol ut i o n f r om 362
World Clim v 2.1 58 , whi ch w as derived fro m the Shut tle Rad ar Topogra phy Mis s io n ( S RTM) glob al 363
ele vatio n da ta . We o btain e d th e foll ow ing soil at t ribu te s from th e S oil and Land sc a pe Gri d of 364
A u st r a lia 59,60 : AW C ( Av ail ab le W ate r Cap a city; % ), B DW (B ulk Den si ty – W ho le ear th; g/c m 3 ), CLY 365
(Clay con ten t; % ), pHc (pH - C a Cl2 ), SL T ( Si lt conte nt; % ) , SN D (Sand c on ten t; %), S O C ( or g anic 366
ca r b on; %) . We downlo a ded the se d ata a t a 3 sec ond re s o lu tion a t a dep th o f 0-5c m and then 367
ag gregated the m by a veragi ng ac r o s s 2.5 -minut e r e solu tion cell s t o ma t c h th e re s olution o f th e 368
cl imatic data . 369
For a nt h ropog enic condi t i on s , w hic h we u s e d t o model sp ec ie s t hr ea t stat us ( s e e s ec tion: Sp ecie s 370
Automa ted A s se ssm ent ), we downloa d e d bas eli ne pr ojec t i on s for 20 10 of h uma n populatio n d en sity 371
( H PD ) at 3 0 se c o nd r es o l ut ion 61,62 , a ggrega ted by averag ing ac ro ss 2.5 -mi nute re solu t io n cel l s , a nd 372
lan d use (LU) at 30-min u t e re s olu tion 63 . We fi l t e re d th e LU ra st er t o o nly i nclude cel ls c od ed a s 373
“c ropland_ bio ene rgy ”, “ croplan d_oth er” , a nd “bu ilt_up” and c on s i d er e d the s e a s human-modi f i ed 374
lan ds (no te t h a t t hi s do e s no t i nclud e th e impac ts o f exte nsive r a ngela n d grazing w hich cove r s a 375
large amoun t of l an d are a in Au st r al ia ). 376
For fu ture p rojec tio ns o f cli mat e ( s e e s e c tion: S p eci e s Di s tr i buti on Mod elli ng – p roje ction s) a nd 377
anthro poge nic c onditi on s ( se e s ectio n: Automa te d A sse s s men t) , w e c ons id e re d t wo Sh ared 378
Soc ioec onomic Pa thwa y s (SS P1. 26 and S SP5. 85) u nder the Had GEM3 -G C31-L L gl o bal c ircula t io n 379
model , whic h has the b e st fi t f or Au strali a’s clima t i c pa ttern s 64 , t o ca ptur e unc er t ainty in projectio ns 380
of fu tur e cl imate s, H P D , and L U . The S S P s r e pr e sent up p er and low e r extrem e s in m odelled 381
trajec torie s in clima te , demog ra phic s, ec onomy, and la nd u se, ba s ed on di ff er e n t clima t e cha nge 382
mitiga tion strat egi e s . The s e range fr om S usta inabili t y (low chall enge s to mitig ati o n and adapt ation ; 383
SS P1) t o Fo s sil-F uell ed Deve l opment (hig h chal lenge s to mitig a tion and l ow c hall e nges t o ada p t a t i on; 384
SS P5) . For e ach S SP w e obt ain ed p rojec t i o ns a t f o u r 2 0-ye ar in terv al s (2021–20 40 ; 2041 –2060; 385
2061 –2080; 2081–21 00) for H P D 61,62 , L U 63 , an d downsca l ed Cou p led Mod el In ter c ompari son P rojec t 386
Pha se 6 ( CMI P6 ) 65 c lima t e dat a fr om Wor ldClim v 2.1 58 . 387
Spec ies Dis tribution Modelling – tr a ining 388
We u sed th e fil t e r ed s pe ci es oc cu rrenc e data t o ge ner at e SDM s f o r e ach sp ec ie s i n the d a ta s e t w ith 389
at le a st 10 occ urr en c e rec ord s 66 (n = 1 ,928; Ta ble S 7) . To reduc e spa tial non -ind ep end ence o f 390
pre sence p oint s we pe rfo rmed spa tial thi nning by samplin g a sing l e point p e r g r i d cel l in the 391
env ironmen tal ra s ter re s olu tion (i f more than on e wa s pr e sen t) u sing th e `g r id Sa mple ` function i n 392
the di s m o p ack age v 1.3.5 67 . 393
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
18
Uneven s a mpling ef f o r t in pre s enc e poi n ts c ould l ea d t o bia s in m odelle d distribu tion s if no t 394
ac counted for w hen s elec ting ba ckg r ou n d point s 68,69 . T o get a n unbias ed m ea su r e acros s spac e of 395
how many s am ple s ar e repo rt ed fo r eac h s i te , we sele c ted ba ckg r ou nd p oint s whil e ac counting f or 396
sampling bia s by t r ea t i n g all occ urr en c e da t a in th at t ax on (amphi bi an s, bird s , ma mmal s , o r repti le s ) 397
as t he ba s el i ne bac kground f o r the ta xon. We pe rf o rmed s pa t i al thin ning 70 o n the bac kground point s 398
by r a ndomly s el e cting 10% of t h e r e co rd s fo r m ammal s, rep tile s , an d amphibi an s, and 1% of the 399
record s for bird s (due to th e muc h hig her number o f bird oc curr enc e s in ALA ). 400
F o r ea c h s p e ci es ( n = 1 ,9 2 8) , w e f it t e d e ns e m b le SD M s 71 usi ng the bi omo d2 pa cka ge v 4.2-6-2 72 . W e 401
fit ted five dif fe rent alg or i t hm s u sing th e `BIO M OD_ modeling ` func tion (Bo o sted R eg r e ssion Tre e s 402
[G BM], Gen era li sed Addi tive Mod el [ GA M], G e ne rali se d L inear Mod el [ G L M ] , Ma ximum Entropy 403
[MAXENT ], and R andom F ore st [RF ]) w ith t h e `bigbo ss` pr ede fin ed m odel pa ram e t e r s . We a ss e ssed 404
good ne ss- of - fit o f indiv idual mo del s via spatia lly e xplic it bloc k v alida t i on 73 using th e bloc kCV pa cka ge 405
v3. 1.5 74 . For ea ch spec ie s, we rand omly di vided t he s a mpling spac e in to h exag on al bl ocks, w hich 406
we r e di s tri bute d a mong k fold s. Bl oc k size (in m) wa s d ete r mi ne d a s the lo w er va l ue of ei t he r the 407
es timat e s s p a t i al autoc o rr el atio n range ( estima te d u s i ng the `c v_ spa tial_ autoco r` func t i on ) or th e 408
squa re ro ot o f th e are a of the speci e s’ ra n ge shap ef il e. The numb er o f fol ds k w a s then de termin ed 409
in a s t epw is e f a shion, st art i n g with 4 and a t tem p t i ng to ge n er a t e s pa t i al blo ck s us ing the ` c v_spa tia l` 410
func t i o n, and d ecr ea sing k by 1 until spa t ial block s wer e succ e ss fully gene rat ed . I f no bloc ks we re 411
succ e ss fully genera ted with k = 2 , SDM fit t ing for thi s speci e s w as s k ipped a nd it wa s ex clu ded in 412
dow ns t ream a n alys e s (n = 4 s pe ci es , al l r eptile s ). A f t e r i ndividu al mod el s w ere fit, an ense mble 413
model was ge n era ted by calc ulati ng a w eighted mean o f a ll mode l s, w eighi ng mode ls by th eir 414
ca lcul ated Tru e S kill Sta tis t ic (TS S; a m e a s u re of mod el a cc ur a cy calc ulat ed from t he s um o f 415
sen s i t i vity [ true p osi tive r a t e ] and speci f i city [tru e ne ga tive r a te ] ), a f t e r f i lt er i ng o ut indi vidua l 416
model s w it h nega t i ve v alu es of TS S, w hic h are indic a tive of pr edicti on s no bet t e r t han r a ndom 417
ch ance . Fiv e specie s (fo ur rep tile s , one bi rd) had no indivi du al mode l s with po sitiv e TSS value s. 418
Ensembl e model fi tting faile d f o r a n addi tion al four speci e s (on e amphibi an, one mammal, two 419
reptil e s) for unknow n r ea son s (e r ro r in r unni ng the `BI OMOD_mo de ling` func tion ). All n in e we r e 420
ex clude d fr o m down s tr e am an a lys e s . En sembl e mod el s wer e c onstruc ted f o r a t o t a l of 1,9 14 sp ecie s 421
(Tabl e S8) . 422
Spec ies Dis tribution Modelling – project ions 423
For e ach m odell ed s pec ie s , we use d th e ens emble mode l to ma ke pr edi ction s o f cl imatic s ui tab ility 424
ac r o ss Au s tr alia f or curr ent cond i t i on s a n d for futu r e condi ti on s i n eac h tim e st e p ( up to 2040 , 2060, 425
2080 , 2100) u nder t he two S S P s exa mine d. We conv erted p r e dict ed suita bi lity int o bi nary pre se nce -426
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
19
abs enc e pr edic t i on s u s ing a th re shol d va lue, w hich was selec ted for each spec ie s by maximi s ing its 427
TS S using the `bm_Fi ndOp timSta t` func ti on. Sum marie s o f SD M s fo r ea ch spec ie s , incl uding sampl e 428
size s, e valua t i on met rics and thr e shold v alue s, c an be fou nd in Ta bl e S 9. 429
We u sed th e bi nary pr edic t ed oc curre nce laye rs f rom the S DM s a s proxie s for pr e dic t e d exte nt o f 430
oc currence in futu re tim e st e p s. How ever , trea ting t h e enti re r ange abov e -t h e - thr esh old a s fu tur e 431
ex t e nt o f occ urrenc e do e s not t ak e int o a cco unt s pe cie s’ di f feri ng abilitie s to di sp er se t o track 432
shi fting c limatic cond i t i on s . In reali ty, cli ma tical ly “ s uit ab le” are a s t h a t are too fa r to re ac h wil l be 433
out side a f u t u re e xt ent o f oc curr ence . There for e, f or fu tur e predic t io n s o f p r e s enc e, we e st a b lis hed 434
an addition al pipeli n e to mak e sure t h at r a nge shif t s were po s s i bl e w ithin re a son able di s p ers al 435
limi t s, d efi n ed by seque n tially apply ing d ispe r sal bu ff er s to s ui t a bili ty l ayer s in ea c h time step . In 436
order to be c oun ted a s pr e senc e ce l ls in a futu re ex t ent o f occ urre nce m ap, c ell s that exc eeded t h e 437
pre sence t h res hold al s o ne eded t o be wi t hi n the limi ts of di sp er sal bu ff er app li ed to the p rev iou s 438
time ste p. 439
Due t o li mited da ta for di s pe r sal c apabi li tie s fo r mo s t o f Au st r a lia t err e s t rial vert e brat e sp ecie s , w e 440
brac ket ed unc er tain ty by consideri ng thre e dif fe ren t di sper sa l sce nario s to ge ne ra te bu ff er s . I n th e 441
fir s t (“ none” ) , t h e di sp er s al bu ff er wa s s e t to 0, di sall owing any r a nge ex pansi on s i n sub s e q uent 442
time ste ps, r e f l ec tive of sp ecie s tha t c ann ot tr ack range shif t s due t o li mited in trin sic di sper s al 443
ca pabili t i es o r h ard ge ograp hic ba rrie rs . I n the sec ond (“l imited” ) , we e stimat ed a fixe d dis per s al 444
limi t f o r a ll sp eci e s b a s e d o n av ailab le m amma l data . W e cal cula t e d t h e di spe r s a l buff er by 445
es timating maxima l di sper sa l rat e per ge nera tion f or Au strali an mamma l s f rom b ody ma ss, tro phic 446
lev el, a nd home r an ge size 75 usi ng publis hed pow er law s 76 . We t hen d ivided t h e max imal rate by 447
gen era tion leng t h (u s ing yea r a t fi rst bir t h as a sur r o gat e 75 ) t o a rr i ve at a p er ye a r dis p e rsal r a t e o f, 448
on a verage , ~7km 77 , or ~ 140km o ver a 2 0-year in t e r v al . We then round ed thi s valu e to 150k m and 449
appl ied it a s the d i s pe rsal bu ff er. S inc e t his value i s aver ag ed ove r sev e r a l mamma l sp eci e s (t he only 450
taxo n fo r whic h the req uired d ata we r e a vai lable), t hi s li kely repre s en t s an u nde re s tima t e o f 451
disp er s a l ca pabili t y fo r highly mobile ani ma ls ( e. g. , la r g e mamma ls, bat s , b i r d s), a nd an ov ere s t ima t e 452
for s p eci e s w ith low mobility ( e.g ., le gle s s l iza rds , fr ogs ). In the third di s p er sal sce na r i o (“unl imit ed”), 453
the di spe r sal buf f er s were d e s ig nat ed a s the e n t ire l andma s s in wh ich spec i e s are pre sen t, only 454
disall ow ing overwate r di sper sal . 455
The s a me S DM pr oce dure s w ere r epe ate d for th e f i ve in va sive spec i es , tre ati ng th e e nti r e pool o f 456
oc currence d ata for the inv a sive s pec ie s as th e bac kground p oi nt s , t o e s tima te cu rr e n t and fu tur e 457
range s ize s f or eac h o f th e inv a sive sp ec i es . We then c alcu la ted for eac h nativ e sp ecie s , a cro ss it s 458
cu r ren t an d pre dicted fut u re ext ent s o f o cc ur renc e, the ran ge size (km 2 ), p erc en ta ge of r a nge 459
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
20
ove r l ap wi t h cell s th at ha ve ov e r 50% human-modi fie d lan d, perc entag e of r ange overlap with c ell s 460
that hav e ove r 100/km2 hum an popul a ti on den sity, me an % o f human -modi fi ed l and ave r a ged 461
ac r o ss the e ntir e rang e , mea n human po pulation d e n s ity ave r a g ed ac ro ss th e ent i re rang e (cut -o ff 462
va lues for t he human pop ulati on d en s i ty and l and u se f ea ture s a s in 5 ), a nd p e r ce nt a ge of r a n ge 463
o v er la p w it h e a c h of t he f i v e i n v a s i ve s p ec i es . Th es e v a lu es w e re t h en u se d a s fea tu r es f or t h e 464
automa t e d a s s e s s m en t algori thm. 465
Automated As sessment of Th reat S t a t us 466
We u sed a machi n e lea r ni ng-ba s ed aut o ma t e d a s s e s s m en t meth od origi nally dev el oped to inf er 467
cu r ren t thr eat s t a tu s for Dat a De fic ien t ( DD) or Not Ev aluat ed (NE) r eptil e s p e cie s 6 , and extende d i t 468
to all ow it t o predi c t IU CN t h rea t c at egor ies (L C: Lea st Co nc ern ; NT: N ear Thr eat e ne d; V U : 469
Vulne rab le ; EN: Enda ng ered ; CR : Cr i t ic all y Enda ngered) und e r futu re c limat e c han g e scena rio s. Due 470
to a l ow sampl e s ize o f C R s p ec ie s in ou r da ta s e t (T abl e S 8), we combi n ed EN and CR into a sing le 471
c at e g o r y . O u r a ut o m a te d as s ess m ent us e s e X t r e me G ra d i e n t B oos ti n g ( X G B oos t 78 ), an e ffic ie nt and 472
hig hly a ccurate s upe rv is e d mach ine le a rni ng al gorithm 79 . The a pproa ch reli e s o n tr a ining a model t o 473
perf orm hier arc hical bina ry clas si ficati on t a s k s in a dec i sio n tre e: fir s t s epara ting t hr e a tene d ( VU, 474
EN/C R) f rom non -t h rea t e n ed (L C, N T ) s p ec ie s; th en, sep ara ting LC fr om NT sp ec ie s; then, se para ting 475
EN/C R from VU spec i es ( s e e det ailed flo wc hart in Fig ure 5 in 6 ). Each bina ry c las si ficatio n t a s k 476
inv olves hy perp arame te r tuning a nd mo del fit t i ng, u sing n e s t ed c ro ss -vali dati on t o te st pre dicti on 477
ac curac y (s ee bel ow). In a dditi on to spec i es omi tte d due t o fa ilu re t o fi t SDM s, we d id not tr ain th e 478
model on s peci e s w ho se pr edic ted r a nge s grea tly exc eede d r a nge shap e f i l e s (≥ 3 t imes la rge r or 479
smalle r; n = 226) . Thi s is t o r e duc e e rror s in cl assific a tion du e t o in fla ted ra ng e si zes , bec au se r ange 480
size i s on e of the k ey pre di ctor s for IUC N thr e a t leve l (ba sed o n c rit e r i on B ). Al tho ugh the se spec i es 481
we r e exc luded f rom mo del t raining , we us e d t he XGb oo s t mod e l to pr edic t the ir t h r e a t leve l s , a nd 482
the se we re u s ed in t he dow n s tr eam anal y s e s. I n t o tal , we train ed X GBoo st mo del s on 1631 spe cie s, 483
76% of Aus t r ali a’ s te r r e str i a l ve r t ebra te f au na (Tab l e S8). W e no te th at speci e s o mi t ted from t raini ng 484
for au toma ted a sse s s me n t ( n = 525) a re disprop ortiona lly t h rea ten e d and na rr ow -r a ng ed (F igur e 485
S10 ), so our a sse s s me nts li kely r e p r e s en t a conse rvativ e e s t i mat e of futu re th rea t lev els . 486
To p r ed ict th rea t st a t u s fo r ea ch speci e s , we inc luded sev en fe a t u re s f or eac h sp e cie s: ma ss (g), 487
range s ize (k m 2 ) , percen tag e o f ra nge overlap with c ell s t hat hav e ove r 5 0% huma n-modi fied land , 488
perce n tage o f ran ge ov e r l ap wi t h cell s t h at ha ve ov er 100 /km 2 huma n popula t i o n den sity, me an 489
perce n tage o f h uman -modifi e d land a ve r age d ac r o s s t h e enti re ra nge , mean hum an pop ulati on 490
den sity a verage d ac r o s s t h e enti re ra nge , and pe r c en tage o f ra nge o verlap w it h e ac h of th e fiv e 491
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
21
inv as i ve spec ie s. G i ven t he r e l atively sma ll numb er of fea tur e s , we di d not p er for m feat ure sel ec t i on 492
as in 6 , so a ll f ea ture s a re inc lude d in the mode l to pr edic t spec i es thre at . 493
In a dditi on to t he se fea tur e s, we al so co nsid ered whic h high er t a xon (c la s s) ea ch spec i e s belong s to , 494
and whic h biome eac h s pe cie s’ di s tribu ti on is pr edomina n t l y found i n . To a s s ign s pec ie s to biom es 495
we overla pped sp ecie s rang e s with sha p efil e s o f terr e st ri a l ec oreg ion s in Au stral i a 80 , and a ssigne d 496
eac h sp eci e s t o t he biom e i n whic h the h i ghes t pr op o r tio n of i ts di stribu tion li e s. We t h e n 497
parti tioned all sp ec ie s int o thre e biome c ateg or i e s , t o en s ur e e nough spec ie s in e a ch c ombination o f 498
biome , cla ss, and t h r ea t leve l . The t hr ee biome c atego rie s ar e: t ropic al (T ropic al a nd Subtropic al 499
Mois t Br o a dlea f F o re s t s ; Tropica l and Su btropica l Gr a sslan ds , Sa vanna s , a nd Shru bla nds), tem pe r a te 500
(Tempe ra t e Br o adl ea f a nd Mixe d For e st s; Te mpera te gra s s l an ds , Sav anna s, an d Shrublan d s ; 501
Montan e G r a s s la nd s and Sh rubl and s), an d dry (Med i t e rrane an For es t s , W o odl and s, and S cr ub; 502
De se r t s and Xeric S h r ub land s ). In te ractio ns amo ng f ea ture s , cl ass e s, and b iome s were ful ly al lowe d 503
in mo del traini ng , sinc e t hr ea ten ing proc e ss es in Au st r al ia va r y regiona lly a cro ss h a bita ts 81 and 504
cl asse s 82,83 . 505
We tr ained t he XGB oo s t mode ls u s ing th e x g bo os t p a cka ge v 1.7.10. 1 84 . We t un ed the fol low ing 506
hyp er p ara met er s : l e arning ra te, maxi mu m tr e e d epth, mi nimum c hild w eight, r o w s am pling, column 507
sampling , we ight b ala ncing , and the re gulari satio n pa r a m ete rs γ , α , and λ, by rand omly s a mpling 508
10,0 00 diff ere nt va lue s for e ac h parame t er. G ive n the size of ou r dat a set we a ppli ed ne sted cro ss-509
va lidation (rat her than s e tting a s i de a pr oporti on of the da ta se t a s vali dati on dat a and te s t dat a as in 510
6 ): the d ata s et w a s f i rs t s pl it in to N f old s; eac h fold w a s u s ed a s t e st d ata with the other N -1 fold s 511
use d as traini ng data to fi t mode l s . The s e N -1 fo l ds w e r e f u rt h e r s p l it i n t o n fol d s; ea ch o f t he s e n 512
fold s wa s us ed a s v alida t i on dat a w ith th e othe r n -1 fol d s u s ed a s trai nin g data to tune 513
hyp er p ara met er s . T hi s ne sted cro ss -val id ation e n s u red t hat traini ng, v alid ation , a nd te st da ta ha ve 514
no ov erlap. 515
We c r e a ted cu st om c ode to en sure tha t diff er ent fold s hav e roughly similar num be rs of sp ec ie s in 516
eac h c ombinatio n of biom e, c la s s , and t h reat l e vel. Wh en t h e number o f spec ie s i n a c ombina t i on 517
wa s small er t h an N , we set a s id e the se spec ie s to t he t e st da ta, so th at t he se und erre pre s ente d 518
speci e s w ill no t cau se mode l in s t a bili ty, a nd the p redi ction a ccu r a cy c an ref le ct ho w much 519
informa tion the ot her s pe ci e s ca r ry on th e thre at ening proce s s o f th ese unde rrep r esen ted spe cie s . 520
For the d ata se t to sepa ra t e t h rea te ned fr om non-t hre aten ed s pec ie s , we set N = n =5. The d ata se ts t o 521
sep ara te L C fr o m N T s pe cie s a nd t o s ep a r a te CR / E N a nd VU s pe cie s w ere much smal ler, so f or t h e m 522
we s e t N =5 an d n =3 (LC f rom NT) an d N = 4 and n =2 ( CR /EN from VU ). 523
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
22
At th e end o f t h e n e ste d cro s s -v alid ation , the mean s o f th e be s t fit hy pe rpar amet er s fo r e ach o f the 524
N fold s wer e us ed a s th e hyperp arame t e rs for t he fi nal model fit ting us i ng all the data . T he av erage 525
of th e predi c tion a ccu r a cy o f each o f th e N te s t d a t a wa s u sed a s the pr edi ction a ccurac y of t he final 526
model (Table S1 ). T o a s s e s s mod el s tabili ty , we r ep e ate d the w hol e proc edu re t hr ee time s, and all 527
replic at e s gav e simila r fi tt ed mode l s a nd predic tio n ac curac y. 528
We th en a sse ssed rel at ive fe a tu r e impor tance for pred iction i n the final model , ra nk ing fe atu r e s by 529
es timat ed g ain ( frac tiona l co n t ribu tion o f eac h fea ture to th e mod el ba s ed on t h e tot a l ga in o f thi s 530
fea tur e’ s sp lit s ) . T hi s r ep re sen t s the con tribution o f e ac h fe atur e to mod el predic t ions, repr e sen ting 531
its r e la tive import ance in a s s i gning s p ec i es to thre a t ca t e go rie s . Fin a lly, w e us ed t he fin al mod el to 532
predic t I U C N thr eat l e vel s. To set a ba s e li ne a gain st whic h to c ompa re fu tur e pred ic t e d th r e a t lev el s , 533
we us e the same me thod to pr e dic t c urrent th rea t l evel s f or all 1,914 spec ie s wit h mod elled 534
dist r i buti on s (1 ,631 spec i e s used t o t rai n the model a nd 28 3 s p e cie s tha t w ere eit her not a s s e s s ed 535
[NE ], a s s e s se d a s data d efici e nt [ D D ], or we r e dropp ed due t o havin g predic t e d ra nges muc h large r 536
than shap e f i l e s ; s ee a bove ) . This i s bec au s e cur ren t IU CN t h rea t l evel s ar e d e ter m i ned empi r i call y by 537
ex pert a sse s s o r s ba sed on a ddition al inf o r m ation (su ch as o bs erved ra te s o f po pul ation c hange ) s o 538
wou ld not be dire ctly comparab l e to mo del - b a s e d in fer ence s o f thr eat l evel. We made pred ic tion s 539
for futu r e th r e a t s ta tu s unde r al l exa min ed c limate chan ge and di s p ers al s c e nario s u s i ng the 540
ca lcul ated va lu es o f th e f eat ure s for e sti ma t e d future r a ng es for a ll 1,914 sp ec ie s . We cla ssifi ed a 541
speci e s a s e xtinct i f it wa s not pr edic t ed t o occ ur in a ny c ell in t h e corr e sponding t i m estep, a f t e r 542
ac counting fo r di spers al a s de s c ribed ab ove. 543
Analysis of chang ing tax onomic a nd s pat ia l patter n s of th reat 544
To e xplore w heth er specie s a r e pre dict e d to move through t he I U C N Re d Lis t s e q uentiall y ( i. e . , 545
inc r e a s e in t hre at th rough time ) , we fitt e d a n order ed log i s tic r egre ssion to t he p re dicted IU C N 546
t h r e a t s ta tu s u s i n g t h e MA SS pa cka ge v 7.3.65 85 . The re s pon se v ari able wa s th e pr edic ted IUC N 547
thre at st a t u s in 2040, 2 060, 208 0 and 2100 as in fe rred f rom a utom at ed a s s e ssme nt. The 548
inde pend ent va r i able s we re t he IUC N thr eat st atu s in th e previ ou s time s t ep, th e modelli ng sce na r i o 549
(com binati on of S S P a nd di sp er s a l s c ena r io; e. g. , S S P2.1 n o di spe r s al ), and th e ta x onomic Cla s s. 550
To e xplore g ene ral p at t e rn s ac r o s s s pe ci es in t he way range s ar e p r ed i cted to shif t with c lima t e 551
ch ange, we c alc ulat ed th e fol lowing th r e e metric s f or pr edic t e d r a nge s in e ach ti me s t ep and und e r 552
eac h c limate c hang e and di s per s al sc en a rio: 553
1. Direc tion o f r a nge shi ft : We c alcul a ted th e rang e ce n t roid f o r the pre dict ed rang e of ea ch 554
speci e s a t e ach t i me s t ep. W e th en c alcul ated the azimu th of e ach cent roid c omp ared t o the 555
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
23
ce ntr oi d in the pr ev iou s t i me ste p f or all predi ct ed rang e s i n 2040, 2060, 2080, a n d 2100, 556
wi t h 0° being due n orth . 557
2. Di st a nc e shi f ted: We c alc ula ted th e Eucli dean di s ta nc e in k m betwe en eac h c en troid to th e 558
ce ntr oi d in the pr ev iou s t i me ste p f or all predi ct ed rang e s i n 2040, 2060, 2 080, a n d 2100, 559
aft er pr ojec t i ng al l rang es to an eq ual a re a Au strali a Alb er s pr ojection ( E PS G:357 7). 560
3. Magn it u de of r ange cont rac t io n / ex pan si on: We c alc ula ted the a r e a in k m 2 of all p r e dict ed 561
range s a fte r projec ti ng to a n equ al are a Aus tr a li a Albe rs projec t i on (EP SG:357 7 ). W e th en 562
ca lcul ated log ( a re a in e ach time step / a rea in the p r e viou s time s t e p ) t o gen era te a 563
symmetrica l in dex whe r e nega tiv e va lue s indic ate ra nge contr ac t i on a nd po sitiv e val ues 564
indi cate r ang e ex pan sion . 565
We ex amined how spa tial pat te rn s of thr eat ene d s p eci e s ric hne ss c hange acr o ss Aus trali a w ith time . 566
For e ach tim e step an d c limate c hange s c enario w e ov e r l ayed all pr edi ct ed rang e s of spec ie s 567
predic te d to b e thr eat ene d in that time step on a map o f Au s t ralia in a n equal -ar e a Austral ia n Alb er s 568
projec t i on (EPS G:357 7) a nd tal li ed th e n u mber of thr e at ened spec i e s per 50x5 0km c ell. Con s e rv ation 569
pla nning i s o ft en ba s e d o n iden tify ing a rea s w ith co ncent rati on s of thre at ened sp ec ie s from 570
diff er ent hig he r-l evel t a xa ( e. g. , 7 ), so the c ongruenc e o f pat tern s o f th r ea t e n ed sp eci e s ric hne s s —571
the de gre e t o whi ch a n area tha t c ontain s many thre at ened sp ec ie s fo r one taxo n is al so exp ect ed to 572
co nt a in many thr eat ened sp ec ie s o f oth e r taxa—ma y hav e impor tant impl ica tion s for th e e ffic i enc y 573
of pla nning for pr otec t e d a rea s t o max imize biodiv er s i ty prot e c tion . We a sse s sed c ongruenc e in 574
thre ate ned spec ie s richne s s by c alc ulatin g s pa t i a lly c orr e ct ed P e ar s o n’ s co r r ela t i o n coef fic ien ts 575
(Tjos theim's c oe f ficie n t 86,87 ) using the `co r. s p ati al` func t i on f rom the S pat ial Pa c k pac kag e v0.4 .1 88 . 576
We c alcu lat ed c ongrue nce be twee n t hr e aten ed s pec ie s r i chn e ss of e ac h pair o f te rre s t rial ver tebr at e 577
cl asse s in e ach time s t ep and c limat e c ha nge s c en ario a fte r om itting double - zero c ells . We u sed an 578
ANC O V A to a s s e s s wh e t h e r congruen c e ( Tjo s the im’s c o ef ficie n t ) b etwe en dif fe r e nt t a xa c hang e s 579
ove r year s. 580
Fi nally, we d efin ed ho ts pot s o f t hr ea ten e d speci e s ric hne s s f or ea ch v erteb rat e cla s s a s th e ce ll s in 581
the to p 10 th perc entil e of thre at ened spe c ies richn e s s. W e down loade d a lay er o f prot e c ted a rea s i n 582
Aus t ral i a fr om Colla bora tive Au st r a lian P r otec ted A rea Dat aba se (C AP AD 89 ) . W e ca lc u l at ed o v e r l a p 583
of th rea ten ed spec ie s ho t s p ots wi t h cur r ent pro t e c ted a rea s a s th e pr op o r ti on o f hotspot s cel l s 584
who s e cen troi ds fall w ithin p rot ecte d are as . We t hen c ompar ed th e ov erlap o f th r eat ene d s p eci e s 585
hot spots w ith p rote cte d are a s in dif fer en t time ste p s a nd cli mat e c hange sc ena rio s t o as se s s t he 586
ade quacy of Au st r ali a’ s curr ent n etwork o f pr o tec t ed ar ea s f or pro tec ting s p e cie s in the fu t u re a s 587
t h e i r ra n g e s an d th r e at l e ve ls c h a n ge d ue t o c l i m at e c h an g e . 588
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
24
589
Acknowledgements
590
T h i s r ese a r c h w as f u nd e d by t h e C o ll e ge of S c ie n c e, A us tr a l i a n N at ion a l U n i v er s it y . B CS w a s 591
suppo rte d by T he Au st r ali an Re s earch Co unci l (AR C) th r ou gh a Di scov e ry Ea r ly Car eer Re s earc h 592
Awa r d (DE2 00100121 ) and a Fu ture F el lo wship (F T250100 113) . 593
594
AUTHOR CONTRIBUTIONS 595
Conc ep tualization – AS, M C, L B, XH, BCS ; D a ta curati on – AS ; S of t w are – AS , XH; F ormal analy s i s – AS ; 596
Fun ding a cquisi tion – MC , LB, XH , B CS; In v es tig a tion – AS ; Visu aliza tion – AS ; Wr i ti ng – origi nal dra ft 597
– AS , M C, LB, XH, BSC ; Wri ting – review & edit ing – AS, M C, L B, XH, BS C . 598
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
25
References
599
1. John son, C. N. et al. Bio dive r sity lo s s e s a n d conse rvatio n r e s pon s es in the An thr o poc ene. 600
S cie nce ( 1 979). 356 , 270–2 75 (2017 ). 601
2. Día z, S . et al . Pe rva sive human- drive n de cl ine of l i fe on Ea rth po in ts to th e n e ed f or 602
tran s for m a tive c hange . Sc ienc e (1979 ). 366, e aax3 100 (201 9). 603
3. F oden, W . B. et a l . Cl imat e c hange vulner abi lity a ss e ssment o f spec i es. W I REs Clim ate Ch a ng e 604
10 , (2019). 605
4. Ca r d ill o, M. , S chee l e, B. C . & T ulloc h, A. I. T . Forec a s ting extinc tion ri sk f or fu ture -p r oof 606
c onse r v ation d e ci sion s. Tren d s Ec ol. Ev ol. 41 , 112–1 19 (2026). 607
5. Ca r d ill o, M. , S keel s , A . & Dinna ge , R. P r io r itie s for c on s e rvi ng the w orld’ s t e rre stri a l mamma ls 608
ba sed on ove r-t he -horizon ex t i ncti on ri sk. Cu r ren t Bio lo gy 33 , 1381 –1388 ( 2 023). 609
6. Ca e tano , G. H . de O. et a l. Au tomat ed a sse ssmen t rev e al s tha t the ex t i nction risk of rep til e s i s 610
w idely undere stim ate d ac ro ss sp ace and phy logeny . PL oS Biol . 20 , e 3001544 (202 2). 611
7. Grenye r , R. et a l. Glob a l dis tr i bu t i on and c onse rvation o f r are and thr eat ened ver te br a t e s . 612
Nat ure 444 , 93–96 (2006 ). 613
8. My er s , N. , Mit term eier , R. A. , Mit t e rmei er, C . G ., Da F on s ec a, G. A . B. & Ke nt, J. B iodiv ersity 614
ho tspot s for c on se r v ati on priori tie s . Na t ur e 403, 853–858 (2000) . 615
9. Araújo , M . B., Ca be za , M ., Thui ller , W. , Hannah , L. & Willi am s , P . H. Wo uld c limat e c hang e 616
driv e sp ecie s ou t o f res erve s? An a sse s sme nt of exis ting re se rve- s elec ti on metho ds. Glo b. 617
Ch an g. Bio l . 10 , 1 618–162 6 (2004 ). 618
10. Hanna h, L . e t al. P r o tec t ed ar ea ne ed s in a chang ing c limate. Fro nt . Ec ol. En viro n. 5 , 131–138 619
(20 07). 620
11. T homas, C. D. & Gilling ham, P. K. The p e rforman c e of p rot ec t e d a rea s for bio div er s ity und er 621
c limate chang e . Biol o gica l Jo urna l of the Linn ea n S ocie t y 115, 7 18–730 (2015). 622
12. T homas, C. D. e t a l. P rot ect ed a r e a s fa cili t a te s p e cie s’ r ange expa n s i on s. P r oc eed i ng s of the 623
Nati onal Acade my o f S cie nces 109 , 14063 –14068 (2012). 624
13. L ehik oinen, P . , San ta ngeli , A . , Jaa tinen , K., Raja s ä r k kä, A. & L ehikoi nen, A . P rot ec te d are as ac t 625
a s a bu f fer aga in st de trimen ta l e f fect s o f cl imate c hange — Ev idenc e fr om large - sca le, 626
l ong-t erm abund a nce da t a . Glob . Ch a ng . Biol . 25 , 30 4–313 ( 2 019 ). 627
14. Co ldre y, K. M . & T urpi e, J. K. T he futur e r epre se nt a t i vene s s o f Mad aga s c ar’ s pr o te c ted ar ea 628
ne t w ork in th e fac e of cli ma te cha nge . Af r. J. E col. 59 , 253– 263 (202 1 ). 629
15. Po pe s c u, V. D., R ozyl owic z , L ., Cog ălnic ea nu, D. , N ic ulae , I. M . & Cucu, A . L. M ov ing i nt o 630
protec t e d a rea s ? S et t in g con s erva t i on pr ioritie s for R o manian rep t i l e s and amph i bians a t ri sk 631
from c limat e c hang e. PL oS O ne 8 , e7933 0 (2013). 632
16. Critc hl ow, R. et al. Mul t i- tax a spati al con se rvation p la nning r e veal s simila r p r i or i ti e s b e t w e en 633
tax a an d improved pro t e c ted a rea re pre s enta tion with c limat e c hange . Biodi v ers. Conse r v . 634
31 , 683– 702 (2022 ). 635
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
26
17. Hole , D. G. et al . P roje cted im pac t s of cl i mate c hange on a c on tine nt -wide pro t e c ted ar ea 636
ne t w ork. Ec ol. L et t. 12 , 4 20–431 ( 2 009). 637
18. Mi, C. e t al. G l obal Pro tec ted A r e a s a s r ef ug es for a mphibi an s a nd re ptile s und er c l imate 638
c hange . N at . Co mm un. 14 , 138 9 (2023 ). 639
19. Pe ng, S . et a l. Incorp ora ting g lobal chang e reve al s extinc t i on ri s k beyo nd the cu r r ent Red Li st . 640
Cu r rent Bi olo gy 33 , 3669 –3678 ( 2 023). 641
2 0 . V is c on t i, P . et al . Fu tur e hot sp ot s o f te rre strial mamma l lo s s . P hilo so phic al Tran s a ction s of the 642
Ro yal S ocie t y B: Biol ogica l Sc ienc es 366, 27 02–296 3 (2011). 643
21. Ca lvin, K. et al. I PCC, 20 23: Clim a t e Ch an ge 2023: Synthe s is R epo rt. Co n t ri bu t i o n o f Workin g 644
Grou ps I, I I an d II I to the Six th A sse ssme n t Repor t o f th e In te r go vern men tal P ane l on Clim a t e 645
Ch an ge [C ore W riti ng Tea m, H . L ee a nd J . Rome r o (E d s . )] . IPCC , Gen ev a , S witz e r l an d. (2023 ) 646
do i:10.593 27/IP C C/AR6-978 9291691 647. 647
22. CS IR O & Bu rea u o f M ete orology . Climat e C hange in Au strali a . P repri nt a t 648
http: / / w ww.c lima t e chang ei nau st r a lia .g o v.au/ (2023 ) . 649
23. Hea r d, G . W. et a l. D rought , fir e, a nd r ain for es t e nd emic s: A c a se st u dy o f two t h r ea ten ed 650
frog s impa ct ed by Australi a ’ s “ Blac k S ummer”. E col. Evol . 13 , e 10069 (20 23 ). 651
24. S opniew s k i, J. , Sc heel e , B. C. & Ca rdil lo, M. Pr edic ting th e di s tri bution of Au s tr a li an frog s and 652
thei r o ve rlap w ith B atrach oc hytrium d en drobati di s under c lima te c hang e . Div ers . Distrib . 28 , 653
12 55–126 8 (2022). 654
25. IUC N. I U CN R e d Lis t C at ego r i es a nd C r i t er i a : Ver s i on 3. 1 . (In te rnati onal U ni on for 655
Co n s e rva t i on o f Na ture , Gl and, S witze r l a nd, 201 2). 656
26. Ande rson , B. J. & Ohlemül l er, R . Cli ma te cha nge a nd pro tect ed a re a s : how w ell d o Br iti sh ra re 657
bryo phyte s f are ? in Br y ophyte Ecol ogy a n d Clima t e Ch ang e (ed s. Tuba, Z ., Sl ack , N . G. & S ta r k , 658
L . R.) 40 9–425 ( Ca mb r i dge Univ ersi ty P r e s s, C ambridg e, UK, 2010 ). 659
27. T huiller, W . et a l. Ar e dif fe ren t fac e t s o f p lant div ersi ty w ell pr otec ted ag ai ns t cli ma t e and 660
l and c over ch ange s? A te st study in the F r e nch Al p s . E cogr aphy 37 , 1 254–126 6 (2 014). 661
28.
Virk kala , R ., P öy r y , J., H eikk inen, R . K., L e hikoi nen, A . & Valk ama, J. P r ot ec te d are as al l evia t e 662
c limate chang e e f fec ts on n o r th ern bi rd s pec ie s of co n serva t i o n conc ern . Ec ol. Ev o l. 4 , 2991–663
30 03 (2014 ) . 664
29. S chee le, B. C . et al . How to improve t hr e aten e d s pecie s ma nag eme nt: An Au strali a n 665
pe rspec t i ve. J . E nviro n. Man a ge. 22 3, 66 8– 675 (201 8). 666
30. IUC N. The I U C N Red Lis t o f Threa ten e d S pec ie s. Ver s i o n 6.2. ht tp s: //ww w.iuc nred list. or g . 667
Dow nload ed on 02 July 2022 . Pr eprin t at (2019). 668
31. Bi r dL ife Int erna tiona l & H a ndb ook o f t he Birds o f t h e Wo r ld . Bir d sp ecie s di st ribut ion map s o f 669
the world . Ve r sion 202 2.2. Pr eprin t at (2 02 2). 670
32. Rol l, U. e t al. The g lob al di stribu tion o f t e trapod s r evea l s a ne ed fo r ta rge te d re ptil e 671
c onse r v ation . Na t. Ec ol. E vo l. 1 , 1677– 16 82 (201 7). 672
33. R Cor e Te am. R: a l a nguag e and e nviron ment fo r s ta ti s tic al c ompu ting. P r e pr i n t a t (2025) . 673
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
27
34. Be lbin, L ., W alli s, E ., Hob ern , D. & Zerger , A. T he Atl a s of Li ving A us trali a: H i st ory, c ur ren t 674
st at e and fu t ur e dir e c tion s. Bi odi v ers. Da t a J. 9 , e65 023 (202 1) . 675
35. We stga te, M ., Kellie , D. , Steve nson, M. & New man, P . gala h: Bio div ersity Data fr o m t he GBI F 676
Node Netwo r k . R pac kag e ve rsion 2. 1 .1. Prep rint a t ht tp s:// CRA N .R -677
proje ct. org/p ack age=g ala h (2025) . 678
36. John son, C. N. A u strali a’ s Mam mal Extin c tions : A 50,000 - Y ear Hi s tory . ( Camb ridge U niv er s ity 679
Pre s s, Cambridge , UK, 2006). 680
37. Ch ambe rlain, S . & Szöc s, E. t a xize - tax on omic s ea r c h and r e tri ev al in R. F100 0Re s. 2 , 191 681
(20 13). 682
38. Ch ambe rlain, S . et a l . taxi ze : T axono mic i nfor ma t i on fr o m ar ound th e w eb. P rep ri nt at 683
http s: //github .c om/rope n s c i / tax ize (202 0). 684
39. T ur tle Tax onomy Working Group . Tur tle s of the Wor l d : Anno tat ed Ch ec kli st and Atla s o f 685
T axonomy , Sy nonymy , Dis tribu tion, and Con se rvation S tatu s ( 9 t h Ed. ). in C ons er v a t i on 686
Bi ology o f Fres hwater Tu rtle s an d Tort oise s : A C ompil ati on Pr oject o f th e I UCN /SS C Tortois e 687
an d Fre shw ater T urtle Spec iali st Gro up (e ds. R hodin, A . G. J. et al .) 1 –472 (Ch e loni an Re s ea r c h 688
Monog r a ph s 8, 2021 ). 689
40. Co on ey, C. R . & Thoma s, G. H. H et erogen eou s rela t io n s hip s b etw e en rat e s of sp ec ia t i on and 690
bo dy size evolu t i on ac ro s s vert ebra te cl a des . N at . E c o l. E v ol. 5 , 10 1–110 ( 2 021 ). 691
41. Ite scu , Y. A bio geogra phi c pe rspec tiv e on turtl e ev olutio n. (20 12 ). 692
42. T homs on , S . & Georg e s, A. A new s p ecie s of fr e shwat er tu rtl e of t he g enu s El seya 693
(Te s tudin a t a : Pl eurodi ra: C helid ae ) fr om t he N o rth ern Te rrito ry of Au s tr a li a. Zo ota x a 4061, 694
18 –28 (201 6). 695
43. Oliv eira, B . F ., S ã o- Pe dro, V. A ., S an to s-B a r re ra, G ., P e non e, C . & Co st a, G . C. Amph iBIO , a 696
g lobal datab a se f or amphi bia n ec ologi cal t rai ts . Sc i. Dat a 4 , 1–7 (20 17 ). 697
44. S lave nko, A. e t al. Ev olut ion o f s ex ual size dimorphi sm in te trap od s is d riven by v arying 698
pa t t ern s o f s ex- speci f i c s elec tion on size . Nat. Ecol . Evo l. 9 , 4 64–473 (2025 ). 699
45. Mei r i, S . Trai ts o f th e l iza r d s o f th e w orld : va r ia t i on around a s uc ce s s ful evo lu t i on ary de s ig n . 700
Glob a l Ec ology and Bi og e ogr aphy 27 , 116 8–1172 (2018). 701
46. S t o r r , G. M . G eograp hic r ac e s of the ag a mid li zard Am ph ib olur us cau dicin ctu s . J . R. So c. Wes t. 702
Au st. 50 , 49 –56 (19 67) . 703
47. F uery, C . J . , W i t h e rs, P. C ., Hobb s , A. A . & G upp y, M. Th e rol e o f prot ein synthe si s during 704
me t a bolic de pre ssion in the Au s t ral i an d es ert fr o g N eob atr achu s c entrali s . Comp . Bi oche m. 705
Ph ysiol . A Mo l. I nte gr. P hy s iol . 119, 4 69– 476 (1998). 706
48. Iv er s on, J. B. , Enne n, J . & Lov ich, J. Life -hi st ory and ecolo gy da ta fo r t he turtle s o f t he world. 707
C he l o n ian C on se r v at i o n and B i o l o g y 24 , (2025). 708
49. Mei r i, S . Squ amB a s e — A da ta ba se o f sq uamat e (R e pt i li a: S quama ta ) t rait s. 709
Glob a l Ec ology and Bi og e ogr aphy 33 , e1 3 812 (202 4). 710
50. Pa r i s h, S ., Ric har ds, G. & Hall , L. A N a tura l Hist ory of Au stral ia n Ba tsWi0[ : Workin g t h e N ig h t 711
Sh i f t . ( CSIR O Publi s hing , M elbou rne, VI C, Australi a , 2012 ) . 712
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
28
51. Pa r n aby, H. E . A t axono mic rev iew of Au stralia n G r ea te r L ong- ea red Ba ts p r ev iou sl y k nown a s 713
Nyc tophilu s ti morie n s i s ( Chi ropt era : Ve s pertili onida e ) and some a ss oci at ed taxa . Au st rali a n 714
Z oologi st 35 , 39 –81 (20 09) . 715
52. Re ardo n, T. B. e t al. A m olecul ar a nd mor phol ogica l inv estig ati on of sp eci e s bound arie s and 716
ph yloge netic r ela t i on s hi p s in Au s t ralia n f r e e -ta iled ba ts Morm op teru s ( Chiro pte ra : 717
Molo s sidae ). A u st . J. Zo ol . 62 , 109–136 ( 2014). 718
53. Re ardo n, T. B. , Adam s, M ., McKe nzie , N. L. & Jenk in s , P. A n ew s p e cie s of Au stra li an fr eet ail 719
ba t Mormo p t e r u s e leryi sp . nov . ( Chir opt era: M olo s s i dae ) an d a tax onomic r e ap pr aisal of M. 720
no r f ol ken si s ( G ray ). Zoot ax a 1875, 1–31 ( 20 08). 721
54. Ri chard s, S. J. & Alf ord, R . A . S tr u ctu re an d d ynamic s of a r ain fore s t f r og ( Litori a 722
ge nim ac ula ta ) p opul a tion in no rthe rn Q ueen sland . A u st. J . Z ool. 53 , 229 –23 6 (20 05). 723
55. S illa, A. J., McF a dden, M. & By r n e , P. G . Hormone -induc e d sp awni ng of th e c ritic al ly 724
e ndange red northe rn c orr o b or e e f r o g P s eudo ph r y ne pe n gill ey i . Rep r o d uctio n, Fer ility and 725
Dev el opm e nt 30 , 13 52–135 8 (2018) . 726
56. S lave nko, A. , Tallow in, O. J. S . , It e s c u, Y., Raia, P . & M eiri, S . Late Qu ate rnary r eptil e 727
e xtinction s: size matt er s , i n s ul a r i t y domi na t e s. G l ob al E co l o g y a nd B i oge og r a ph y 25 , (2016 ). 728
57. Ca r di ll o, M. & Bromh am, L. Body size an d risk o f ex tinctio n i n Au s tr alia n mamma l s . 729
Co nse r v a tio n Biol o gy 15 , 1435– 1440 (200 1). 730
58. F ick , S. E . & H ij mans , R. J . Wo rldClim 2: n ew 1km spatial re s olu tio n cli mat e su r f ac es for globa l 731
l and are as . In te r n ati o n al Jo urnal o f Clim a tolo g y 37 , 4302–43 15 (2017). 732
59. Vi sca r r a Ro ssel, R . A. et a l . So il a nd La nd sc ape Grid Nati onal Soil A t t ribut e Map s (3 " 733
re soluti on) - Re l ea se 1. v 6. CS IR O. Da ta C o llec t i on . 734
http s: //doi. org/h ttp s:/ / doi .or g / 1 0 .4225/ 08/546ED604 ADD8 A (2014 ) 735
do i:ht t p s: // d oi. o r g / 10. 4225 /08/546E D60 4AD D8A. 736
60. Mal one , B. & S e arl e, R. S oil a nd La nd s c ap e Grid N atio nal Soil Attrib ute Map s (3 " re s ol u t i on ) - 737
Re lea s e 2. v 4. CS IRO. Dat a C ollec t i o n. (20 22 ) . 738
61. Jone s , B . & O’N eill, B . C. Sp a t i ally expli cit gl obal pop ulat ion scen ari o s c onsi sten t w ith the 739
S hared S ocio ec onomic P athwa y s. Env iro nmen tal Re s e a r c h L ett e rs 11 , 84003 (2 01 6). 740
62. Gao, J . G l ob al 1-Km D o wn scale d Pop ula t i o n Ba se Yea r a nd Projec tio n Grid s B as e d on th e 741
S hared S ocioe cono mic P ath ways, Re v isi o n 01 . (NASA S ocio e conomi c Da ta a nd Ap plic ation s 742
Ce n t e r (S EDA C ) , Pali s ade s, NY, 2 020). 743
63. F ujimori, S ., Ha segaw a , T., I t o , A ., Tak aha shi, K . & Ma s ui, T. Gr i dd ed emi ssion s and la nd-u s e 744
da t a for 2005-210 0 und er div er se soc ioe co nomic and c limate mi t i ga tion scen a r i o s . Sci . Da t a 745
5 , 180210 (2018). 746
64. Adhi kari, R . K., Yilm az, A. G . , Mainal i , B. & Dy s o n, P. P e rformanc e ev alua tion o f C MIP6 mod el s 747
for a ppl ica t i on to hydrolog ica l mod ell ing stu die s – A c a se study of A ustrali a . Sc ien c e of T he 748
T ot a l Env ir onm ent 945 , 17401 5 (2024 ). 749
65. E yring, V. et al. Ove rvi ew of th e Cou pled Mode l In terc ompa r i s on P r oje ct Pha s e 6 ( CM I P6) 750
e xperiment al de sign a nd org aniza tion. G eo s c i. Model D e v. 9 , 1937 –1 958 (2016 ). 751
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
29
66. S antini, L ., Ben í t e z-Ló pez, A . , M aior ano, L ., Če ngić, M . & H uij bre gt s , M. A. J. A sse ssing th e 752
reli abili ty of s pe cie s di s tr i bu tion projec t i o ns i n cli mat e ch ange r es earc h . Di vers . D is tri b. 27 , 753
10 35–105 0 (2021). 754
67. Hijma ns , R. J., Phil lip s , S . & Elith , J . di s m o : Spec ie s Di stribu tion Mod e ling. P r e p r in t a t (2021) . 755
68. Ph illip s, S. J . et al. S ampl e sel ec tion bi a s and pre senc e -o nly di stribu tion mode l s: i mplica t i on s 756
for ba ckg r ou n d and p seudo -a b senc e da t a. Eco logic al Ap pli cati on s 19 , 1 81–197 (2 009). 757
69. E lith, J. et a l . A sta ti stica l exp lan ation o f M a xEnt for ec ologi s t s . Di v e r s . D i st r i b . 17 , 43–57 758
(20 11). 759
70. Krame r-S chadt , S . et al. The i mporta nc e of corre cting fo r sampl ing bia s i n MaxE nt s pec ie s 760
di s t ribut ion mode l s. Di v e r s . Di st r i b . 19 , 1 366–1 379 (2013 ). 761
71. Araújo , M . B. & N e w, M . E nsemb le f orec a sting of s pe ci e s dis tr i bu t i on s. Tr e nd s Eco l. Ev ol. 22 , 762
42 –47 (200 7). 763
72. T huiller, W . et a l. biomo d2: En sembl e Pla tfo rm for Sp ec ie s Di stribu tion Mod el ing. R pack age 764
v ers io n 4.2-6 -2. Pr eprin t a t ( 2 025 ). 765
73. Rob ert s, D . R. et al . Cro ss -val idati on s tr a te gies for d ata w ith tempo ral, spa tial , hi e rarchi cal, o r 766
ph yloge netic st r uc ture . Ec ogra phy 40 , 913– 929 (201 7). 767
74. Va lavi, R ., Eli t h, J., La hoz-M onf ort , J. J. & G ui ller a- Arroita , G . bloc k 768
C VSČqF : An r pack a ge f o r g en era ting s p atia lly o r e nvironme ntal ly 769
sep ara ted f o ld s for k - fold c r o s s-v alida t i o n of sp ecie s di st ributi on mode l s . M et h ods Ec o l . E v o l. 770
10 , 225– 232 (2019 ). 771
75. Jone s , K. E. et al. PanT H ER I A : a sp ecie s -l e v el da t a ba se of li fe h i s t ory, e colog y, an d geog r a phy 772
of extan t and re cently extinc t mammal s. E colog y 90 , 2648 (200 9). 773
76. S antini, L . et a l. Ec ologic al c or r el a te s o f d is p e rsal di s ta nc e in t e r re s tr ia l mamma l s. Hy st r i x : th e 774
Ital i an J our nal o f M a mmal ogy 24 , 18 1–1 86 (2 013). 775
77. S chlos s, C. A ., Nuñ ez, T. A. & L awl er, J. J. Di s pe rsa l w ill l imit abil ity of ma mmal s to tr a ck 776
c
limate chang e in th e W e ster n Hemi sphe re. Proc eedin g s o f t he N a t i on al Ac ade my of Sc ienc es 777
10 9, 8606 –8611 ( 2 012). 778
78. Ch en, T. & Gu e st ri n, C . XGBoo st: a s c alab le tre e boo s ting sy st em. in P r oc ee di ng s o f th e 22n d 779
ACM S IGKDD In tern ati on al Con fe r e nce o n Know ledg e D i s c overy and Da ta Mi nin g 785–79 4 780
(AC M, N ew York, NY, USA, 2016 ). doi :1 0. 1145/2939 672.2939 785. 781
79. Niel sen , D . Tre e boo s ting w ith XGBo ost - w hy doe s XGBo o st win ‘every’ mach in e lea r ni ng 782
c ompetiti on? ( N orwe gia n Univ er s ity o f S ci ence and Te chnolo gy, T r on dh eim, No r way , 2016). 783
80. Ols on, D . M. e t a l. Te rre strial ecoreg ion s of th e world: a n ew map of l if e on Ea rth. B io s ci e n ce 784
51 , 933– 938 (2001 ). 785
81. Cre s s w ell, I. D ., J anke , T. & John s t o n, E. L . Au st r ali a Sta te o f t he Envi r o n me nt 2021 : O v erview , 786
In de pen d en t Rep ort to t he Au s tr a lia n G o ve r n me nt Mi nis ter for th e Env iro nme nt, 787
Co mmo nwe alt h o f Au s t ral ia , Can be r ra . (2 021). 788
82. Duc at ez, S. & Shin e , R. D river s o f extin cti on risk in te r r e str i a l vert ebra te s . Con serv . L ett. 10 , 789
18 6–194 ( 2 017). 790
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
30
83. Harf oot, M . B. J. e t al. U sing th e IU CN R e d L ist to map thr eat s to t e rre s t ria l ver teb rate s a t 791
g lobal s c al e. N at . Ec ol. E vol. 5 , 1510– 151 9 (20 21). 792
84. Ch en, T. et a l. x gboos t : Extreme Gradi ent Boo sting. R pac kage ver si on 1.7. 10 .1. P r eprint a t 793
http s: //doi. org/10 .326 14/C RAN.pa cka ge . xg boos t (20 25 ). 794
85. Ve na ble s, W . N. & Ripley , B . D . Mod ern A ppl ied S ta t i s tics w ith S . (S pr i nger , New Y or k , USA, 795
20 02). 796
86. T jøs th ei m, D . A mea sur e of a s soci a t io n f or s p atia l va r i a ble s . Biome t rik a 65 , 1 09– 114 ( 1 978). 797
87. Hubert , L . & Goll edge , R. G . Me as ur i ng a s s o cia tion be tween s pa tia lly de fined v ari able s: 798
T jøs th ei m’ s c oef fic ie nt index and some e xtensi on s. Ge ogr . An al. 14 , 273– 278 (198 2). 799
88. Va llejo s, R ., O so r io , F . & Bev ilac qu a, M. S pati al Re la tio ns hip s Be twee n Two G e oref erence d 800
Va r i abl es : W it h Ap p l i cat io ns i n R. (S p r in g er, N ew Yo r k , NY, USA, 202 0 ). 801
89. Co mmonwea l th of Au s tr alia . Coll a b orat i ve Au stra lian Pro tec ted Ar ea s D at a ba s e ( CAPA D ) 802
20 20 . (2021 ). 803
804
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.13.718320doi: bioRxiv preprint
Text is read by the "Ask this paper" AI Q&A widget below. Extraction quality varies by source — PMC NXML preserves structure cleanly, OA-HTML may include some navigation residue, and OA-PDF can have broken hyphenation. The publisher copy (via DOI) is the canonical version.
My notes (saved in your browser only)
Ask this paper
Answers must be backed by verbatim quotes from this paper's full text. Hallucinated quotes are dropped automatically; if no verbatim passage answers the question, we say so. How this works
Citation neighborhood (no data yet)
We don't have any in-corpus citations linked to this paper yet. This is a recent paper (2026) — citers typically take a year or two to land, and the OpenAlex reference graph may still be filling in.
Source provenance
- europepmc
- last seen: 2026-05-20T01:45:00.602351+00:00
- unpaywall
- last seen: 2026-05-22T02:00:06.705733+00:00
License: CC-BY-NC-ND-4.0