(Fig. 1B). A mass extinction in this emissions
scenario is projected across all potential
values of the extinction threshold except in
the unlikely case that the average species can
maintain a viable population in <10% of
their initial habitat volume (fig. S10; 19 ). By
contrast, limiting warming to 2°C would cut
the severity of extinctions by >70%, avoiding a

marine mass extinction across all extinction

thresholds.

Climate-driven species losses vary widely

among ocean biomes, with important implications

for fisheries and the patterns of biological

richness (Fig. 2). These spatial patterns also

provide crucial observational tests of the un-

derlying model.

Extirpation risk is greatest where climate

anomalies are strong (fig. S2) or species are

living close to their ecophysiological thresholds,

typically equatorward near the warm and/or

low-O 2 edges of their ranges (Fig. 2, A and B).

Vulnerable regions include highly productive

ecosystems where background O 2 is already low,

such as the north Pacific, eastern boundary

SCIENCEscience.org 29 APRIL 2022¥VOL 376 ISSUE 6592 525

0 0.5 1

Marine Biological Richness

0 50 100

Extinction (%)

Latitude

(^0) 60°E 120° E 180° 120°W 60°W 0
Longitude
30° S
0
30° N
60° N
90° N
Latitude
0 25 50 75 100
Extirpation (%)
2100 CE high emissions
mean (line)
range (shading)
2020 historical emissions
2100 low emissions
2100 high emissions
2300 high emissions
60° S
0 50 100
Extirpation (%)
60°S
30°S
0
30°N
60°N
90°N
60°S
30°S
0
30°N
60°N
90°N
60°S
30°S
0
30°N
60°N
90°N
Latitude Latitude
Model
Observations 2300 high emissions
mean (lines)
s.d. (shading)
ABC D
Fig. 2. Spatial variation in species losses and marine biological richness.
(A) By 2100, regions of strong extirpations overlap past productive fisheries
(blue points), where catch rates exceed the global median from 1950 to 2014
( 36 ). (BtoD) Patterns of extirpation risk (B), marine biological richness (C), and
global extinction risk [(D); averaged across colonization scenarios] are shown
versus latitude. Observed biological richness (number of species) estimated
using rarefaction ( 31 ) is reproduced by the trait-based habitat model
applied to climatological distributions of temperature and O 2 ( 37 , 38 ) across a
range of maximum summation depths [(C) line is 500 m; shading is 0
and 5000 m; ( 19 )]. Richness is normalized to the maximum observation.
Extirpation and global extinction risks are averaged from 0 to 500 m, across
Earth system models, and across longitude in (B) and (D).
0 50 100
0
20
40
60
80
100
Extinction (%)
5 th
50 th
75 th
95 th

Past Occurrences

End-Permian
Late Ordovician
End-Cretaceous
Late Devonian
End-Triassic
The “Big 5”
All time
Percentiles
A BC
0 5 10 15 20
2300 CE
2100 CE
2020 CE2.6 8.5
Permian
(model)
Threatened (IUCN)
0% co
loniza
tion
100% colonization
0 5 10 15 20
0
20
40
60
80
100
Extirpation (%)
2300 CE
2100 CE
2020 CE 2.6 8.5
Permian
(model)
Threatened (IUCN)
CO
2
1900 2000 2100
0
15
30
emissions
High emissions
(8.5 W m-2)
Low emissions
(2.6 W m-2)
Historical
(^001234567)
10
20
30
Extinction
Late Devonian
End-Triassic 1850-2100 CE
Fig. 1. Marine extinctions in the past and risk from climate warming.
(A) Extinction intensity (percent losses) from the fossil record of marine animal
genera over the past ~542 million years ( 12 ), including the“Big 5”mass extinctions
( 35 ) (fig. S1A). (BandC) Projected global extinction (B) and global mean extirpation
risks (averaged over 1° latitude × longitude, 0 to 500 m) (C) rise with increases in
annual mean global surface air temperature (thick lines are intermodel averages; thin
lines are individual Earth system models) and are plotted under historical greenhouse
gas emissions (petagrams of carbon per year) and divergent future scenarios [(C),
inset], yielding radiative forcings of 2.6 W/m^2 [i.e., RCP/SSP1-2.6; ( 6 , 19 , 21 )] versus
8.5 W/m^2 in 2100 [i.e., RCP/SSP5-8.5; ( 6 , 19 , 21 )]. Inset in (B) zooms in to show end
of the century changes. Percent colonization refers to the fraction of new habitat that a
species can instantly disperse to and inhabit (thin dashed line is the median case).
Extinction risks from current anthropogenic threats are estimated from IUCN
vulnerability assessments (orange bars; tables S3 and S4; 19 ). Simulated end-Permian
extinction risks are shown across colonization scenarios in (B) and in (C) display
the global mean and spatial SD of extirpation. Points and horizontal lines on the top
axis show the average and range of warming across Earth system models evaluated
here, respectively, for different time horizons and emissions scenarios.
RESEARCH | REPORTS