Add files using upload-large-folder tool

.gitattributes

@@ -60,3 +60,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 dd9fa622529abdee9160c93e7c3ad7399577e19fb9862692ba9a7f882cc52c8a/preprint/images/Figure_4.tif filter=lfs diff=lfs merge=lfs -text
 dd9fa622529abdee9160c93e7c3ad7399577e19fb9862692ba9a7f882cc52c8a/preprint/images/Figure_1.tif filter=lfs diff=lfs merge=lfs -text
 dd9fa622529abdee9160c93e7c3ad7399577e19fb9862692ba9a7f882cc52c8a/preprint/images/Figure_2.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_6.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_7.tif filter=lfs diff=lfs merge=lfs -text
+507a30887b7df476a3d7501cf7fb4c37ab7e6aa40feae7e15a141aa968a843e1/preprint/images/Figure_8.tif filter=lfs diff=lfs merge=lfs -text

05225d17b4449ec0559648b67741e3748a3e400b9813ff34e06b11744539c34d/preprint/images_list.json

@@ -0,0 +1,42 @@
+[
+    {
+        "type": "image",
+        "img_path": "images/Figure_1.png",
+        "caption": "Focal regions and example climate of the Early Jurassic, north-west Tethys. (A) Symbols indicate all fossil occurrences between Margaritatus and Bifrons ammonite zones, grouped into regions (coloured, delineated, and labelled in the legend) by hierarchical clustering based on occurrence paleocoordinates. Coordinates, maximum sea level coastlines (thin black lines), and deeper waters (dark blue, demarcated by \u22121400 m contour) were reconstructed according to Pliensbachian (185 Ma) paleogeography of the PaleoMAP model 36. The landmass of Iberia is labelled. (B) An example of the utilised CLIMBER-X downscaled mean summer sea surface temperatures at the 185 Ma (Pliensbachian) paleoconfiguration and 750 ppm atmospheric CO2. Global location shown as box in world map (inset top left) alongside lines of latitude every 30 degrees including the equator. BM = Bohemian Massif; MC = Massif Central; AM = Armorican Massif; SM = Scottish Massif; E., W. and N. Iberia are east, west, and north of Iberia.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_2.png",
+        "caption": "Species\u2019 thermal bias is correlated with their escalating response over the two warming and transitional phases, but not over the two stasis phases. (A) Summary of the climatic phases under study, with ammonite (sub)zone time bins on the x-axis. The solid line shows the main CO2 scenario, while the dotted line shows more extreme estimates. The stage boundary absolute timing has an error \u00b1 0.4 Ma37. (B-F) Each panel shows two regressions: the solid line regressions run across immigrant, persisting, and extirpated species only; the dashed line regressions run across all five ordered response levels. Regressions use regions nested within time zones. Circles show species responses with a small horizontal jitter to avoid overplotting of points against their thermal biases per region, the numbers of which are given along x-axis, with box plots showing the medians and interquartile ranges.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_3.png",
+        "caption": "Stronger thermal determination of species occupancy change (i.e. increasing slope coefficients, one per spatiotemporal scenario) under higher magnitudes of regional temperature change. Different lines show the relationship\u2019s dependence on sampling intensity (slope parameters generated from more species are more likely to support the relationship but few points meet these higher thresholds). Filled circles are those with at least 20 species, their regression shown by the solid black line, R = -0.25, 95% Cis = -0.49\u2014-0.01, P = 0.04, while the slopes of the other lines were insignificant (next closest was threshold = 10 with R = -0.23, P = 0.06, see SM). Direct exploration of the effect of species number threshold on the relationship is shown in (Fig. S2). Each point is a spatiotemporal assemblage (n = 25) with error bars being the standard error of the y-axis slope parameter, representing species response variation within each spatiotemporal assemblage. Standard errors were used for inverse weighted least squares regression. Extinct species\u2019 occurrences are here merged with those of extirpated species and originating species occurrences are merged with those of immigrating species (the same result was achieved treating these groups separately, R = -0.18, P < 0.05, with threshold = 20).",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_4.png",
+        "caption": "Assessment of whether the thermal bias of an existing assemblage is related to its future response to environmental change, or whether direct descriptors of environmental change magnitude are more useful. Figure should be read from row to column, with the intersecting panel showing the correlation and its significance, expressed by the legend as the associated %-change in the column variable. For example, reading the first row: for each degree of assemblage thermal bias below ambient (during warming and transition phases), the share of originating species in the new assemblage increases by ~1.3% (P = 0.001). The panel ellipses represent coefficients from nested random effects models by colour (see legend) and direction of elongation. ** is P < 0.01, * is P < 0.05, dot is P < 0.2. The first two rows have a unidirectional hypothesis between change and response; 4 regions nested in 3 time zones, n = 12 assemblages, when Germanic basins responses were unavailable. Other rows cover all five time zones with a bi-directional hypothesis between change and response; 5 clusters nested in 5 time zones, n = 22 assemblages, with Germanic basins responses unavailable for three zones. The last two rows are %-change of occurrences per zone and cluster that are categorised as primarily carbonate lithology or \u2018deep\u2019 depositional environment, indicating larger changes in habitat sampling within a region.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    },
+    {
+        "type": "image",
+        "img_path": "images/Figure_5.png",
+        "caption": "Designation of region-specific species responses observed around a focal boundary (emboldened, with the ammonite zone immediately preceding it being time i). We focus on the responses of regional two-timer species, specifically the lower two-timers for extirpated or extinct species, boundary crossers for persisting species, and upper two-timers for immigrating or originating species. These responses are ordered with respect to expectations of thermal bias. FAD = First Appearance Date. LAD = Last Appearance Date.",
+        "footnote": [],
+        "bbox": [],
+        "page_idx": -1
+    }
+]

05225d17b4449ec0559648b67741e3748a3e400b9813ff34e06b11744539c34d/preprint/preprint.md

@@ -0,0 +1,243 @@
+# Abstract
+
+A mismatch of species thermal preferences to their environment may forewarn that some assemblages will undergo greater reorganization, extirpation, and possibly extinction, than others under climate change. Here, we examined the effects of regional warming on marine benthic species occupancy and assemblage composition over one-million-year time steps during the Early Jurassic. Thermal bias, the difference between modelled regional temperatures and species’ long-term thermal optima, predicted species responses to warming in an escalatory order. Species that became extirpated or extinct tended to have cooler temperature preferences than immigrating species, while regionally persisting species fell midway. Larger regional changes in summer seawater temperatures (maximum + 10°C) strengthened the relationship between species thermal bias and the escalatory order of responses, which was also stronger for brachiopods than bivalves, but the relationship was overridden by severe seawater deoxygenation. At + 3°C seawater warming, our models estimate that around 5% of an assemblage’s pre-existing benthic species was extirpated, and around one-fourth of the new assemblage being immigrated species. Our results validate thermal bias as an indicator of future extinction, persistence, and immigration of marine species under modern magnitudes of climate change.
+
+Earth and environmental sciences/Ecology/Climate-change ecology  
+Earth and environmental sciences/Ecology/Palaeoecology  
+Earth and environmental sciences/Ocean sciences/Marine biology  
+Earth and environmental sciences/Ecology/Macroecology  
+climate change  
+extinction  
+community temperature index  
+niche  
+extirpation
+
+# Introduction
+
+A suitable temperature is one of the most commanding habitat requirements for species, especially at broad spatial scales<sup>1</sup>. Human activity has set isotherms on the move globally<sup>2,3</sup>, leading to poleward shifts of marine species’ geographical ranges<sup>4,5</sup> and species departing the tropics<sup>6</sup>, with substantial repercussions for human well-being and ecosystems<sup>7</sup>. Warming is projected into the coming centuries<sup>8</sup> when climate change is anticipated to supplant land use change as the dominant driver of species extinction<sup>9</sup>. However, species range shifts may leave clues to predict extinction risk<sup>10</sup>. Range shifts are expected to begin with an extension of the leading edge, as a species arrives into new habitat<sup>4,11</sup>, while trailing edge populations suffer performance decline. Marine heat waves can cause physiological stress to trailing edge populations<sup>12</sup> sufficient to cause their extirpation<sup>10,13</sup>. The proximity to a species’ thermal niche edge should therefore indicate how a given population might react to warming<sup>14,15</sup>, particularly for marine ectotherms, whose distributions tend to be closely associated to their thermal tolerances<sup>16</sup>. Additional, future observations will provide greater predictive confidence but a species is irretrievable once extinct and climate-induced extirpations are already widespread<sup>5</sup>. Rather than waiting for climate-induced extinction to manifest, the rich fossil record has great potential to explore links between climate-induced extirpations and global extinctions<sup>17</sup>, especially given the recurring Earth system responses to a rapid addition of atmospheric CO<sub>2</sub><sup>18,19</sup>.
+
+Ecological assemblage change may be the rule over long time scales, allowing the fossil record to elucidate links between species range shifts, turnover, and extinction<sup>20</sup>. Climate change is consistently tied to organism latitudinal range shifts and regional turnover, covering multiple marine fossil groups and time scales<sup>21–23</sup>. Global warming also encourages seawater deoxygenation in both the modern and the past<sup>18,24</sup>, which can either make populations more sensitive to warming<sup>25</sup> or supersede the impacts of warming completely as anoxia<sup>26</sup>. However, it remains unclear the degree to which fossil thermal preferences or niches can be associated during warming with the regional suitability or vulnerability of populations, species, and assemblages.
+
+Thermal optima can be estimated for a species based on its geographical distribution (species temperature index, STI), which can be combined to estimate assemblage-level net preferences (community temperature index, CTI)<sup>27,28</sup>. An STI or CTI falling behind environmental change signifies a thermal bias, the difference between species long-term median temperatures (STI) and ambient seawater temperatures<sup>11,28</sup>. Thermal bias can indicate more populations to be further from their respective species thermal optimum, potentially making the assemblage more vulnerable to species turnover than others<sup>11,28</sup>. In marine shallow-water fauna, assemblage thermal bias may even be more indicative of species loss than regional warming rates<sup>28</sup>. Thermal bias, STI, CTI, and thermal niches are commonly used measures for species or community vulnerability under climate change. Although the thermal bias of fish species has been correlated with changes in their local abundance and occupancy<sup>11</sup>, the wider validity of these metrics is rarely tested, especially at the assemblage level and their links to extinction risk.
+
+We expect that, (A) under warming, species’ occupancy responses are ordered with respect to, and dependent on their thermal bias (regression: <em>response</em> ~ <em>thermal_bias</em>). This means that regional immigrant species tend to have relatively warm thermal biases i.e., on average they have preferences for warmer waters than the ambient conditions, while extirpated species and those going extinct tend to have relatively cool thermal biases. Finally, persisting species tend to have relatively intermediate thermal biases. Species originating or going extinct could be considered the climaxes of this escalatory ordering of responses (originating = 1, immigrating = 2, persisting = 3, extirpated = 4, extinct = 5), which indicates how well-adapted a species was to the new environment. (B) Thermal determination of species occupancy response is stronger (hypothesis A regression slope becomes steeper) with greater regional climate change and for climate sensitive clades (brachiopods more sensitive to warming than bivalves<sup>e.g. 29</sup>). (C) If a region is warmer than the thermal optima of many of its species (i.e. regional assemblage is cool-biased), further warming will summon extensive assemblage change. Conversely, a region with little assemblage thermal bias, or occupied by species with warmer optima than ambient temperatures (warm-biased assemblage), will change little under further warming. We test these expectations using mixed effects models to account for nested species, regions, and time zones. To guard inferences against changes in sampling intensity between time bins, we calculate regional rates of species immigration, persistence or extirpation as the numbers of two-timer species (see Methods). Extinctions and, for completeness, originations were identified by dataset-wide last or first appearance dates (LADs or FADs) of two-timer species.
+
+Our study system consists of the bivalve, brachiopod, and gastropod species of the epicontinental seas adjacent to the north-western Tethys during an Early Jurassic extinction event<sup>30</sup>. We identified major clusters of sampling and focus on these as discrete ‘regions’ (Fig. <span class="InternalRef" refid="Fig1">1</span>), being similar in area to modern regions used to investigate thermal bias<sup>31</sup>. The Late Pliensbachian to Early Toarcian interval of the Early Jurassic covers a transition from cool global temperatures, potentially with polar ice sheets<sup>32,33</sup>, through rapid global warming with potential modern relevance<sup>18</sup>, to stabilisation as a greenhouse climate. This can be generalised, at ammonite zone temporal resolution (mean = 1.1 myr), into the following phases (Fig. <span class="InternalRef" refid="Fig2">2</span> A): little change between the two Late Pliensbachian zonal means (‘cold stasis’); warming into the earliest Toarcian zone (‘warming phase 1’); further warming during the Toarcian Ocean Anoxic Event (T-OAE; ‘warming phase 2’); an initial continuation of peak warm conditions before cooling slightly (‘transitional phase’, having the highest mean temperature); a stable, warm climate (‘warm stasis’). We used literature estimates of CO<sub>2</sub> concentrations, or geochemical proxies to indicate target seawater temperatures, for forcing the climate model, CLIMBER-X<sup>34</sup>. We focus on the derived spatial variation in summer mean temperatures because maximum temperatures may drive species extirpation<sup>13</sup>. We apply our models to estimate responses to + 3°C regional warming because, although there are many complications to apply paleobiological models to modern change, this value is projected under high emissions scenarios (RCP 8.5) by the end of the century in the North Sea<sup>35</sup>.
+
+# Results
+
+## Thermal bias associated with escalating species responses to warming
+
+Over the two warming phases and the transitional phase, species’ occupancy responses formed an escalatory order in relation to their thermal bias (Table 1; Fig. 2 C-E). The significance of this regression coefficient was robust to whether origination and extinction responses were included separately, treated respectively as immigrations and extirpations, or excluded entirely (Fig. 2, Tables S1 & S2). During these phases, negative (cool) thermal biases prevailed; immigrating species’ thermal optima approximated ambient water temperatures (species mean thermal bias = + 0.5°C), whereas extirpated species had much cooler thermal biases of -4.3°C. This observed mean fell below the linear model expectation (thermal bias expectation for extirpated species = -3.0°C, conditional mean 95% CIs = -4.1—-1.9°C), while all other response levels approximated linear expectations. Persisting species’ thermal optima were significantly below local temperatures during the climate warming and transition phases (mean = -1.1°C), species going extinct had the coolest biases (mean = -5.0°C), and originating species had the warmest preferences (mean = + 1.8°C).
+
+Thermal bias was a stronger predictor of species occupancy response than the climate model-derived magnitude of regional warming during the warming and transitional phases (Table 1; when individually modelled as fixed effects, *R*<sup>2</sup><sub>marginal</sub> = 0.18 vs. *R*<sup>2</sup><sub>marginal</sub> = 0.004, respectively). These results were maintained under alternative CO<sub>2</sub> and paleogeographical scenarios (Table S3), an alternative approach to control for sampling variation (Fig. S1), and no evidence could be found for impacts of changes in habitat substrate (Table S4). Changes in water depth over the first warming phase coincided with an apparent immigration event to Eastern Iberia (Tables S5 and S6) but the effect of thermal bias remained when this was accounted for, including if this region over the first warming phase was removed from analysis (see SM, Table S4).
+
+## Sources of variation in species responses to thermal bias
+
+Support for the escalatory relationship between species occupancy and thermal bias was spatiotemporally and taxonomically variable. Brachiopods were more affected than bivalves by the magnitude of regional warming and marginally so for their thermal bias (Table 1). Greater regional warming strengthened the escalatory response of species to thermal bias, expressed by a steeper regression slope. This was best supported when sample sizes were larger (i.e. more species), representing better sampling but also more oxic conditions (Fig. 3, S2). Specifically, in both phases of climatic stasis, there was no significant relationship between thermal bias and occupancy response (Fig. 2). Support for the relationship was also weak to absent in the British basins and north of Iberia during the widespread bottom anoxia of the transitional phase, and throughout the Toarcian in the Germanic basins because of dwindling occurrences (Fig. S3). This was despite the northern regions (Germanic and British basins and north of Iberia) experiencing the largest warming magnitudes, with climate scenarios estimating + 7—10°C over the two combined warming phases (up to + 14°C in a less likely CO<sub>2</sub> scenario). Following our climate modelling, the British and Germanic regions were initially the coolest at 19—21°C and warmed the least in the first warming phase (+ 1.3°C and + 1°C, respectively), but experienced massive warming over the second warming phase (+ 6—9°C and + 7.5°C, respectively, vs. +4—5°C in other regions; Fig. S3).
+
+**Table 1**  
+A species’ occupancy response is dependent on its thermal bias, regional temperature change, and clade membership during the climate warming and transition (warmest) phases. Interaction terms test for differences in these relationships between bivalves and rhynchonellid brachiopods. Here, extinction responses were treated as extirpations and originations as immigrations but the significant coefficients remain similar whether this treatment was removed or whether extinctions and originations were removed entirely (Tables S1 & S2). Results from a mixed-effects model with species nested within regional cluster, and cluster nested within time zone. Number of observations n = 431, bivalves n = 275, brachiopods n = 146. 10 gastropods and 1 lingulid brachiopod observations removed. *R*<sup>2</sup><sub>marginal</sub> = 0.30, *R*<sup>2</sup><sub>conditional</sub> = 0.62, estimating the variance explained by the fixed effects alone, and both fixed and random effects, respectively.
+
+|  | Value | S.E. | *t* | *p* |
+|---|---|---|---|---|
+| (Intercept) | 2.97 | 0.11 | 26.04 | < 0.0001 |
+| Thermal bias °C | -0.09 | 0.01 | -10.00 | < 0.0001 |
+| Regional temperature change °C | -0.03 | 0.03 | -1.19 | 0.268 |
+| Clade_Rhynchonellata | -0.05 | 0.07 | -0.74 | 0.458 |
+| Thermal bias:Clade Rhynchonellata | -0.03 | 0.02 | -1.89 | 0.076 |
+| Regional temperature change:Clade Rhynchonellata | 0.06 | 0.02 | 3.19 | 0.002 |
+
+## Assemblage-level thermal bias and responses
+
+Faunal responses to climate change are often measured or projected at the level of assemblage. To assess thermal bias at the assemblage level, we take the mean thermal bias over species regionally present before climate change and correlate it with the proportion of a given assemblage that were extirpated or went extinct, the proportion of species added via immigration or origination, and the overall turnover. Over all climate phases combined, assemblages accumulated thermal bias moderately as the ambient temperature changed, adding − 0.41°C thermal bias (95% Cis = -0.77—-0.05°C) for each degree of warming, rather than maintaining perfect equilibrium (0°C thermal bias per degree) or not responding at all (-1°C thermal bias per degree; calculated by a mixed effect model between regional temperature change and thermal bias). Relationships were not different if assemblage thermal bias was weighted towards cool- or warm-adapted members of the assemblage (Table S7).
+
+Focussing on the climate warming and transition phases, a cooler assemblage thermal bias consistently increased the proportions of species changing, either going extinct, being extirpated, or subsequently immigrating or originating, but was only significantly correlated with an increase in originations (+ 1.3% in the subsequent assemblage per − 1°C assemblage thermal bias, 95% Cis = 0.6—2.0%; Fig. 4). The magnitude of regional warming was significantly correlated with an increase in immigrating species as a proportion of the subsequent assemblage (+ 8.5% per 1°C increase in water temperature, 95% Cis = 4.2—12.8%; Fig. 4). The influence of regional warming magnitude on the escalating occupancy response (e.g. Figure 3) was supported at the assemblage level by the correlation between the proportion being extirpated and the proportion going extinct increasing to R = 0.73 (P = 0.006) during the climate warming and transition phases, up from R = 0.40 (P = 0.058) across all climatic phases (R values from mixed effect models of standardised variables). Meanwhile, the shares in a new assemblage of immigrated or originated species were moderately correlated both during the climate warming and transition phases (R = 0.49, P = 0.092) and across all phases (R = 0.45, P = 0.033; mixed effect models of standardised variables). No significant effect of changes in broad habitat substrate or water depth was found on assemblage level responses (Fig. 4, explored further in SM).
+
+Projecting the models in Fig. 4 to + 3°C seawater warming estimates that 4.74% (0.03—9.45%) of an assemblage’s pre-existing benthic species to be extirpated and 25.5% (95% Cis = 12.5—38.4%) of a new assemblage to be newly immigrated (see SM ‘Application of results’).
+
+# Discussion
+
+## Does a species thermal bias predict its removal under climate change?
+
+Regional species loss is often correlated with climatic changes<sup>13</sup> but considering climate change relative to species’ thermal niches leverages additional information to assess population vulnerability<sup>28,31</sup> (this study). We present empirical evidence from the fossil record that immigration, persistence, extirpation, and likely the extinction of species form an escalating response gradient linked to species suitability to regional conditions, as estimated by their thermal bias. A species’ thermal bias is thus a useful attribute to predict its likely response to warming, though it is likely insufficient alone (e.g. here <em>R</em><sup>2</sup> = 18%), with magnitude of regional warming and taxonomic membership explaining additional variation. Thermal biases of individual species were highly variable across responses despite the spatiotemporal extent and focal species being well-sampled, which supports the validity of responses such as extirpation. In some times and regions, seawater anoxia likely overruled the importance of temperature in determining assemblage membership. The otherwise consistent temperature–response relationship supports cautious use of habitat suitability models based on temperature, ideally alongside additional niche variables<sup>31</sup>, to provide evidence on extinction risk as a statistical tendency over multiple species<sup>38</sup>.
+
+For simplicity, especially given the large time scales, we assumed a linear relationship between a species’ thermal bias and its regional occupancy response, from immigration, through persistence, to extirpation<sup>detailed in 10</sup>, and potentially extinction. This assumed an open thermal landscape over which species could disperse, which was relatively well supported, though sampling was bookended between approximately 15 and 34°C (Fig. S4), with geographical barriers to the north (discussed below). Non-temperature habitat (and sampling) heterogeneity is likely to dominate at scales beneath our effective spatial resolution of 1000s km. As climate changes, species distributions should move through a region, following shifting isotherms according to their thermal tolerance, other habitat variables permitting (see below)<sup>10</sup>. Though we observed occupancy responses, these are likely to have additional dimensions that also escalate according to thermal bias, such as larger body sized species being regionally replaced by smaller, opportunist species<sup>39</sup> and being more likely to go extinct<sup>29</sup>.
+
+The influence of thermal bias on occupancy response was more pronounced in rhynchonellid brachiopods than bivalves, the latter having a lower mean vulnerability to extinction during rapid warming<sup>29,39,40</sup>. Thermal performance has ecophysiological underpinnings<sup>25,41,42</sup>, with some organisms having more specific or limited physiological adaptations. Evidence is mounting that different ecophysiological adaptations among taxa lead to different performance outcomes, including extinction risk<sup>25</sup>, though quantitative comparisons of the thermal performance of brachiopods and bivalves are scarce<sup>43</sup>. Our results therefore support the view that ecophysiology predisposes vulnerable taxa and traits to greater species immigration and extirpation at multiple scales, and their extinction risk is predictable via habitat loss<sup>42,44,45</sup>. Groups vulnerable to warming may thus be more likely to show strong range shift responses, where habitat permits, such as bony fish<sup>4,25</sup>, relying on escape rather than tolerance. Other vulnerable groups, such as reef corals, may be more restricted in their rate of habitat tracking (though see<sup>22</sup>). Identification of vulnerable clades or traits, alongside spatial projections of their habitat loss from physiological principles and environmental factors<sup>42</sup>, may aid understanding of the pressures regional warming will place on species<sup>44–46</sup>.
+
+## Linking climate-driven range shifts to extinction risk
+
+Species may avoid climate-induced extinction by colonising newly suitable habitat<sup>13</sup>, hence marine fauna are expected to consistently trace their thermal preferences during climate change<sup>21</sup>. Therefore, an escalating occupancy response in line with a species’ thermal bias may be a null expectation for a region under warming<sup>10,15</sup>, leading species with particularly cool thermal biases to be vulnerable to local and global extinction. However, there are modern examples of how disequilibria between ambient temperatures and a population or assemblage can be stable rather than a dispersal failure, especially when observed at finer spatiotemporal scales than sampled here<sup>15,28</sup>. Even at our scales, we observed an especially large variation in thermal bias for persisting species, suggesting either that finer scale thermal heterogeneity played a substantial role in permitting species to persist (i.e. refuges), or that many species were temperature generalists. Nevertheless, our finding that regional warming increased the slope of the relationships between thermal bias and response implies that greater magnitudes of warming on average increase the cost of disequilibria between species and climate. Furthermore, the overwhelming negativity of thermal biases across responses during warming phase 2, which coincided with the highest ratio of extirpations and extinctions to persisting species, may also have been amplified by the warming on-top-of warming climatic context, which increases extinction risk<sup>47</sup>.
+
+The largely overlapping thermal bias values for species being extirpated and going extinct, as observed here, may betray a cause of extinction. The mean thermal bias for extirpated species fell below the expectations of our linear model and instead fell within 95% confidence intervals for species going extinct. Meanwhile, observed thermal biases for immigrating and originating species aligned well with linear expectations. Either the thermal bias values for extirpations were unusually high, or the thermal bias values of the extinctions were unusually low. The first option assumes the true relationship was linear and thus the thermal bias expectations for extinctions were valid, but anomalous values for extirpated species alone are difficult to explain. The second option implies a non-linear true relationship, with the thermal bias values for extirpated species being valid but there being more extinctions observed than expected given their thermal bias. Given the anoxia of northern waters of the north-western Tethys, especially during the T-OAE (discussed below)<sup>26</sup>, we suspect the anoxia overruled temperature in habitat suitability, leaving species going extinct with unusually low observed thermal bias values. Poor dispersal capabilities and/or dispersal barriers can lead to a species’ failure to lessen its population thermal biases by shifting distributions, thereby shrinking its geographical range<sup>48</sup>, and making it vulnerable to global extinction<sup>45,49</sup>. Mechanisms of and limitations to habitat tracking should be explored during other intervals with changes in climate, sea level, and geography<sup>e.g. 49</sup>.
+
+## Overriding effects of anoxic waters and terrestrial runoff
+
+The well-sampled, oceanic-influenced regional clusters of north and east of Iberia best supported thermal determination of assemblage membership, where any deoxygenation prior to the T-OAE<sup>50</sup> apparently did not preclude a signal of thermal bias. Analyses of well-oxygenated environments such as outcrops from the south-west of Europe implicate Early Toarcian warming as the main regional driver of species loss, changes in bivalve-brachiopod assemblage structure, and their body size<sup>39,40,51</sup>. During peak T-OAE (‘warming 2’ into ‘transitional’ phases), support for thermal determination of assemblage membership dwindled in the Germanic and British regions alongside the number of occurrences. Although aquatic deoxygenation can amplify the influence of warming on ectotherm performance<sup>25,52</sup>, bottom water anoxia is likely to supersede the ecological influence of increased temperature. Several regions during the T-OAE are characterized by black shale deposition, where hypoxic and anoxic waters have long been associated with faunal turnover and extinctions<sup>26</sup>. Accordingly, benthic macrofaunal recovery only began after seafloor ventilation resumed, and remained incomplete in the British region by the end of our study<sup>53</sup>. During the T-OAE, the northern waters may have essentially been unavailable as habitat for species tracing their thermal niche. This may exemplify how species ranges can be compressed as they trace thermal preferences. Although fully marine (see Table S8), the more restricted northern waters likely had greater terrestrial influence, such that bottom-water anoxia was probably dependent on productivity, as nutrients were delivered from warming-enhanced weathering<sup>54</sup>, rather than simply temperature-dependent deoxygenation. The HadCM3 model estimated slightly lower salinity in the Germanic and British basins<sup>also 55</sup>, ranging between 33.3 and 34.6ppt across scenarios, than the other regions, while salinity was always highest east of Iberia, ranging between 34 and 35.6ppt (see SM). The semi-enclosed setting, especially of the Germanic and British basins, also likely increased the influence of local processes that global models are unlikely to capture, with the reality likely being warmer and more seasonal than estimated by our models<sup>56</sup>. Alongside changes in sea-level-dependent seafloor ventilation<sup>53</sup>, water density differences from freshwater input may have also encouraged stratification<sup>55,57</sup>. Enhanced capacity to model biochemical processes and extract variables, such as oxygen levels, should expand the ability of predictive models to account for additional niche requirements. While modern oxygen minimum zones continue to spread<sup>24</sup>, our results show how regional-scale processes can complicate the predictability of assemblage responses to temperature change.
+
+Besides temperature and salinity, other habitat requirements for a benthic species include suitable water depth and substrate conditions, which also dictate the conditions under which a species can be sampled. The northern regions were the only ones dominated by siliciclastic substrates, which could have blocked the immigration (alongside anoxia, see previous paragraph) of carbonate-affinity species. The largest and most consistent non-temperature change occurred at the Spinatum-Tenuicostatum transition, when substantial sea-level rise<sup>33</sup> led to increases in the frequency of deep habitat occurrences from 20–50% to 90–96% regional share. However, species thermal bias remained highly significantly associated with its occupancy response through different statistical treatments to explore the importance of this spatiotemporal scenario (see Tables S4 and S5). Being 100s km across, our regions tended to cover substantial substrate and depth variation, such that finer scale analyses may be needed to detect the influence of non-temperature habitat variables. Our focus on two-timer species also emphasised longer-term changes of the more common and better-preserved species, of which our analyses support temperature change being a key driver at broad spatial scales.
+
+## Temporal and spatial scaling
+
+Temporal and spatial resolutions in our study were ~1 million years and ~2000 km respectively, which need appreciation to compare our results with other studies. Finer scale variations were averaged out, such as the warming at the Pliensbachian–Toarcian stage-boundary<sup>58</sup>, although permanent ecological changes such as extinctions from short-term pulses remain. Despite the myriad of factors influencing a species occurrence at fine spatial scales, climate is expected to be one of the dominating factors at broader scales<sup>15,59</sup>. Significant effects of thermal bias have been assessed for modern assemblages at spatial scales from surveyed sites<sup>11</sup> to biogeographic ‘ecoregions’, more similarly sized to our regions<sup>28,59</sup>. At intermediate spatial scales, Flanagan et al.<sup>31</sup> found larger thermal biases of fish assemblages over decadal scales than inter-annual scales, which might encourage expectations that marine communities rapidly maintain equilibrium with temperature, despite evidence often to the contrary<sup>31</sup>. At much longer time scales and with spatially coarse temperature estimates, our data also supported equilibrium between assemblage mean thermal optima (CTI) and environment temperatures (Fig. S5). However, geographical context affected observations of thermal equilibrium in a study of planktonic foraminifera over thousands of years: mid latitude assemblages tracked climate change by turnover, but decreasing assemblage turnover at high latitudes under warming and low latitudes under cooling accumulated assemblage thermal bias<sup>23</sup>. Regions of high climate velocity, such as the tropics and poles, are likely to demand faster species’ niche-tracking than lower climate velocity regions, which is more likely to push populations of multiple species nearer to their thermal niche edges<sup>45</sup>. However, increasing thermal bias may only increase extirpations and extinctions when changes exceed species recent climatic experience<sup>47</sup>. Temporal resolution is not a problem per se for the application of paleontological insights to modern issues<sup>17</sup>, but limits the mechanisms for which we can observe evidence. Future work should be directed to understanding the mechanisms underlying observed palaeontological patterns and the transferability of those mechanisms to modern climate change and the current biodiversity crisis<sup>17</sup>.
+
+Based on our results for the northwestern Tethys, we expect regional species extirpations and especially immigrations to be already considerable (~5% and ~26%, respectively) at +3°C warming, such as forecast under high emissions scenarios (RCP 8.5) by the end of the century for the North Sea<sup>35</sup>. These extirpation and immigration values are similar to projections of a paleo-validated biodiversity model for the shelf seas of Europe by 2100<sup>60</sup>. Although an application of our results to modern warming ignores the very different time scales (=observed rates of change), the loss of a species’ thermally suitable habitat can respond directly to the magnitude of warming, regardless of the rate of warming, such as supported by empirical patterns of high latitude extinctions during hyperthermal events<sup>45</sup>. Rates of ancient climate changes may have been sufficiently slow for most species to track habitat availability but the extremely rapid anthropogenic rates of change are likely to divide response severity between species with greater and lesser dispersal abilities<sup>45</sup>. This may be especially the case in the tropics where climate velocities are highest<sup>61</sup>, leaving paleobiological extrapolations most likely as underestimates.
+
+Rare species, both range-restricted or locally uncommon, are unlikely to make it into the fossil record and thereby into our analysis. If rare species are at higher extinction and extirpation risk or tend to have narrower thermal tolerances, the overall magnitude of assemblage change including rare species can be expected to be higher than we predict. Again, this implies that inferences based on paleobiology will tend to give underestimates of whole community responses.
+
+# Conclusions
+
+We show a distinct relationship between the thermal suitability of Jurassic benthic species for its occupied region and its occupancy changes in that region during warming, which aggregated to substantial assemblage-level responses. Species thermal bias provides more information than the magnitude of regional warming alone and thus can be a stronger predictor of species extirpation, persistence, or immigration. Temperature-focused models may be less effective at finer (more local) spatial scales, where additional habitat variables may become more important, and in semi-enclosed coastal waters, which may be more inclined to anoxia. Predictions may be further refined by species-specific modelling and using climate models that handle processes at regional or finer scales, such as tidal mixing, where permitted by reliable, high resolution paleogeographic reconstruction. Our results support that greater magnitudes of warming tend to increase the cost of disequilibria between species and climate, increasing the rate of extirpation and extinction, especially if thermal habitat loss is not replaced elsewhere. Meanwhile, ambient warming was most clearly linked to species immigrations. Given potentially unprecedented modern rates of global warming<sup>62,63</sup>, paleobiology likely presents conservative warnings of future changes in marine species’ regional occupancy.
+
+# Materials and Methods
+
+## Study interval and region
+
+We focus on the climate changes from the cool Late Pliensbachian to the warm Early and Middle Toarcian (Early Jurassic), covering the hyperthermal Toarcian Ocean Anoxic Event (T-OAE), when some ocean basins became anoxic. We used the finest regionally-consistent temporal resolution for our occurrence data, the ammonite zone (the Serpentinum Zone was further split into Exaratum and Falciferum subzones; Table 2; mean 1.1 myr), at which the literature on temperature proxies was used to estimate local climates (particularly Müller et al. 64 and Ullmann et al. 65; more detail in SM Methods, ‘Climate conditions of our major time steps’). After the cool, low-CO₂ Late Pliensbachian Margaritatus and Spinatum zones, the Early Toarcian was associated with the release of greenhouse gases from the intense volcanism of the Karoo-Ferrar magmatic province 66–68. Emplacement of the Karoo-Ferrar large igneous province occurred over ~9 million years between 183.4 and 176.8 Ma, with bulk magmatism occurring from ~183.4 to ~183.0 Ma, coinciding with the T-OAE 69. Note that we consider the T-OAE to be equivalent in time to the well-known negative excursion of carbon isotopes (see Erba et al. 70 for discussion and alternative definitions). Analyses of thallium isotopes suggest that global marine deoxygenation of ocean water started sooner 50, alongside rapid, short-lived warming across the Pliensbachian/Toarcian boundary 66 at ~184 Ma 71. The Tenuicostatum Zone of the earliest Toarcian remained on average warmer than the Late Pliensbachian. Further warming in the T-OAE proper of the Exaratum subzone, possibly as the consequence of a rapid release of thermogenic and/or biogenic methane adding to the volcanic CO₂ release, is associated with the main extinction phase 72,73. After this peak of warming and CO₂ concentrations, the Falciferum Zone represents a transitional climate, starting warm but later cooling to a level warmer than the Tenuicostatum Zone 64, which is maintained into the Bifrons Zone.
+
+Our regional focus follows a roughly north-south trending oceanic transect from Scotland via the western European epicontinental sea to north-western Tethys including Morocco, Tunisia, and Algeria (Fig. 1). Terrestrial influence (nutrients, turbidity, freshwater input) was higher in northern, more restricted water bodies, especially the Cleveland Basin 26, with less mixing and less oxygenation of bottom waters 55,74. This is particularly expressed during the Exaratum subzone (T-OAE proper) when sites in England and Germany are dominated (though not completely) by hypoxic to anoxic sediments, while other basins were less affected by deoxygenation.
+
+## Seawater temperature maps
+
+CO₂ scenarios per ammonite zone were either allocated directly, where CO₂ estimates were available (Tenuicostatum and Exaratum sub/zones) 33,72, or indirectly based on approximating relative temperature change estimates, especially Müller et al. 64 and Ullmann et al. 65, which together traced relative temperature change via oxygen isotopes over our whole temporal duration. Temperature changes output by the CLIMBER-X climate model were then checked against proxy temperature changes at the appropriate paleocoordinates and water depth (see SM). Secondary CO₂ scenarios were based on maximum possible temperature changes (Table 2; see SM section ‘Climate conditions of our major time steps’ for a wider discussion of the evidence).
+
+We ran equilibrium climate simulations at fixed pCO₂ scenarios using the CLIMBER-X Earth-system model 34. CLIMBER-X is particularly useful as a fast and flexible paleoclimate model and provides simulated temperatures in the ocean and atmosphere on a 5°x5° horizontal grid, among other parameters. Early Jurassic boundary conditions were represented by a reduced solar constant (1340.5 W/m²). For the model paleogeography, we used the bathymetric topography of Kocsis & Scotese 36, which matched the coastline to marine occurrences in the Paleobiology Database (see below), primarily using the Toarcian map (180 Ma) and secondarily using the Pliensbachian (185 Ma). Deep seafloor depth was set to -3700m, shallow marine / continental shelf to -200m, and land to +200m. Local shelf features are not well represented in these reconstructions and the coarse resolution model results are not expected to be perfect, but we expect the derived niche estimates to be better than a simple dependence on paleolatitude. We also downloaded the sea surface temperature maps simulated with the HadCM3 model, though these were limited to CO₂ scenarios of 560 and 950 ppm 75. Despite being affected by similar limitations, HadCM3 is a more complex and highly resolved model than CLIMBER-X, and its outputs were used as a benchmark. This supported the upscaling of the July mean temperature maps from CLIMBER-X to the finer spatial resolution of the HadCM3 maps via bilinear interpolation. In general, correlations between the two models were high (Rho ≥ 0.8) with a root mean square error (RMSE) that increased, as expected, as the modelled CO₂ scenarios deviated (see SM section ‘Climate model (dis)agreement’, Tables S9, S10, Figs. S6, S7). While CLIMBER-X has an equilibrium climate sensitivity close to the best estimate of 3°C per pCO₂ doubling 2, the HadCM3 model is more sensitive and generally yields higher temperatures at elevated CO₂ levels.
+
+**Table 2**  
+CO₂ scenarios used for modelling climates over time steps of ammonite zones, from late Pliensbachian (Margaritatus Zone) into Middle Toarcian (Bifrons Zone). See SM for more detail on determining the CO₂ scenarios.
+
+| Ammonite (sub)zone | Main CO₂ scenario (ppm) | Secondary CO₂ scenario (ppm) | Notes |
+|--------------------|--------------------------|-------------------------------|-------|
+| Bifrons            | 750                      | 750                           | Outputs should be warmer than Tenuicostatum* |
+| Falciferum         | 750                      | 1000                          | Outputs should be warmer than Tenuicostatum* |
+| Exaratum           | 1000                     | 1250, 1500                    | 1000 as low estimate. 1500 as peak estimate |
+| Tenuicostatum      | 500                      | 500                           | Secondary scenario with Pliensbachian map |
+| Spinatum           | 400                      | 300                           | 300 as cold estimate |
+| Margaritatus       | 400                      | 300                           | 300 as cold estimate |
+
+*Following oxygen isotopes covering our temporal range in Müller et al. 64 or Ullmann et al. 65.
+
+## Species occurrence data
+
+On 24th May 2022, we downloaded marine-only occurrences of bivalves, gastropods and brachiopods from the Paleobiology Database (PaleoDB, https://paleobiodb.org/), representing benthic assemblages, and binned them to stratigraphic stages using R package divDyn 76. Our analyses used species-level occurrences, but occurrences initially had to be accepted at least at the genus level. They also required modern geographical coordinates, which were used for paleogeographical rotation and for isolating north-west Tethys occurrences by a bounding box around modern Europe, east-west from Turkey to Portugal, and north-south from Scotland to the Mediterranean coast of Africa. Confidently identified species names that were taxonomically unaccepted by the PaleoDB underwent automatic checks for spelling mistakes. Of these, persistent unaccepted species names of the Pliensbachian and Toarcian were then taxonomically vetted by M. Aberhan to catch more accepted species occurrences and prevent artifacts in geographic distribution patterns, such as synonymous species names. To achieve ammonite (sub-)zone temporal resolution, we explored PaleoDB download columns ‘early_interval’, ‘zone’, and ‘stratcomments’ for temporal resolution information, especially ammonite zone or subzone allocation. Some data-rich entered references were investigated manually for lacking temporal, paleoenvironmental or lithological information (see R code in SM). A separate, global dataset was used to establish species’ First and Last Appearance Dates (FADs and LADs), ideally at ammonite (sub-)zone resolution, within the Pliensbachian and Toarcian stages.
+
+Determination of species’ thermal preferences may be confounded if species have significant affinities for particular substrate or bathymetric paleoenvironments. Substrate or bathymetric categories were combined using the keys in divDyn, before environmental affinities were tested for using binomial tests with alpha = 0.1 (function ‘affinity’ in divDyn) 76.
+
+Temperature estimates were sampled per taxon occurrence from modeled seawater temperature paleogeographical maps from 180 Ma (Toarcian, primary scenario) and 185 Ma (Pliensbachian, secondary scenario) separately. This avoided switching between maps in the same analytical time series, which could result in a sudden, artificial shift in paleocoordinates and influencing the thermal bias. Accordingly, we reconstructed coordinates and coastlines using the *rgplates* interface 77 to Gplates v2.3 78 to both Toarcian and Pliensbachian rotations as separate columns, based on the PaleoMAP model 36.
+
+## Spatial clusters
+
+Spatial clusters of sampling, or ‘regions’, were expected to be more similar in mean temperature and species composition within than among clusters per time zone. The species recorded in each of these clusters per time zone then became the spatiotemporally cohesive ‘assemblage’ of interest (analogous to quantification of thermal bias for spatiotemporally cohesive sampling transects in 11). Collections were pooled into unique spatial coordinates per time zone. Objective and non-overlapping clusters were identified using hierarchical clustering of Euclidean distance matrices of occurrence paleocoordinates of all time zones pooled. We expected these clusters to arise mainly from sampling patterns, given that they use no ecological data, but separate assemblages should ideally be ecologically distinct, having more differences between them than within them. To assess ecological similarity among the clusters defined by Euclidean distance of coordinates, we also estimated groupings of late Pliensbachian occurrences by hierarchical clustering of Jaccard distance matrices of species presence/’absence’ (i.e. using ecological co-occurrence but ignoring spatial coordinates). Jaccard distance clusters with less than 14 species were removed to balance the tendency of small samples to drive dissimilarity (via species absences) against persistent and more relevant, larger groupings. Ecological clusters validated the use of the separately identified spatial clusters as distinct species assemblages, such as from separate bodies of water or habitat. Adopting ten spatial clusters maximised the agreement between the two approaches.
+
+Finally, practical requirements for spatial clusters included (1) being sampled in different time steps, ideally throughout, and (2) having sufficient occurrences. This was the case for four of the ten spatial clusters: the northern and most likely terrestrially influenced British basins cluster, and three clusters surrounding the landmass of Iberia: to the west, to the north, and to the east (likely to be the most pelagic influenced cluster). The benthic fauna of a fifth, Germanic cluster were well-sampled in the late Pliensbachian, but not in the early Toarcian. However, its outcrops are exposed throughout our temporal focus, suggesting that species absences were driven by anoxic bottom waters rather than by poor sampling, so this cluster was also used for analysis. Clusters had different thermal regimes (see Results) and variables like terrestrial influence (see Discussion).
+
+## Rates of species responses
+
+As a precaution against spurious features of sampling patterns (see Fig. S8), we focus on comparing numbers of regional two-timer species, that is, species that were observed in a region for at least two time bins consecutively (Fig. 5) 79. These are the better-sampled species, whose observed responses may be more reliable. The same can be done using three-timers (species must be observed in a region for three time bins consecutively; see Supplementary Methods, Fig. S9, and Supplementary Results). However, using two-timers has the advantage that the temporal focus of change is a single boundary between two time bins, which fits understanding of the timing of the climatic changes investigated here, rather than change over a central bin and both of its demarcating boundaries for three-timers. The well-sampled nature of two-timers and high sampling completeness of the focal ammonite (sub)zones of European regions for this interval means the observed times of extinction, extirpation, immigration or origination are relatively reliable (e.g. against Signor-Lipps effect).
+
+Focussing on cluster two-timers (Fig. 5), immigrating species were those observed in the cluster in time i + 1 AND time i + 2 but not in i. Originating species were the same but also had their dataset-wide First Appearance Date (FAD) in time i + 1. Extirpated species were those similarly observed in the cluster in time i AND time i – 1 but not in i + 1, with those having time i as their Last Appearance Date (LAD) were classed as going extinct. Persisting species were observed in the cluster in times i AND time i + 1. There were fewer occurrences before the first time bin (i.e. in the Davoei zone, which preceded Margaritatus) and after the last bin (in the Variabilis zone, which followed Bifrons), limiting the quantity of cluster two-timer species, so their two-timers were simply required to have a presence in times i – 1 and i + 2, respectively, regardless of spatial cluster. Species still needed a cluster occurrence around the focal boundary, either in time i or i + 1, to be assigned a response category (e.g. extirpated), so this step did not artificially increase numbers of species in any response category, but simply allowed more species to pass the sampling threshold in the earliest and latest time bins. Note that in all cases, due to incomplete sampling, extirpation and immigration are probabilistic events rather than definite.
+
+Two-timers without a LAD or FAD have occurrences in the future and past, respectively, of time i, such that their species thermal niche is averaged over past and future distributions. Meanwhile, the thermal niches of extinct and originating species were inherently limited to only past or only future distributions, respectively. To address the potential criticism of extinct and originating species having a fixed thermal niche, we focus our analysis on extirpation, immigration and persistence responses, and only secondarily including extinctions and originations.
+
+## Analysing assemblage temporal change
+
+Analyses were separate between species and assemblage levels. A species’ thermal bias was defined as the difference between the cluster median temperature for a time zone and the species’ thermal median (temperatures averaged over all zone-level occurrences of the species from the Margaritatus to Bifrons zones, the complete interval when occurrences were matched to temperature maps). An assemblage thermal bias, often assumed to indicate net vulnerability, was thus the difference between the median of the constituent species’ thermal medians 11,27 and the cluster median temperature for a time zone.
+
+We expected an escalatory order of occupancy responses relative to thermal bias in a warming scenario (Fig. 5), with extinct and extirpated species at one extreme having relatively cool thermal biases, originating or immigrating species at the other extreme having relatively warm thermal biases, and persisting species having relatively intermediate thermal biases. Species-level regressions therefore used species occupancy response as an ordered continuous dependent variable and species thermal bias as a continuous independent variable, Occupancy_response ~ Thermal_bias. Mixed effects accounted for the nested analysis structure, where necessary, with species nested within clusters and clusters nested within time zones (i.e. a single species can have one response and thermal bias per cluster per time zone, a single cluster can occur in multiple time zones). To guard against criticism that originating and extinct species’ thermal niches were pre-decided (e.g. since species going extinct in time i can only have occurrences in the past relative to time i, when climates tended to be relatively colder in our study), we compare regression results with extinction or origination responses left out vs. included. Being at the extremes of the regression line, species originating or going extinct also have a stronger effect on the regression slope than persisting, extirpated (but surviving) or immigrating species (with past occurrences). To assess how much the observed thermal bias values for the different species occupancy responses deviated from linear expectations, the above regression equation was reversed into, Thermal_bias ~ Occupancy_response, to calculate thermal bias confidence intervals.
+
+For assemblage-level analyses, we recorded the percentage of a current assemblage that was categorized at the species escalatory response levels of persisting, extirpated, or extinct, and the percentage of a new assemblage that was categorized as immigrating or originating. The turnover of the current assemblage into the new assemblage (i.e. from time i to i + 1) was also quantified by Jaccard distance. These were each used separately as dependent variables. Independent variables were assemblage thermal bias, regional temperature change magnitude, or the difference in occurrence proportions of occurrences from carbonate or offshore substrates (the most frequent substrate types). Here, mixed effect models nested clusters within time zones, but had a low sample size (5 clusters x 5 time zones = 25 assemblage data points maximum) and thus a weaker potential for inference. These regressions were applied in an exploratory framework akin to a correlation matrix to weigh evidence for further research. Models using assemblage thermal bias as an independent variable were inverse weighted for the standard deviation of species’ thermal bias. We chose to apply these model expectations for regional species responses at a modern-relevant level of warming (+3°C). All analyses were performed in R 80 with packages *divDyn* v0.8.2, *corrplot* v0.92 to present assemblage-level analyses, *nlme* v3.1-162 for mixed effects models, *vegan* v2.6-4 for clustering 76,81,82.
+
+# References
+
+1. Tittensor, D. P. et al. Global patterns and predictors of marine biodiversity across taxa. *Nat. 2010 4667310* **466**, 1098–1101 (2010).
+2. IPCC. *Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change*. (Cambridge University Press, 2021). doi:10.1017/9781009157896.
+3. Burrows, M. T. et al. The pace of shifting climate in marine and terrestrial ecosystems. *Science* **334**, 652–655 (2011).
+4. Poloczanska, E. S. et al. Global imprint of climate change on marine life. *Nat. Clim. Chang.* **3**, 919–925 (2013).
+5. Bindoff, N. L. et al. Changing ocean, marine ecosystems, and dependent communities. in *IPCC special report on the ocean and cryosphere in a changing climate* 477–587 (IPCC, 2019).
+6. Chaudhary, C., Richardson, A. J., Schoeman, D. S. & Costello, M. J. Global warming is causing a more pronounced dip in marine species richness around the equator. *Proc. Natl. Acad. Sci. U. S. A.* **118**, (2021).
+7. Pecl, G. T. et al. Biodiversity redistribution under climate change: Impacts on ecosystems and human well-being. *Science* **355**, eaai9214 (2017).
+8. IPCC. *Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change*. (Cambridge University Press, 2013).
+9. Newbold, T. Future effects of climate and land-use change on terrestrial vertebrate community diversity under different scenarios. *Proc. R. Soc. B* **285**, (2018).
+10. Bates, A. E. et al. Defining and observing stages of climate-mediated range shifts in marine systems. *Glob. Environ. Chang.* **15**, 351–389 (2014).
+11. Day, P. B., Graham, R. D. S. & Amanda, J. E. Species’ thermal ranges predict changes in reef fish community structure during 8 years of extreme temperature variation. *Divers. Distrib.* **24**, 1036–1046 (2018).
+12. Smale, D. A. & Wernberg, T. Extreme climatic event drives range contraction of a habitat-forming species. *Proc. R. Soc. B Biol. Sci.* **280**, (2013).
+13. Román-Palacios, C. & Wiens, J. J. Recent responses to climate change reveal the drivers of species extinction and survival. *Proc. Natl. Acad. Sci. U. S. A.* **117**, 4211–4217 (2020).
+14. Lenoir, J. & Svenning, J. C. Climate-related range shifts - a global multidimensional synthesis and new research directions. *Ecography* **38**, 15–28 (2015).
+15. Bonachela, J. A., Burrows, M. T. & Pinsky, M. L. Shape of species climate response curves affects community response to climate change. *Ecol. Lett.* **24**, 708–718 (2021).
+16. Sunday, J. M., Bates, A. E. & Dulvy, N. K. Thermal tolerance and the global redistribution of animals. *Nat. Clim. Chang.* **2**, 686–690 (2012).
+17. Kiessling, W., Smith, J. A. & Raja, N. B. Improving the relevance of paleontology to climate change policy. *Proc. Natl. Acad. Sci. U. S. A.* **120**, e2201926119 (2023).
+18. Foster, G. L., Hull, P., Lunt, D. J. & Zachos, J. C. Placing our current ‘hyperthermal’ in the context of rapid climate change in our geological past. *Philos. Trans. A. Math. Phys. Eng. Sci.* **376**, 20170086 (2018).
+19. Song, H. et al. Thresholds of temperature change for mass extinctions. *Nat. Commun.* **12**, 1–8 (2021).
+20. Powell, M. G., Moore, B. R. & Smith, T. J. Origination, extinction, invasion, and extirpation components of the brachiopod latitudinal biodiversity gradient through the Phanerozoic Eon. *Paleobiology* **41**, 330–341 (2015).
+21. Reddin, C. J., Kocsis, Á. T. & Kiessling, W. Marine invertebrate migrations trace climate change over 450 million years. *Glob. Ecol. Biogeogr.* **27**, 704–713 (2018).
+22. Kiessling, W., Simpson, C., Beck, B., Mewis, H. & Pandolfi, J. M. Equatorial decline of reef corals during the last Pleistocene interglacial. *Proc. Natl. Acad. Sci. U. S. A.* **109**, 21378–83 (2012).
+23. Mathes, G. H. et al. (in review) Fossil planktonic foraminifera responses to climate change vary by latitude. *Glob. Ecol. Biogeogr.*
+24. Breitburg, D. et al. Declining oxygen in the global ocean and coastal waters. *Science* **359**, eaam7240 (2018).
+25. Reddin, C. J., Nätscher, P. N., Kocsis, Á. T., Pörtner, H. O. & Kiessling, W. Marine clade sensitivities to climate change conform across time scales. *Nat. Clim. Chang. Chang.* **10**, 249–253 (2020).
+26. Danise, S., Twitchett, R. J., Little, C. T. S. & Clémence, M. E. The Impact of Global Warming and Anoxia on Marine Benthic Community Dynamics: An Example from the Toarcian (Early Jurassic). *PLoS One* **8**, (2013).
+27. Devictor, V., Julliard, R., Couvet, D., Jiguet, F. & B, P. R. S. Birds are tracking climate warming, but not fast enough. *Proc. R. Soc. B Biol. Sci.* **275**, 2743–2748 (2008).
+28. Stuart-Smith, R. D., Edgar, G. J., Barrett, N. S., Kininmonth, S. J. & Bates, A. E. Thermal biases and vulnerability to warming in the world’s marine fauna. *Nature* **528**, 88–92 (2015).
+29. Reddin, C. J., Kocsis, Á. T., Aberhan, M. & Kiessling, W. Victims of ancient hyperthermal events herald the fates of marine clades and traits under global warming. *Glob. Chang. Biol.* **27**, 868–878 (2021).
+30. Little, C. T. S. & Benton, M. J. Early Jurassic mass extinction: A global long-term event. *Geology* **23**, 495–498 (1995).
+31. Flanagan, P. H., Jensen, O. P., Morley, J. W. & Pinsky, M. L. Response of marine communities to local temperature changes. *Ecography* **42**, 214–224 (2019).
+32. Ruebsam, W. & Schwark, L. Impact of a northern-hemispherical cryosphere on late pliensbachian–early toarcian climate and environment evolution. *Geol. Soc. Spec. Publ.* **514**, 359–385 (2021).
+33. Nordt, L., Breecker, D. & White, J. Jurassic greenhouse ice-sheet fluctuations sensitive to atmospheric CO2 dynamics. *Nat. Geosci.* **15**, 54–59 (2022).
+34. Willeit, M., Ganopolski, A., Robinson, A. & Edwards, N. R. The Earth system model CLIMBER-X v1.0 – Part 1: Climate model description and validation. *Geosci. Model Dev.* **15**, 5905–5948 (2022).
+35. Cornes, R. C. et al. *Climate change impacts on temperature around the UK and Ireland. MCCIP Science Review 2023*. (2023) doi:10.14465/2022.reu08.tem.
+36. Kocsis, Á. T. & Scotese, C. R. Mapping paleocoastlines and continental flooding during the Phanerozoic. *Earth-Science Rev.* **213**, 103463 (2021).
+37. Ruhl, M. et al. Astronomical constraints on the duration of the Early Jurassic Pliensbachian Stage and global climatic fluctuations. *Earth Planet. Sci. Lett.* **455**, 149–165 (2016).
+38. Zurell, D., Fritz, S. A., Rönnfeldt, A. & Steinbauer, M. J. Predicting extinctions with species distribution models. *Cambridge Prism. Extinction* **1**, e8 (2023).
+39. Piazza, V., Ullmann, C. V. & Aberhan, M. Temperature-related body size change of marine benthic macroinvertebrates across the Early Toarcian Anoxic Event. *Sci. Rep.* **10**, 1–13 (2020).
+40. Piazza, V., Duarte, L. V., Renaudie, J. & Aberhan, M. Reductions in body size of benthic macroinvertebrates as a precursor of the early Toarcian (Early Jurassic) extinction event in the Lusitanian Basin, Portugal. *Paleobiology* **45**, 296–316 (2019).
+41. Pörtner, H.-O., Bock, C. & Mark, F. C. Oxygen- and capacity-limited thermal tolerance: bridging ecology and physiology. *J. Exp. Biol.* **220**, 2685–2696 (2017).
+42. Deutsch, C., Ferrel, A., Seibel, B., Pörtner, H.-O. & Huey, R. Climate change tightens a metabolic constraint on marine habitats. *Science* **348**, 1132–1136 (2015).
+43. Clark, M. S. et al. Biodiversity in marine invertebrate responses to acute warming revealed by a comparative multi-omics approach. *Glob. Chang. Biol.* **23**, 318–330 (2017).
+44. Penn, J. L., Deutsch, C., Payne, J. L. & Sperling, E. A. Temperature-dependent hypoxia explains biogeography and severity of end-Permian marine mass extinction. *Science* **362**, eaat1327 (2018).
+45. Reddin, C. J., Aberhan, M., Raja, N. B. & Kocsis, Á. T. Global warming generates predictable extinctions of warm- and cold-water marine benthic invertebrates via thermal habitat loss. *Glob. Chang. Biol.* **28**, 5793–5807 (2022).
+46. Penn, J. L. & Deutsch, C. Avoiding ocean mass extinction from climate warming. *Science (80-. ).* **376**, 524–526 (2022).
+47. Mathes, G. H., van Dijk, J., Kiessling, W. & Steinbauer, M. J. Extinction risk controlled by interaction of long-term and short-term climate change. *Nat. Ecol. Evol.* **5**, 304–310 (2021).
+48. Payne, J. L. & Finnegan, S. The effect of geographic range on extinction risk during background and mass extinction. *Proc. Natl. Acad. Sci.* **104**, 10506–10511 (2007).
+49. Saupe, E. E. et al. Extinction intensity during Ordovician and Cenozoic glaciations explained by cooling and palaeogeography. *Nat. Geosci.* **13**, 65–70 (2020).
+50. Them, T. R. et al. Thallium isotopes reveal protracted anoxia during the Toarcian (Early Jurassic) associated with volcanism, carbon burial, and mass extinction. *Proc. Natl. Acad. Sci. U. S. A.* **115**, 6596–6601 (2018).
+51. Gómez, J. J. & Goy, A. Warming-driven mass extinction in the Early Toarcian (Early Jurassic) of northern and central Spain. Correlation with other time-equivalent European sections. *Palaeogeogr. Palaeoclimatol. Palaeoecol.* **306**, 176–195 (2011).
+52. Götze, S. et al. Single and combined effects of the “Deadly trio” hypoxia, hypercapnia and warming on the cellular metabolism of the great scallop Pecten maximus. *Comp. Biochem. Physiol. Part - B Biochem. Mol. Biol.* **243**–**244**, 110438 (2020).
+53. Atkinson, J. W., Little, C. T. S. & Dunhill, A. M. Long duration of benthic ecological recovery from the early Toarcian (Early Jurassic) mass extinction event in the Cleveland Basin, UK. *J. Geol. Soc. London.* **180**, (2023).
+54. Krencker, F. N. et al. Two-phased collapse of the shallow-water carbonate factory during the late Pliensbachian–Toarcian driven by changing climate and enhanced continental weathering in the Northwestern Gondwana Margin. *Earth-Science Rev.* **208**, 103254 (2020).
+55. Bjerrum, C. J., Surlyk, F., Callomon, J. H. & Slingerland, R. L. Numerical paleoceanographic study of the Early Jurassic Transcontinental Laurasian Seaway. *Paleoceanography* **16**, 390–404 (2001).
+56. Judd, E. J., Bhattacharya, T. & Ivany, L. C. A Dynamical Framework for Interpreting Ancient Sea Surface Temperatures. *Geophys. Res. Lett.* **47**, (2020).
+57. Dera, G. & Donnadieu, Y. Modeling evidences for global warming, Arctic seawater freshening, and sluggish oceanic circulation during the Early Toarcian anoxic event. *Paleoceanography* **27**, (2012).
+58. Suan, G. et al. Secular environmental precursors to Early Toarcian (Jurassic) extreme climate changes. *Earth Planet. Sci. Lett.* **290**, 448–458 (2010).
+59. Devictor, V. et al. Differences in the climatic debts of birds and butterflies at a continental scale. *Nat. Clim. Chang.* **2**, 121–124 (2012).
+60. Beaugrand, G., Edwards, M., Raybaud, V., Goberville, E. & Kirby, R. R. Future vulnerability of marine biodiversity compared with contemporary and past changes. *Nat. Clim. Chang.* **5**, 695–701 (2015).
+61. Garciá Molinos, J. et al. Climate velocity and the future global redistribution of marine biodiversity. *Nat. Clim. Chang.* **6**, 83–88 (2015).
+62. Hönisch, B. et al. The geological record of ocean acidification. *Science* **335**, 1058–1063 (2012).
+63. Foster, G. L., Royer, D. L. & Lunt, D. J. Future climate forcing potentially without precedent in the last 420 million years. *Nat. Commun.* **8**, 1–8 (2017).
+64. Müller, T. et al. Ocean acidification during the early Toarcian extinction event: Evidence from boron isotopes in brachiopods. *Geology* **48**, 1–5 (2020).
+65. Ullmann, C. V. et al. Warm afterglow from the Toarcian Oceanic Anoxic Event drives the success of deep-adapted brachiopods. *Sci. Rep.* **10**, (2020).
+66. Suan, G., Mattioli, E., Pittet, B., Mailliot, S. & Lécuyer, C. Evidence for major environmental perturbation prior to and during the Toarcian (Early Jurassic) oceanic anoxic event from the Lusitanian Basin, Portugal. *Paleoceanography* **23**, (2008).
+67. Dera, G. et al. Climatic ups and downs in a disturbed Jurassic world. **39**, 215–218 (2011).
+68. Pálfy, J. & Smith, P. L. Synchrony between Early Jurassic extinction, oceanic anoxic event, and the Karoo-Ferrar flood basalt volcanism. *Geology* **28**, 747–750 (2000).
+69. Greber, N. D. et al. New high precision U-Pb ages and Hf isotope data from the Karoo large igneous province; implications for pulsed magmatism and early Toarcian environmental perturbations. *Results in Geochemistry* **1**, 100005 (2020).
+70. Erba, E. et al. Carbon-and oxygen-isotope signature of the Toarcian Oceanic Anoxic Event: insights from two Tethyan pelagic sequences (Gajum and Sogno Cores – Lombardy Basin, northern Italy). *Newsletters Stratigr.* **55**, 451–477 (2022).
+71. Ait-Itto, F. Z., Martinez, M., Price, G. D. & Ait Addi, A. Synchronization of the astronomical time scales in the Early Toarcian: A link between anoxia, carbon-cycle perturbation, mass extinction and volcanism. *Earth Planet. Sci. Lett.* **493**, 1–11 (2018).
+72. McElwain, J. C., Wade-Murphy, J. & Hesselbo, S. P. Changes in carbon dioxide during an oceanic anoxic event linked to intrusion into Gondwana coals. *Nature* **435**, 479–482 (2005).
+73. Hesselbo, S. P. et al. Massive dissociation of gas hydrate during a Jurassic oceanic anoxic event. *Nature* **406**, 392–395 (2000).
+74. Van De Schootbrugge, B. et al. Enhanced Arctic-Tethys connectivity ended the Toarcian Oceanic Anoxic Event in NW Europe. *Geol. Mag.* **157**, 1593–1611 (2020).
+75. Valdes, P. J. et al. The BRIDGE HadCM3 family of climate models: HadCM3@Bristol v1.0. *Geosci. Model Dev.* **10**, 3715–3743 (2017).
+76. Kocsis, Á. T., Reddin, C. J. C. J., Alroy, J. & Kiessling, W. The R package divDyn for quantifying diversity dynamics using fossil sampling data. *Methods Ecol. Evol.* **10**, 735–743 (2019).
+77. Kocsis, Á. T. & Raja, N. B. rgplates: R interface for the GPlates web service and desktop application. at https://rdrr.io/cran/rgplates/ (2021).
+78. Müller, R. D. et al. GPlates: Building a Virtual Earth Through Deep Time. *Geochemistry, Geophys. Geosystems* **19**, 2243–2261 (2018).
+79. Alroy, J. Dynamics of origination and extinction in the marine fossil record. *Proc. Natl. Acad. Sci.* **105**, 11536–11542 (2008).
+80. R Development Core Team. *R: A language and environment for statistical computing*. (R Foundation for Statistical Computing, 2023).
+81. Pinheiro J, Bates D, DebRoy S, Sarkar D & R Core Team. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-131.1. at (2018).
+82. Oksanen J, Kindt R, Legendre P, O’Hara B & Stevens, M. The vegan package. Community ecology package. at http://cran.r-project.org/package=VEGAN (2007).
+
+# Supplementary Materials
+
+Supplementary Materials (Supplementary Figures, Supplementary Tables, Supplementary Methods and Supplementary Results) are not available with this version.
+
+# Supplementary Files
+
+- [rs.pdf](https://assets-eu.researchsquare.com/files/rs-3796284/v1/541499fd9ff86decf3d83409.pdf)  
+  Reporting Summary