Correction

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES, EVOLUTION Correction for “Flat latitudinal diversity gradient caused by the – mass extinction,” by Haijun Song, Shan Huang, Enhao Jia, Xu Dai, Paul B. Wignall, and Alexander M. Dunhill, which was first published July 6, 2020; 10.1073/pnas.1918953117 (Proc. Natl. Acad. Sci. U.S.A. 117, 17578–17583). The editors wish to note that the following competing interest statement regarding personal associations with the authors should have been disclosed: Editor M.J.B. shares a research grant with P.B.W. and has coauthored papers with A.M.D. in 2019 and 2020, and with H.S. in 2017 and 2019. The editors apologize for the oversight in failing to disclose this information to readers. The competing interest statement in the article has been corrected.

Published under the PNAS license. First published August 10, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.2015344117

20334 | PNAS | August 18, 2020 | vol. 117 | no. 33 www.pnas.org Downloaded by guest on September 28, 2021 Flat latitudinal diversity gradient caused by the Permian–Triassic mass extinction

Haijun Song (宋海军)a,1, Shan Huangb, Enhao Jia (贾恩豪)a, Xu Dai (代旭)a, Paul B. Wignallc, and Alexander M. Dunhillc aState Key Laboratory of Biogeology and Environmental Geology, School of Earth Sciences, China University of Geosciences, 430074 Wuhan, China; bSenckenberg Biodiversity and Climate Research Center, 60325 Frankfurt am Main, Germany; and cSchool of Earth and Environment, University of Leeds, LS2 9JT Leeds, United Kingdom

Edited by Michael J. Benton, University of Bristol, Bristol, UK, and accepted by Editorial Board Member Neil H. Shubin May 21, 2020 (received for review October 29, 2019) The latitudinal diversity gradient (LDG) is recognized as one of the of marine species (16) and caused profound temporary and most pervasive, global patterns of present-day biodiversity. How- permanent ecological restructuring of marine ecosystems (23), ever, the controlling mechanisms have proved difficult to identify which ultimately catalyzed the transformation of marine com- because many potential drivers covary in space. The geological munities dominated by Paleozoic faunas to those dominated by record presents a unique opportunity for understanding the mecha- the Modern fauna (15). The effect of the largest mass extinction nisms which drive the LDG by providing a direct window to deep-time in Earth’s history on global marine biogeography is largely un- biogeographic dynamics. Here we used a comprehensive database known, but the rich marine fossil record from this time can be a containing 52,318 occurrences of marine fossils to show that the shape powerful tool for illuminating the fundamental principles that of the LDG changed greatly during the Permian–Triassic mass extinc- shape global biodiversity. tion from showing a significant tropical peak to a flattened LDG. The LDG dynamics across the P-Tr mass extinction has received flat LDG lasted for the entire (∼5 My) before reverting to little attention except for a few case studies that have all sug- a modern-like shape in the . The environmental ex- gested a strong impact of environmental changes on global di- tremes that prevailed globally, especially the dramatic warming, likely versity patterns. For example, the early-middle Permian diversity induced selective extinction in low latitudes and accumulation of di- of terrestrial tetrapods reportedly peaked in temperate regions

versity in high latitudes through origination and poleward migration, as a result of profound climate-induced biome shift (24). By the EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES which combined together account for the flat LDG of the Early Triassic. Early Triassic, the distribution of terrestrial tetrapods had moved 10 to 15° poleward (25, 26), and the group was apparently absent biogeography | end-Permian mass extinction | global warming | ocean from 40 °S to 30 °N as a result of tropical overheating (18). anoxia | biodiversity Phylogenetic network analysis also found a marked increase in biogeographic connectedness, resulting in a more homogeneous he increase in species richness from the poles to the tropics, composition of diversity across latitudes in tetrapods during this long known as the latitudinal diversity gradient (LDG), is one time (27). Similar poleward migrations were also found in ma- T EVOLUTION of the most pervasive first-order biological patterns on Earth rine in the Northern Hemisphere during the Early today (1, 2), both on land (3) and in the oceans (4–6). Yet, the Triassic (28), and a variable LDG during this time was reported relative importance of the diverse ecological and evolutionary mechanisms (e.g., reviews in refs. 7–9) for generating this pattern Significance remains unclear. Paleontological data provide a unique per- spective in the search for the dominant driver(s) of LDGs, The deep-time dynamics of the latitudinal diversity gradient allowing diversity and distribution dynamics to be tracked (LDG), especially through dramatic events like mass extinctions, through the long history of environmental changes (10, 11). In can provide invaluable insights into the biotic responses to particular, climate (e.g., temperature or precipitation) is regar- global changes, yet they remain largely underexplored. Our ded as a key driver of the LDG, and its large-scale changes have study shows that the shape of marine LDGs changed sub- been postulated to have altered the general shape of the LDG stantially and rapidly during the Permian–Triassic mass ex- through time (12, 13). Steep, normal LDGs (i.e., with a signifi- tinction from a modern-like steep LDG to a flat LDG. The flat cant tropical peak like today) have been found primarily during LDG lasted for ∼5 My and was likely a consequence of the icehouse times, whereas bimodal or even reverse LDGs with extreme global environment, including extreme warming and diversity peaks at mid to high latitudes occurred during green- ocean anoxia, which ensured harsh conditions prevailing from house climates (13, 14). These findings call for assessments of the tropics to the poles. Our findings highlight the fundamental the relative importance of climate per se and environmental role of environmental variations in concert with severe bio- stability, especially through comparing the dynamics across sev- diversity loss in shaping the first-order biogeographic patterns. eral time intervals with different environmental templates. The dramatic changes in environmental conditions and the Author contributions: H.S., P.B.W., and A.M.D. designed research; H.S., E.J., and X.D. performed research; H.S., S.H., and X.D. analyzed data; and H.S., S.H., P.B.W., and severe mass extinction at the end of the Permian provide an A.M.D. wrote the paper. excellent opportunity for investigating LDGs and their control- The authors declare no competing interest. – ling mechanisms. The Permian Triassic (P-Tr) mass extinction, This article is a PNAS Direct Submission. M.J.B. is a guest editor invited by the which occurred ca. 252 My ago, was the largest extinction event Editorial Board. of the Phanerozoic (15, 16). This biological crisis was linked to Published under the PNAS license. extreme and prolonged environmental changes, many of which Data deposition: All data used to conduct analyses and plot figures are available for are probably the most serious of the past 500 My. The contem- download at Dryad, https://datadryad.org/stash/dataset/doi:10.5061/dryad.41ns1rn9z. poraneous eruption of the Siberian Traps large igneous province 1To whom correspondence may be addressed. Email: [email protected]. (17) drove ∼10 °C global warming within ca. 30,000 to 60,000 y This article contains supporting information online at https://www.pnas.org/lookup/suppl/ through greenhouse gas emissions (18, 19) and widespread doi:10.1073/pnas.1918953117/-/DCSupplemental. oceanic anoxia (20–22). These disastrous events killed over 90% www.pnas.org/cgi/doi/10.1073/pnas.1918953117 PNAS Latest Articles | 1of6 in ammonoids (29). These findings suggest a great change in 201.3 Rh biogeographic distribution during the P-Tr crisis, which would (Ma) have had a profound impact on LDG. In this study, we in- Late No vestigate LDG dynamics through the P-Tr mass extinction event, Triassic and the later recovery of biodiversity, using the most compre- hensive fossil database thus far for this time period. Ca 237 Global Changes in Biogeography La Middle In order to assess the effect of the P-Tr extinction and associated Triassic environmental extremes on latitudinal diversity patterns, we An 247.2 analyzed biogeographic distributions using a database consisting Sp Sm of occurrences of marine genera (including 20 major clades; see Early Methods for details) from the late Permian (Changhsingian, Triassic Di 254.1 Ma) to the Late Triassic (Rhaetian, 201.3 Ma). This da- Gr tabase is an update of an earlier P-Tr marine fossil database (23) 251.90 Late and includes 52,318 generic occurrences at a substage- or stage- Permian Ch level resolution from 1,768 literature sources (Dataset S1). 254.14 Among these, 7,752, 12,676, 13,456, and 18,634 occurrences are 0-15º 15-30º 30-45º 45-90º in the late Permian and Early, Middle, and Late Triassic, re- spectively. Based on reconstructed paleolatitudes, the collections were binned into four paleolatitudinal zones for each hemi- 46.4 65.5 84.6 103.7 122.8 sphere: 0 to 15°, 15 to 30°, 30 to 45°, and 45 to 90°. The larger size Subsampled diversity of the 45 to 90° bin was chosen to accommodate the relatively Fig. 2. Rarefied genus-level diversity trends related to latitude from the low sample density at higher latitudes in most time intervals. At late Permian to the end of the Triassic. Data are standardized by repeatedly the epoch level, data from the paleolatitudinal zones of the subsampling from a randomly generated set until reaching a quota of 136 Northern and Southern Hemispheres are analyzed separately occurrences in each time bin at each latitudinal interval (SI Appendix, Table (Fig. 1 and SI Appendix, Fig. S1 and Tables S1 and S2). At the S3). Diversities are drawn as a contour map by using Origin Pro-2017 soft- stage/substage level, data from the Northern and Southern ware. Ch, Changhsingian; Gr, Griesbachian; Di, Dienerian; Sm, Smithian; Sp, Spathian; An, ; La, Ladinian; Ca, Carnian; No, Norian; Rh, Rhaetian. Hemispheres are amalgamated to ensure sufficient sample sizes (Fig. 2 and SI Appendix, Figs. S2–S4 and Tables S3 and S4). We standardized genus diversity in each paleolatitudinal zone for LDG during the Early Triassic (Fig. 1), indicating a critical role each time bin using both incident-based rarefaction and extrap- Methods of environmental stability in maintaining a rich tropical fauna. olation methods (see more details in ). Extrapolating diversity estimators (Jackknife 1 and Chao 2; SI We found the biodiversity peaks in the 30 °N-15 °S bins in both Appendix, Fig. S1 B and C) show similar biogeographic patterns hot (Middle Triassic) and cold (late Permian) times but a flatter with raw data for the four time intervals (SI Appendix, Fig. S1A). In the Northern Hemisphere, genus diversity decreases from low to high paleolatitudes in the late Permian and Middle Triassic. In 325 A 400 B the Southern Hemisphere, the 15 to 30° bin has the lowest di- versity, which could be partially explained by insufficient sample Late 250 size (Fig. 1A). Most of the Southern Hemisphere bins from the 200 Triassic late Permian and Middle Triassic intervals have generic occur- 175 rences of less than 380. By contrast, both rarefied data and 0 shareholder quorum subsampling (SQS) diversity show that 250 Early Triassic time bins were characterized by a nearly uniform 250 Middle genus richness from low to high latitudes, except for the 45 to Triassic 125 90 °N bin (Fig. 1). The low diversity in the 45 to 90 °N bin is likely 200 due to a species-area effect (30), because this bin contained a smaller shelf area than other bins and included an area of ap- 150 0 proximately half the shelf size compared to the 45 to 90 °S bin 250 200 during the Early Triassic (31). Additionally, there are more oc- Early currences in the lower latitudes than there are in the mid and Triassic 175 100 high latitudes for the Early Triassic (e.g., 6,057, 1,730, 3,788, and 1,101 occurrence records from low to high latitudes, re- 100 0 spectively), suggesting that the flat LDG is not a merely a sam- 250 200 pling artifact. The Late Triassic interval is characterized by a diversity peak in the 15 to 30 °N latitudinal bin but exhibits a Late declining trend in diversity toward the polar region. Permian 175 100 The raw data show slightly different LDG patterns in the late Permian and the Middle Triassic (SI Appendix, Fig. S1A), in- 100 0 dicating that controlling for sampling variation is necessary for a 90ºN0º 90ºS 90ºN 0º 90ºS rigorous investigation on fossil diversity patterns even with such a Paleolatitude Paleolatitude rich record. During the late Permian, 282 genera have been Fig. 1. LDGs for late Permian and Triassic intervals. (A) Subsampled diversity found from the regions in the 15 to 30 °N zone, while 766 were using a quota of 380 occurrences for each time interval. Vertical bar presents found in the 0 to 15 °N zone. Both rarefied and SQS diversities the SD. (B) SQS diversity with a quorum level of 0.5. Dashed line represents show less difference between the two latitude zones (Fig. 1), with the discontinuous case. the late Permian pattern more closely resembling the Middle

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1918953117 Song et al. Triassic pattern, implying a sampling effect in the raw data. 3) Frequent recurrence. Extreme warming, oceanic anoxia, and Nevertheless, all subsampling methods have shown unmistakable enhanced weathering occurred recurrently and such condi- flattening of LDG during the Early Triassic (Fig. 1), which tions lasted for the entire Early Triassic, ca. 5 My (18, 20, provides compelling evidence for the strong impacts of mass 42, 43) (see Fig. 4). Together, these unstable environmental extinction and dramatic environmental changes on global bio- conditions prevented diversity recovery, even in the tropics, geography. Fossil data in the Early Triassic interval have a spatial and destroyed the advantage of this region as both the cradle distribution similar to other intervals (SI Appendix, Figs. S5 and and the museum for global biodiversity (45). S6), suggesting that the flat LDG during the Early Triassic is not Under the stress of such an extreme environment during the due to uneven spatial sampling and species-area effects (30). P-Tr crisis, preferential extinction at low latitudes may have Observed and estimated diversities at finer temporal resolution played an important role in the transformation of LDGs. To (i.e., 17 time bins including early Changhsingian, late Changhsin- evaluate this mechanism, we selected the Changhsingian and gian, early Griesbachian, late Griesbachian, Dienerian, Smithian, early Griesbachian as the interval to calculate extinction mag- Spathian, early Anisian, late Anisian, early Ladinian, late Ladinian, nitude because the major extinction pulses straddled the P-Tr early Carnian, late Carnian, early Norian, middle Norian, late boundary (46). The extinction magnitudes (calculated by the Norian, and Rhaetian) also show significant variations of LDGs number of extinct genera/the number of total genera) in the SI Appendix – (Fig. 2 and ,Figs.S2 S4), that is, from significant low- Changhsingian and early Griesbachian interval are 78.7% and latitudepeaksinthelatePermiantoflatLDGsintheEarlyTriassic 72.4% in the 0 to 15° and 15 to 30° zones, respectively, which are before returning back to LDGs with clear low-latitude peaks in the higher than the 30 to 45° zone with extinction rates of 60.9% but Middle Triassic. Remarkably, diversity recovery occurred in all not the 45 to 90° zone with 72.8% taxa extinct (Fig. 3B and SI latitudes during the Smithian and Spathian intervals (late Early Appendix, Table. S5). Other studies have suggested that the Triassic) and started in high latitudes, that is, 30 to 90° zones Changhsingian has a peak extinction in the higher latitudes, (Fig. 2). The midlatitude peak in observed diversity during the which is not seen in our data, but the (Griesbachian and Carnian (SI Appendix, Fig. S2) is not entirely an artifact of Dienerian) shows higher extinction in the tropics than in tem- sampling, because subsampled data also show a similar, but perate regions (22, 47), which better agrees with our work. A weaker, peak (Fig. 2). In addition, the end-Norian marine biota similar flat LDG pattern has also been found in mammals after experienced a short-term change in LDG, with lower diversity in the /Paleogene extinction event (48). the 0 to 15° zone than in midlatitude regions (Fig. 2 and SI In addition, our results show higher origination and invasion Appendix, Figs. S2–S4). –

rates at high latitudes in the late Griesbachian Smithian interval EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES (Fig. 3C), suggesting that high-latitude regions had become the Drivers of the Dynamic LDG refuge and cradle for marine organisms after the P-Tr mass ex- The similar LDGs during the hot Middle Triassic and cold late tinction. The higher invasion rates toward high latitudes (Fig. 3C) Permian contradict the notion that icehouse climates, which can are also consistent with the expectation of a pervasive tendency of maintain strong environmental gradients across space, are nec- migrating poleward. Therefore, both higher diversification rates at essary to produce such LDGs (13). Previous studies have gen- high latitudes and poleward migration would have played signifi- erally associated steep, normal LDG patterns with icehouse cant roles in producing the flat LDG in the Early Triassic. climates, such as the late Cenozoic including the present day (5, Our findings of a temporally dynamic LDG in response to EVOLUTION 32–34), late Paleozoic (35), and late (36, 37), sug- environmental changes coincident with the P-Tr mass extinction gesting a negative relationship between global temperature and is clear evidence for a strong role of major environmental crises the strength of LDG (13), albeit with some clade-specific ex- and massive climate perturbations in producing flatter LDGs. ceptions (38). The late Permian (Changhsingian Stage) was a For example, the delay of metazoan reef recovery in the Early cold period, during which the temperature was only slightly Triassic was an important factor in suppressed equatorial di- higher than during the late Paleozoic glaciation (39); the versity (49, 50) (Fig. 4), leading to a flatter LDG. Stable envi- strength of Changhsingian LDG is similar to the late Cenozoic ronments allow diversity to accumulate with higher speciation LDG for marine with markedly elevated generic richness rates and/or lower extinction rates (51). However, because en- in the tropics (32). However, the Middle Triassic is commonly vironmental stability and many other factors, especially climatic classified as a hothouse period with the average sea surface conditions, all covary with latitude, their relative importance in temperature ∼8 °C higher than seen at the present day (40), and shaping the LDG are difficult to compare based on spatial yet the steepness of LDG increased after the reestablishment of analyses alone. environmental stability. The flattened shape of Early Triassic Unstable and harsh environments and the major loss of spe- LDG matches well with ecological diversity data, which showed cies during the P-Tr extinction resulted in unstable global com- that the tropics had the highest level of functional diversity munities (52) and weak biotic interactions in both low- and high- during the late Permian but, following the extinction, a level of latitude regions. The blooms of opportunists (e.g., small fora- minifers, linguloid , microgastropods, and Claraia functional diversity similar to higher latitudes (41). – In contrast to the global environment before and after this bivalves) in the aftermath of the P-Tr extinction (23, 53 55) in- dicate that r-strategists dominated the marine realm. Despite an period, the dramatic environmental changes that began in the apparent lack of selection for larger geographic range sizes P-Tr boundary interval and lasted for ∼5 My are likely the during the P-Tr mass extinction (56), biogeographic cosmopoli- leading causes of the lack of a discernable LDG in Early Triassic tanism increased in both terrestrial (27) and marine (29) realms marine biota. These changes have three notable features: in the Early Triassic, likely because surviving opportunistic taxa 1) Strong intensity. Climatic/environmental conditions were the were able to proliferate geographically in the absence of intense most severe of the past 500 My, for example a ∼10 °C in- competition and their larger niche breadths allowed them to crease of sea surface temperature in ca. 30,000 to 60,000 y cope with harsh and variable conditions. (18, 19), rapid shifts between oxia and anoxia in shallow The hothouse climates during the Early Triassic probably also waters (21), and a significant increase of continental weath- contributed to weakening the marine LDG. Sea-surface tem- ering rates and nutrient delivery to the oceans (42, 43). perature of low latitudes in the Early Triassic was ∼15 °C higher 2) Global reach. Some environmental events, including anoxia than at present (40). An Earth system model of P-Tr climate and warming, affected most habitats and regions (21, 22, 44). suggests that the amplitude of warming in high paleolatitudes is

Song et al. PNAS Latest Articles | 3of6 A 0.9 B C 0.8 0.8

0.8 Extinction Origination 0.6 0.6 rpation/ rpation/

0.7 0.4 0.4 nation & invasion i 0.6 Invasion 0.2 Extirpation 0.2 Orig Extinction & exti 0.5 0 0 0-15º 15-30º 30-45º 45-90º 0-15º 15-30º 30-45º 45-90º 0-15º 15-30º 30-45º 45-90º

Fig. 3. Extinction and extirpation magnitudes in the Changhsingian and early Griesbachian interval and origination and invasion magnitudes in the late Griesbachian–Smithian interval. (A) The combined rates of extinction–extirpation and origination–invasion. (B) Extinction and extirpation rates in the Changhsingian and early Griesbachian interval. (C) Origination and invasion rates in the late Griesbachian-Smithian interval. Vertical bars represent SEs.

much higher than that in low paleolatitudes (22), resulting in a 59). High temperatures in tropics are beyond the thermal optima weak latitudinal temperature gradient. Additionally, extreme for some taxa, especially during global warming intervals. Pa- seawater temperatures (up to 35 °C) and associated anoxia leogeographic data show shelf area increased remarkably in the would have been lethal to many tropical organisms (Fig. 4). An midlatitudes of the Northern Hemisphere during the Late Tri- analog is happening in modern oceans, that is, the declining assic (31), which would have provided more habitats for marine oxygen caused by global warming and eutrophication is influ- organisms and accordingly contributed to higher diversity. encing marine life from gene to ecosystem levels (57). The midlatitude peak in diversity during the Carnian is Global temperatures declined by ∼4 °C in the early Middle probably due to extreme climate events that happened in the Triassic when compared to the Early Triassic hothouse (18), mid-Carnian interval. Carbon isotope records show a major which may have been enough to facilitate the reestablishment of negative excursion in both organic and carbonate δ13C at this a normal LDG in the mid-Triassic. Our findings suggest that only time (60, 61), reflecting significant perturbations of the carbon extreme and variable hothouse climates produce flat LDGs while cycle. Sea-surface temperature increased about 6 °C in this in- a stable greenhouse world can still have enough pole-to-equator terval (62, 63), which coincided with the major negative shift of climatic gradient to produce a significant, normal LDG. Among δ13C (63), suggesting a causal linkage between the injection of other factors, the yearly seasonal instability, including the variation pCO2 and global warming. The mid-Carnian event also affected of solar radiation and daylight, is likely a major cause of lower the marine biota and resulted in a biodiversity decline (23) and biodiversity at higher latitudes even during greenhouse periods. some ecological changes (64). The drop of diversity in the low latitudes during the end- Causes of Late Triassic Biogeographic Changes Norian interval is probably associated with environmental dis- During the Late Triassic, marine organisms were most diverse in turbance (62, 65). Conodont oxygen isotope data suggested a ∼7 the 15 to 30 °N region (Fig. 1). A midlatitude peak of marine °C warming in the late Norian, which lasted about 7 My (62). biodiversity has been reported in modern taxa (14), post- Significant negative excursions of organic carbon and nitrogen Paleozoic brachiopods (58), and Early and Middle Ordovician isotope ratios near the Norian–Rhaetian boundary suggest the taxa (37). Temperature and shelf areas have been proposed as development of widespread oceanic stagnation during this the primary variables influencing the bimodality of LDG (14, 37, interval (65).

Time scale Biotic changes Environmental changes 243 AB C D (Ma) 250 c i e l 244 s dd as An 125 17 245 i Mi Tr 246 0 247 200

248 p

Sp a

249 sic 100 G 1 s

f

250 ia e Early r

T 0 Sm e 251

Di R Gr 200 252 warming anoxic enhanced

253 n Ch 22 weathering

a 100 i 254 e m

r 15 20 25 30 35 x1 x2 x3 Lat 255 e 300 500 0.2 0.4 P Wu 0 SST (°C) -2 -1 0 1 2 3 F/F Generic richness Proportion of 13 10 10 10 10 10 10 RM 90ºN 0º 90ºS nekton Conodont Th/U

Fig. 4. Biotic and environmental changes throughout the late Permian to the Middle Triassic. (A) SQS diversities across latitudinal zones. (B) Genus richness and proportion of nekton (23). (C) The number of sites yielding metazoan reefs (50). (D) Sea-surface temperature (SST), ocean redox, and continental weathering. SST values are derived from conodont oxygen isotope data (SI Appendix, Table S6 and Dataset S2). Redox states of seawater are from conodont

Th/U ratios (20). Riverine-to-mantle Sr flux ratios (FR/FM) calculated from conodont Sr isotopes reflect continental weathering change (43).

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1918953117 Song et al. Implications for Modern Ecosystem Changes The fossil occurrences for each time bin were grouped into four paleo- Identifying the main drivers of global biogeographic patterns is a latitudinal zones in each hemisphere: 0 to 15°, 15 to 30°, 30 to 45°, and 45 to 90°. The high-latitude zone covers a total of 45° of latitudes because the critical step toward predicting future responses to projected sample sizes for 15° regions at high latitudes in most time bins were in- environmental changes. In particular, our results support the sufficient for rigorous analyses of diversity patterns. However, we note that previous suggestion that extreme climatic events, particularly higher latitudinal bands tend to cover smaller geographic areas than lower when combined with other anthropogenic effects, will lead to bands of the same number of degrees, which reduces the issue of uneven severe consequences for biodiversity (57), although a super sampling areas. Further, we employed statistical methods to account for the greenhouse Triassic-like world is a distant and perhaps unlikely sampling effects across latitudinal zones. Spanning the whole focal period in prospect. We show a flattening of the LDG after the biggest our study, the four paleolatitudinal zones contained a total of 22,526, mass extinction, which indicates a collapse of tropical ecosystems 13,649, 11,571, and 4,572 occurrences, respectively, from low to high including tropical reefs. We already know that modern reefs are latitudes. highly stressed (66, 67) and it seems that they will likely be the Rarefaction Method. first major victims of warming and, given that these are the most We applied the rarefaction method to compare generic diverse of all marine ecosystems, this will contribute to a flat- richness across latitudinal zones and time bins (72, 73), using the program PAleontological STatistics (PAST, version 3.16) (74). Because our dataset in- tening of the modern marine LDG (66, 68). cludes both micro- and macrofossil groups that systematically differ in the Methods abundance of individuals in each collection (23), abundance does not make an appropriate unit for the subsampling procedures for comparing total Fossil Database. We substantially updated an earlier database of P-Tr marine marine diversity. Instead, we treated each generic occurrence (the unique fossils (23) by adding 89 publications for 1,263 fossil occurrences including stratigraphic unit in which this genus occurred) as an individual sampling data from the Paleobiology Database. Fossil occurrences were compiled for unit, which serves as the analytical unit for rarefaction. We randomly sub- 17 substage-/stage-level time bins, following GSA Geological Time Scale v. sampled the fossil occurrences from each latitudinal zone in each time bin 5.0 (69), from the late Permian Changhsingian (starting 254.1 Ma) to the until a specific quota based on the minimum sample size in latitudinal pools. Late Triassic Rhaetian (ending 201.3 Ma), including, in sequential order, early We generated rarefaction curves in two temporal resolutions to compare Changhsingian, late Changhsingian, early Griesbachian, late Griesbachian, LDGs, that is, in the four epochs (the late Permian and Early, Middle, and Dienerian, Smithian, Spathian, early Anisian, late Anisian, early Ladinian, Late Triassic) and in the more refined 17 time bins as explained above. The late Ladinian, early Carnian, late Carnian, early Norian, middle Norian, late latitudinal faunas in each epoch were rarefied using a quota of 380. The Norian, and Rhaetian. The and biostratigraphy were rigorously fossil occurrences in the 17 time bins were subsampled until a quota of 136 validated to ensure consistency across the database. We based our analyses occurrences in each latitudinal zone. on genus-level occurrences because species-level identification is often in-

accurate and spotty. EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES SQS Method. The SQS approach (75) was applied to estimate diversity vari- The resulting global fossil database contains 52,318 generic occurrences ation across latitudes in the late Permian and Early, Middle, and Late Triassic from 4,875 collections in 1,768 publications (Dataset S1) (70). The total 4,342 intervals. SQS diversities were calculated with the divDyn R package at a genera belong to 20 major groups including 3 clades of algae (benthic cal- quorum level of 0.5 (76). careous algae, coccoliths, and dinoflagellates), 2 clades of protozoa (fora- minifers and radiolarians), 12 clades of invertebrates (annelids, bivalves, Data Access and Availability. brachiopods, bryozoans, cephalopods, corals, echinoderms, gastropods, hy- All data used to conduct analyses and plot figures drozoans, ostracods, nonostracod crustaceans, and sponges), and 3 clades of are available for download at https://datadryad.org/stash/dataset/doi:10. vertebrates (conodonts, fishes, and marine reptiles). 5061/dryad.41ns1rn9z. EVOLUTION

Paleolatitude Reconstruction. Paleolatitudes (and paleolongitudes) were ACKNOWLEDGMENTS. We thank the contributors to the Paleobiology reconstructed using PointTracker v7 rotation files published by the PALEO- Database and Philip Mannion and two anonymous reviewers for construc- tive reviews. This is Paleobiology Database publication number 372. This MAP Project (71) based upon the present-day georeference data and a research was supported by the National Natural Science Foundation of China model of global tectonic history. Paleolatitude data were reconstructed for (41821001), the State Key R&D project of China (2016YFA0601100), the Strategic every 10 My, for example with midpoints at 250 Ma for time bins from Priority Research Program of Chinese Academy of Sciences (XDB26000000), a Changhsingian to early Anisian, 240 Ma for late Anisian and Ladinian, 230 Marie Curie Fellowship (H2020-MSCA-IF-2015-701652), the Natural Environment Ma for Carnian, 220 Ma for early and middle Norian, and 210 Ma for late Research Council (NE/P0137224/1), and the German Science Foundation (DFG, HU Norian and Rhaetian. 2748/1-1).

1. E. R. Pianka, Latitudinal gradients in species diversity: A review of concepts. Am. Nat. 12. E. E. Saupe et al., Non-random latitudinal gradients in range size and niche breadth 100,33–46 (1966). predicted by spatial patterns of climate. Glob. Ecol. Biogeogr. 28, 928–942 (2019). 2. H. Hillebrand, On the generality of the latitudinal diversity gradient. Am. Nat. 163, 13. P. D. Mannion, P. Upchurch, R. B. J. Benson, A. Goswami, The latitudinal biodiversity 192–211 (2004). gradient through deep time. Trends Ecol. Evol. 29,42–50 (2014). 3. M. R. Willig, D. M. Kaufman, R. Stevens, Latitudinal gradients of biodiversity: Pattern, 14. C. Chaudhary, H. Saeedi, M. J. Costello, Bimodality of latitudinal gradients in marine process, scale, and synthesis. Annu. Rev. Ecol. Evol. Syst. 34, 273–309 (2003). species richness. Trends Ecol. Evol. 31, 670–676 (2016). 4. K. Roy, D. Jablonski, J. W. Valentine, G. Rosenberg, Marine latitudinal diversity gra- 15. J. J. Sepkoski Jr., A kinetic model of Phanerozoic taxonomic diversity, III. Post- dients: Tests of causal hypotheses. Proc. Natl. Acad. Sci. U.S.A. 95, 3699–3702 (1998). Paleozoic families and mass extinctions. Paleobiology 10, 246–267 (1984). 5. S. Huang, K. Roy, J. W. Valentine, D. Jablonski, Convergence, divergence, and paral- 16. D. H. Erwin, The Permo-Triassic extinction. Nature 367, 231–236 (1994). lelism in marine biodiversity trends: Integrating present-day and fossil data. Proc. 17. D. P. Bond, P. B. Wignall, “Large igneous provinces and mass extinctions: An update” Natl. Acad. Sci. U.S.A. 112, 4903–4908 (2015). in Volcanism, Impacts, and Mass Extinctions: Causes and Effects, G. Keller, A. C. Kerr, 6. M. Yasuhara, G. Hunt, H. J. Dowsett, M. M. Robinson, D. K. Stoll, Latitudinal species Eds. (Special Paper 505, Geological Society of America, 2014), Vol. 505, pp. 29–55. diversity gradient of marine zooplankton for the last three million years. Ecol. Lett. 18. Y. Sun et al., Lethally hot temperatures during the Early Triassic greenhouse. Science 15, 1174–1179 (2012). 338, 366–370 (2012). 7. G. G. Mittelbach et al., Evolution and the latitudinal diversity gradient: Speciation, 19. S.-Z. Shen et al., A sudden end-Permian mass extinction in South China. Geol. Soc. Am. extinction and biogeography. Ecol. Lett. 10, 315–331 (2007). Bull. 131, 205–223 (2018). 8. D. Jablonski, S. Huang, K. Roy, J. W. Valentine, Shaping the latitudinal diversity 20. H. Song et al., Geochemical evidence from bio-apatite for multiple oceanic anoxic gradient: New perspectives from a synthesis of paleobiology and biogeography. Am. events during Permian–Triassic transition and the link with end-Permian extinction Nat. 189,1–12 (2017). and recovery. Earth Planet. Sci. Lett. 353–354,12–21 (2012). 9. P. V. Fine, Ecological and evolutionary drivers of geographic variation in species di- 21. P. B. Wignall et al., Ultra-shallow-marine anoxia in an Early Triassic shallow-marine versity. Annu. Rev. Ecol. Evol. Syst. 46, 369–392 (2015). clastic ramp (Spitsbergen) and the suppression of benthic radiation. Geol. Mag. 153, 10. M. Yasuhara, G. Hunt, T. M. Cronin, H. Okahashi, Temporal latitudinal-gradient dy- 316–331 (2016). namics and tropical instability of deep-sea species diversity. Proc. Natl. Acad. Sci. 22. J. L. Penn, C. Deutsch, J. L. Payne, E. A. Sperling, Temperature-dependent hypoxia U.S.A. 106, 21717–21720 (2009). explains biogeography and severity of end-Permian marine mass extinction. Science 11. M. Yasuhara, D. P. Tittensor, H. Hillebrand, B. Worm, Combining marine macro- 362, eaat1327 (2018). ecology and palaeoecology in understanding biodiversity: Microfossils as a model. 23. H. Song, P. B. Wignall, A. M. Dunhill, Decoupled taxonomic and ecological recoveries Biol. Rev. Camb. Philos. Soc. 92, 199–215 (2017). from the Permo-Triassic extinction. Sci. Adv. 4, t5091 (2018).

Song et al. PNAS Latest Articles | 5of6 24. N. Brocklehurst, M. O. Day, B. S. Rubidge, J. Fröbisch, Olson’s Extinction and the lat- 51. M. Pontarp et al., The latitudinal diversity gradient: Novel understanding through itudinal biodiversity gradient of tetrapods in the Permian. Proc. Biol. Sci. 284, mechanistic eco-evolutionary models. Trends Ecol. Evol. 34, 211–223 (2019). 20170231 (2017). 52. P. D. Roopnarine, K. D. Angielczyk, Community stability and selective extinction 25. M. Bernardi, F. M. Petti, M. J. Benton, Tetrapod distribution and temperature rise during the Permian-Triassic mass extinction. Science 350,90–93 (2015). during the Permian-Triassic mass extinction. Proc. Biol. Sci. 285, 20172331 (2018). 53. Y. Peng, G. R. Shi, Y. Gao, W. He, S. Shen, How and why did the Lingulidae (Bra- 26. M. J. Benton, Hyperthermal-driven mass extinctions: Killing models during the chiopoda) not only survive the end-Permian mass extinction but also thrive in its permian-triassic mass extinction. Philos. Trans. A Math. Phys. Eng. Sci. 376, 20170076 aftermath? Palaeogeogr. Palaeoclimatol. Palaeoecol. 252, 118–131 (2007). (2018). 54. M. L. Fraiser, D. J. Bottjer, The non-actualistic early Triassic gastropod fauna: A case 27. D. J. Button, G. T. Lloyd, M. D. Ezcurra, R. J. Butler, Mass extinctions drove increased study of the lower Triassic sinbad limestone member. Palaios 19, 259–275 (2004). global faunal cosmopolitanism on the supercontinent Pangaea. Nat. Commun. 8, 733 55. E. Petsios, D. J. Bottjer, Quantitative analysis of the ecological dominance of benthic (2017). disaster taxa in the aftermath of the end-Permian mass extinction. Paleobiology 42, 28. C. J. Reddin, Á. T. Kocsis, W. Kiessling, Marine migrations trace climate 380–393 (2016). change over 450 million years. Glob. Ecol. Biogeogr. 27, 704–713 (2018). 56. P. G. Harnik, C. Simpson, J. L. Payne, Long-term differences in extinction risk among 29. A. Brayard et al., The Early Triassic ammonoid recovery: Paleoclimatic significance of the seven forms of rarity. Proc. Biol. Sci. 279, 4969–4976 (2012). diversity gradients. Palaeogeogr. Palaeoclimatol. Palaeoecol. 239, 374–395 (2006). 57. D. Breitburg et al., Declining oxygen in the global ocean and coastal waters. Science 30. R. A. Close, R. B. J. Benson, P. Upchurch, R. J. Butler, Controlling for the species-area 359, eaam7240 (2018). effect supports constrained long-term Mesozoic terrestrial vertebrate diversification. 58. M. G. Powell, B. R. Moore, T. J. Smith, Origination, extinction, invasion, and extir- Nat. Commun. 8, 15381 (2017). pation components of the latitudinal biodiversity gradient through the 31. C. R. Scotese, Atlas of Permo-Triassic paleogeographic maps (Mollweide Projection, – Maps 43 - 52, Volumes 3 & 4 of the PALEOMAP Atlas for ArcGIS (PALEOMAP Project, Phanerozoic Eon. Paleobiology 41, 330 341 (2015). Evanston, IL, 2014), 10.13140/2.1.2609.9209. 59. H. Saeedi, T. E. Dennis, M. J. Costello, Bimodal latitudinal species richness and high – 32. J. Alroy et al., Phanerozoic trends in the global diversity of marine invertebrates. endemicity of razor clams (). J. Biogeogr. 44, 592 604 (2017). δ13 Science 321,97–100 (2008). 60. J. Dal Corso et al., Discovery of a major negative C spike in the Carnian (Late Tri- – 33. D. Jablonski, K. Roy, J. W. Valentine, Out of the tropics: Evolutionary dynamics of the assic) linked to the eruption of Wrangellia flood basalts. Geology 40,79 82 (2012). latitudinal diversity gradient. Science 314, 102–106 (2006). 61. S. Mueller, L. Krystyn, W. M. Kürschner, Climate variability during the Carnian Pluvial 34. A. Miraldo et al., An Anthropocene map of genetic diversity. Science 353, 1532–1535 phase—A quantitative palynological study of the Carnian sedimentary succession at (2016). Lunz am See, Northern Calcareous Alps, Austria. Palaeogeogr. Palaeoclimatol. Pa- 35. M. G. Powell, Climatic basis for sluggish macroevolution during the late Paleozoic ice laeoecol. 441, 198–211 (2016). age. Geology 33, 381–384 (2005). 62. J. A. Trotter, I. S. Williams, A. Nicora, M. Mazza, M. Rigo, Long-term cycles of Triassic 36. A. Z. Krug, M. E. Patzkowsky, Geographic variation in turnover and recovery from the climate change: A new δ18O record from conodont apatite. Earth Planet. Sci. Lett. Late Ordovician mass extinction. Paleobiology 33, 435–454 (2007). 415, 165–174 (2015). 37. B. Kröger, Changes in the latitudinal diversity gradient during the great ordovician 63. Y. D. Sun et al., Climate warming, euxinia and carbon isotope perturbations during biodiversification event. Geology 46, 127–130 (2017). the Carnian (Triassic) Crisis in South China. Earth Planet. Sci. Lett. 444,88–100 (2016). 38. P. D. Mannion et al., Climate constrains the evolutionary history and biodiversity of 64. A. M. Dunhill, W. J. Foster, J. Sciberras, R. J. Twitchett, Impact of the Late Triassic mass crocodylians. Nat. Commun. 6, 8438 (2015). extinction on functional diversity and composition of marine ecosystems. Palae- 39. B. Chen et al., Permian ice volume and palaeoclimate history: Oxygen isotope proxies ontology 61, 133–148 (2018). revisited. Gondwana Res. 24,77–89 (2013). 65. M. A. Sephton et al., Carbon and nitrogen isotope disturbances and an end-Norian 40. H. Song, P. B. Wignall, H. Song, X. Dai, D. Chu, Seawater temperature and dissolved (Late Triassic) extinction event. Geology 30, 1119–1122 (2002). – oxygen over the past 500 million years. J. Earth Sci. 30, 236 243 (2019). 66. T. P. Hughes et al., Global warming and recurrent mass bleaching of corals. Nature 41. W. J. Foster, R. J. Twitchett, Functional diversity of marine ecosystems after the Late 543, 373–377 (2017). – Permian mass extinction event. Nat. Geosci. 7, 233 238 (2014). 67. G. De’ath, J. M. Lough, K. E. Fabricius, Declining coral calcification on the great barrier 42. T. J. Algeo, R. J. Twitchett, Anomalous Early Triassic sediment fluxes due to elevated reef. Science 323, 116–119 (2009). – weathering rates and their biological consequences. Geology 38, 1023 1026 (2010). 68. A. R. Harborne, A. Rogers, Y.-M. Bozec, P. J. Mumby, Multiple stressors and the 43. H. Song et al., Integrated Sr isotope variations and global environmental changes functioning of coral reefs. Annu. Rev. Mar. Sci. 9, 445–468 (2017). through the Late Permian to early Late Triassic. Earth Planet. Sci. Lett. 424, 140–147 69. J. Walker, J. Geissman, S. Bowring, L. Babcock, GSA geological time scale v. 5.0. (2015). (Geological Society of America, 2018), https://www.geosociety.org/documents/gsa/ 44. S. Grasby, B. Beauchamp, A. Embry, H. Sanei, Recurrent early triassic ocean anoxia. timescale/timescl.pdf. Geology 41, 175–178 (2013). 70. H. Song, S. Huang, E. Jia, X. Dai, P. B. Wignall, A. M. Dunhill, Data from: Flat lat- 45. D. Jablonski et al., Out of the tropics, but how? Fossils, bridge species, and thermal itudinal diversity gradient caused by the Permo-Triassic mass extinction. Dryad. ranges in the dynamics of the marine latitudinal diversity gradient. Proc. Natl. Acad. Sci. U.S.A. 110, 10487–10494 (2013). https://doi.org/10.5061/dryad.41ns1rn9z. Deposited 8 February 2020. 46. H. Song, P. B. Wignall, J. Tong, H. Yin, Two pulses of extinction during the Permian- 71. C. Scotese, The PALEOMAP Project PaleoAtlas for ArcGIS (PALEOMAP Project, Ar- Triassic crisis. Nat. Geosci. 6,52–56 (2013). lington, TX, 2008). 47. C. J. Reddin, Á. T. Kocsis, W. Kiessling, Climate change and the latitudinal selectivity of 72. A. I. Miller, M. Foote, Calibrating the ordovician radiation of marine life: Implications – ancient marine extinctions. Paleobiology 45,70–84 (2019). for phanerozoic diversity trends. Paleobiology 22, 304 309 (1996). 48. P. J. Rose, D. L. Fox, J. Marcot, C. Badgley, Flat latitudinal gradient in Paleocene 73. J. Alroy et al., Effects of sampling standardization on estimates of Phanerozoic marine mammal richness suggests decoupling of climate and biodiversity. Geology 39, diversification. Proc. Natl. Acad. Sci. U.S.A. 98, 6261–6266 (2001). 163–166 (2011). 74. Ø. Hammer, D. Harper, P. Ryan, PAST: Paleontological statistics (Version 3.16, National 49. E. Flügel, “Triassic reef patterns” in Phanerozoic Reef Patterns, W. Kiessling, E. Flügel, History Museum, University of Oslo, 2017), p. 256. J. Golonka, Eds. (Society for Sedimentary Geology, Tulsa, OK, 2002), Vol. 72, pp. 75. J. Alroy, Fair sampling of taxonomic richness and unbiased estimation of origination 391–463. and extinction rates. Paleontol. Soc. Pap. 16,55–80 (2010). 50. W. Kiessling, Reef expansion during the Triassic: Spread of photosymbiosis balancing 76. A. T. Kocsis, C. J. Reddin, J. Alroy, W. Kiessling, The R package divDyn for quantifying climatic cooling. Palaeogeogr. Palaeoclimatol. Palaeoecol. 290,11–19 (2010). diversity dynamics using fossil sampling data. Methods Ecol. Evol. 10, 735–743 (2019).

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1918953117 Song et al.