The paleoecological disturbances recorded during the Early Toarcian warming event (183 Myr ago), including marine anoxia, sea level rise, seawater acidification, carbonate production crisis, and species extinctions, are often regarded as past examples of Earth's possible responses to the rapid emergence of super greenhouse conditions. However, physical mechanisms explaining both the global and local expressions of paleoenvironmental events are still highly debated. Here we analyze the paleoclimatic and paleoceanographic consequences of increases in atmospheric pCO2 levels at a multiscale resolution using a fully coupled ocean–atmosphere model (FOAM). We show that, in association with stronger high-latitude precipitation rates and enhanced continental runoff, the demise of polar sea ice due to the global warming event involved a regional freshening of Arctic surface seawaters. These disturbances lead to progressive slowdowns of the global oceanic circulation accountable for widespread ocean stratification and bottom anoxia processes in deep oceanic settings and epicontinental basins. In agreement with very negative oxygen isotope values measured on fossil shells from the NW Tethys, our simulations also show that recurrent discharges of brackish and nutrient-rich Arctic surface waters through the Viking Corridor could have led to both vertical and geographical gradients in salinity and seawater δ18O in the NW Tethyan seas. Locally contrasted conditions in water mass density and rises in productivity rates due to strong nutrient supplies could partly explain the regional severity of the anoxic event in the restricted Euro-boreal domains, as it has been previously suggested and modeled regionally.
 For decades, the Early Toarcian global warming event has been especially well documented by δ18O analyses performed on belemnites, brachiopods, and fish teeth from European outcrops. The statistical analysis of current data show a rapid and significant decrease in δ18O values at the Pliensbachian-Toarcian boundary [Dera et al., 2011a], which is characterized by geographical differences in amplitudes (Figure 1). The range of isotopic shifts reach 6‰ in England and northeastern France [Dera et al., 2009b; McArthur et al., 2000], 3‰ in southern Germany, Bulgaria, and Spain [Bailey et al., 2003; Gómez et al., 2008; Metodiev and Koleva-Rekalova, 2008; Rosales et al., 2004; van de Schootbrugge et al., 2005a], and 2‰ in Portugal and southern France [Suan et al., 2008; van de Schootbrugge et al., 2010]. If solely linked to temperature changes, these isotopic variations would express drastic increases in surface seawater temperature (SST) exceeding more than 10°C. Because this degree of temperature change appears unrealistic for subtropical domains, which are less prone to climatic variations than poles, additional factors linked to warmer conditions have been suggested. These include enhanced rainfall and fluvial freshwater input from surrounding continents [Bailey et al., 2003; McArthur et al., 2008; Röhl et al., 2001; Saelen et al., 1996], southward discharge of low salinity water from Arctic seas [Bjerrum et al., 2001; Hesselbo et al., 2000; Mattioli et al., 2008; van de Schootbrugge et al., 2005a], and reductions in polar ice volumes monitoring the global seawater δ18O values [Dera et al., 2011a; Suan et al., 2010]. Though such changes appear likely for driving regional anoxic conditions (through higher nutrient supply and salinity stratification), these hypotheses are mainly based on geochemical proxies and still remain speculative. This is mostly because they do not contain physically based models, which provide confirmation of suspected disturbances and evaluation of respective amplitudes. For the moment, only Bjerrum et al.  have specifically tested the relations between global and regional water mass circulations by using a simple partly coupled ocean–atmosphere model. Consistently with the initial hypothesis of Hesselbo et al. , these authors founded that when the Tethyan oceanic waters become relatively denser, strong boreal currents could flow southward in the NW Tethyan domain and locally trigger reductions in salinity. At larger scale, a better appraisal of atmospheric and oceanic processes is however of prime importance for understanding the mechanisms, which underlie the global expression of bottom anoxia in deep oceanic settings and in most epicontinental basins.
 Given the strong rise in atmospheric pCO2 levels during the T-OAE, the aim of this integrative study is to explore the paleoclimatic and paleoceanographic consequences of an increase in the greenhouse gas concentration during the Early Toarcian. To this end, a series of pCO2 sensitivity experiments have been carried out using a fully coupled ocean–atmosphere model. Our goal is to specify the nature of overall mechanisms linking greenhouse gas concentration, paleoclimate, and anoxia by evaluating the role of ocean–atmosphere dynamics on the disturbances recorded during this period at the global and regional scale. Furthermore we also attempt to provide clarity regarding the spatial heterogeneity of larges fluctuations in δ18O recorded by fossil shells from the NW Tethyan domain. For this, we address three main questions: (1) How air temperature, moisture, surface seawater temperature, and salinity evolved during the Toarcian warming event?, (2) What were the factors that governed their fluctuations?, and (3) What were the consequences for the ocean circulation dynamics?
 The continental distribution used for experiments derives from the Early Jurassic paleogeographic reconstruction of R. Blakey (http://cpgeosystems.com) (Figure 2). Given the importance of the sea level rise during the Early Toarcian [Hallam, 1981, 1997, 2001], the Viking and Hispanic corridors were slightly enlarged for allowing surface water exchanges between the NW Tethyan, Arctic, and Panthalassan seas. Even if their precise opening times and depths remain controversial (especially for the Hispanic Corridor), it is generally suspected that these epicontinental seaways were already sufficiently deep for allowing the migration of ammonites, bivalves, ostracods, foraminifers or dinoflagellates in the middle of the Early Jurassic [Aberhan, 2001, 2002; Arias, 2006; Dera et al., 2011b; Hallam, 1977; Mattioli et al., 2008; Nikitenko, 2008; van de Schootbrugge et al., 2005a; Venturi et al., 2006]. Also, the NW Tethyan paleogeography was simplified to few emerged lands owing to the limitation in resolution imposed by the oceanic model. For the same reasons but also for uncertainty problems, spatial disparities in basinal depth were not considered for the NW Tethyan seas. Even if these details as well as the location of sills may influence the local circulation patterns in high-resolution models [Bjerrum et al., 2001], we consider that the consequences are negligible at the scale of our medium-resolution oceanic grid. Consequently, we assume a mean depth of −200 m for European basins and all epicontinental shelves, a gradual transition for the oceanic slope, and a flat bottom bathymetry of −4500 m for deep oceans. Similarly to GCM simulations of Sellwood et al.  and consistently with the lack of major plate collision during the Jurassic, a maximum elevation of 2000 m is ascribed to the main mountain chains. Finally, boundary conditions were completed by calculating a reduction of −1.53% in the solar constant relative to the present-day value (i.e., 1344.07 against 1365 W.m−2) [Gough, 1981], and by applying the present-day orbital parameters.
 In order to simulate the Early Toarcian warming event, we ran the model at three different atmospheric pCO2 levels. Consistently with pCO2 values inferred from stomatal indexes measured by McElwain et al. , atmospheric CO2 concentrations were set to 2×, 4×, and 6× the preindustrial level (i.e., 560, 1120, 1680 ppm). These three successive atmospheric pCO2 levels are expected to capture the paleoclimatic changes ranging from the cool Late Pliensbachian period to the Early Toarcian thermal optimum. The experiments were integrated for 1000 years without flux corrections or deep ocean acceleration. During the last 100 years of model integration, there is no apparent drift in the upper ocean (between the surface and 300 m depth), and <0.001°C/year change in globally averaged ocean temperature.
3. Results and Discussion
3.1. Influences of pCO2 Levels on the Toarcian Paleoclimate
 Whatever the pCO2 levels estimated for the Pliensbachian-Toarcian transition, it is noteworthy that the latitudinal distribution of annually averaged air temperatures inferred from simulations appears more equable than for the present-day (Figure 3a). Interestingly we observe, on the one hand, extended subtropical areas with annual air temperature of 20 to 30°C up to 40° latitude (especially for the 4× and 6× experiments), and cold to temperate subpolar climates characterized by annual temperatures of −10 to 5°C on the other. Though heavily contrasted, these results fully agree with previous Jurassic temperature patterns modeled by Donnadieu et al.  and Sellwood and Valdes , although their simulations relied on GCMs neglecting deep ocean circulation effects. Our results are also particularly compatible with the current databases of paleobotanical, paleontological, and sedimentary proxies available for the Early Jurassic, which indicate both globally extended subtropical climatic belts and cold to temperate biomes toward the higher latitudes [Hallam, 1984, 1985, 1993; Rees et al., 2000; Ziegler et al., 1993, 2004].
 Our experiments also show that, by assuming a tripling of pCO2 levels during the Early Toarcian, the average global air temperature increases by +4.5°C, ranging from 15.5°C at 2× CO2 to 20°C at 6× CO2 (Figure 3a). These results are greater than previous estimates (+3°C) based on simple thermodynamic relationships between pCO2 values and temperature [Beerling and Brentnall, 2007]. Our GCM simulations consequently reveal the importance of feedback processes on the modulation of global temperatures. Toward the high latitudes, the air heating exceeds +10°C owing to reductions in ice cover and albedo, which amplify the mechanism (Figure 4a). Such a drastic heating in subpolar areas is corroborated by substantial paleofloristic reorganizations in the Siberian area characterized by northward dispersals of thermophylous ferns and gymnosperms of Euro-Sinian origin [Zakharov et al., 2006]. In the southern hemisphere, similar plant turnovers have been documented in continental deposits from eastern Australia, with relative rises in the abundance of Araucarian and Podocarp conifers typical of more temperate domains (see Turner et al.  for a review). In the midlatitudes and tropical domains, the magnitude of the warming event appears much lower and reaches +5°C and +3°C respectively. Nevertheless, this climate change seems to have been strong enough to trigger major modifications in the southern Laurasian vegetation, as documented by the increasing domination of Cheirolepidiaceans and subtropical ferns in western Chinese basins [Wang et al., 2005]. Importantly, this latitudinal heterogeneity in air heating is very coherent with previous experiments simulating an increase of ∼1000 ppm during the Late Triassic and Middle Cretaceous [Huynh and Poulsen, 2005; Poulsen et al., 2007].
 In parallel to the global warming event, our simulation results reveal that successive rises in pCO2 during the Early Toarcian drove a slight increase of +9 cm/year in the global precipitation rates (Figure 3b). This rise is marked by a global redistribution of net moisture (i.e., precipitation minus evaporation rates) over lands and seas (Figure 4b). At 6× CO2, high latitudes and most equatorial regions of Gondwana and Tethys get substantially more humid, with precipitations rates regionally rising by +10 to +20 cm/year. Likely marked by amplifications of strong monsoonal events typical of Pangean supercontinents [Chandler et al., 1992; Kutzbach and Gallimore, 1989; Loope et al., 2001; Parrish and Peterson, 1988], these regional changes involve a stronger runoff on the boreal landmasses such as the Fenno-Scandian and Siberian shields, as well as on the northwestern and southern parts of the Gondwanan continent (Figures 3c and 4c). Consistently with our results showing global rises of +3.5 cm/year in the mean annual continental runoff (locally ranging from −20 to +25 cm/year), stronger weathering rates of continental crusts are confirmed by independent geological proxies. These latter include excursions toward more radiogenic values of strontium and osmium isotopes [Cohen et al., 2002], higher kaolinite contents in marine sediments deposited close to Euro-boreal landmasses [Dera et al., 2009a], or thick detrital deposits and important phosphorus influxes from northwestern margins of Gondwana [Bodin et al., 2010]. Conversely, the eastern parts of the Panthalassan Ocean and the marine peri-Tethyan margins get substantially drier at 6× CO2 (Figure 4b). In details, evaporation rates would have been markedly higher over the marine Mediterranean areas of the Tethyan domain during the Early Toarcian warming event (i.e., on the eastern Maghrebian margins of Gondwana). To our knowledge, no sedimentological feature currently proves such a regional rise in aridity at this time. It is probably because the sea level was too elevated for recording obvious evaporation indexes [Hallam, 1981, 1997, 2001]. Nevertheless, it is worth noting that regional evaporite deposits and carbonate platforms were very common at times of lower sea level as in the preceding Sinemurian and Pliensbachian periods [Rees et al., 2000; Ziegler et al., 2004]. Also, the atmospheric GCM simulations of Chandler et al.  and Sellwood and Valdes  confirm the presence of a seasonally dry climate over this marine domain during the Early and Late Jurassic. Consequently, we suggest that the marine Mediterranean domain could well have been particularly prone to high evaporation rates all over the Jurassic and more especially during the Toarcian and Late Oxfordian–Early Tithonian climatic optimums [Dera et al., 2011a].
3.2. Arctic Seawater Warming and Freshening
 Along with air temperature patterns, our modeling approach shows that the evolution of surface seawater temperatures was zonally heterogeneous across the globe (Figures 5a and 6a). Based on model considerations, SSTs increased by +2 to +3.5°C in subtropical domains, testifying that variations calculated from δ18O values (i.e., exceeding +10°C) may be strongly overestimated for the Tethyan seas. However, seawater temperatures could have gained +8°C to the poles, so that the sea ice formation could have been importantly impacted (Figure 6b). In agreement with occurrences of glendonites and exotic boulders in Late Pliensbachian sediments from northern Siberia [Rogov and Zakharov, 2010; Suan et al., 2011], our simulations attest that at 2× CO2, and in a context of reduced solar constant, the temperature of polar surface seawater could have reached 0 to −2°C. This would then allow the formation of an extended sea ice volume, stretching south to ∼60° latitude, prior to the well-known global warming event. These cold to freezing seawater conditions could well have been favorable to the first emergence of endemic Jurassic ammonites species in the Arctic province [Dera et al., 2011b]. During the subsequent rise in SST, our simulations show that the annual sea ice extension decreased by half at 4× CO2 then 90% at 6× CO2 (Figure 6b). This change could have led to important disturbances in the Arctic biota, as reflected by strong biotic turnovers and regional migration events [Dera et al., 2010, 2011b; Nikitenko and Mickey, 2004; Nikitenko and Shurygin, 1992; Suan et al., 2011; Zakharov et al., 2006]. Nevertheless, this does not give evidence for waxing and waning of large ice volumes on continents leading to strong decreases in seawater δ18O values. Recently, simulations have shown that small Jurassic ice sheets were possible over boreal uplands but that extended ice caps likely occurred when pCO2 dropped below 400 ppm or at times of minimal seasonal forcing [Donnadieu et al., 2011; Valdes et al., 1995]. Given the slightly higher pCO2 levels estimated by McElwain et al.  and in a context of relatively temperate polar climates, we suggest that demises of partial continental ice sheets formed during the Late Pliensbachian cooling event likely involved minor variations in seawater δ18O values (below 0.5‰).
 In conjunction with regionally higher evaporation rates, our simulations show that the Early Toarcian warming event involved a slight rise in surface salinity in most subtropical and midlatitude seas, with regional maximums of variations reaching +1 to +2 psu in the Mongol-Okhotsk Ocean (i.e., northeast of Laurasia) and in the southwestern part of Gondwana (Figures 5b and 6c). Conversely, a striking surface water freshening locally reaching −6 psu occurred in the Arctic seas, with average regional decreases in salinity from 32 to 28 psu. To our knowledge, this major regional event, affecting Arctic surface waters, remains difficult to confirm. This is because there is a lack of geochemical and geological salinity proxies from boreal outcrops [Nikitenko and Shurygin, 1992; Suan et al., 2011]. Nevertheless, it is worth mentioning that compared to benthic and nektobenthic faunas, the extinction rates of marine organisms inhabiting surface to intermediates waters such as ammonites were slightly higher in the Arctic seas than in the NW Tethyan domains [Dera et al., 2010]. Also, this high-latitude freshening process seems realistic as it occurred with stronger amplitudes during the PETM and the warm Eocene “Azolla event” [Brinkhuis et al., 2006; Cope and Winguth, 2011; Roberts et al., 2009]. Similarly, the regional drop in salinity would be influenced by multiple factors, including a rapid thaw of polar sea ice volumes under warmer conditions, strong fluvial discharges from the Siberian and Fenno-Scandian shields, and a paleogeographical restriction of the Arctic basin preventing important mixing with salty Panthalassan waters.
3.3. Oceanic Circulation Slowdown During the T-OAE
 Whatever the pCO2 levels used for sensitivity experiments, our Early Jurassic simulations show that the global surface ocean circulation was quite symmetrical between the two hemispheres (Figure 7a). The main simulated features consist in: 1) strong equatorial currents flowing westward in the Tethyan and Panthalassan oceans, drifting along the northern Gondwanan margin, then passing through the Hispanic Corridor; and 2) two large Panthalassan subtropical gyres with respectively clockwise and anticlockwise rotations in the northern and southern hemispheres. These global results confirm the paleobiogeographically and physically based conceptual model of Arias  for the Early Jurassic. Also, they are very consistent with previous ocean circulation patterns based on Permian or idealized Pangean continental configurations [Kutzbach et al., 1990; Winguth et al., 2002]. The only difference concerns the velocity of surface currents, which could have been markedly more vigorous in the southern hemisphere than to the North during the Permian [Winguth et al., 2002]. Excepting potential model dependences, we suggest that this divergence could be linked to the relatively more polar position of Gondwana at the end of the Paleozoic.
 Under low atmospheric CO2 concentrations and lower derived temperatures (i.e., during the spinatum ammonite zone), our results indicate that the subtropical Panthalassan currents drifting along the eastern shelves of Laurasia and Gondwana followed extended poleward convections in the two hemispheres. By cooling, these salty currents then sank toward the bottom to the contact of high-latitude sea ice. This density-related process is reflected in profound mixed layer depths in the Arctic and southwestern parts of the Panthalassan Ocean (Figure 7b), inducing strong and symmetrical meridional overturning circulations in the two hemispheres (Figure 8). Importantly, this corroborates previous Pangean simulations based on similar landmass configurations and low pCO2 levels [Huynh and Poulsen, 2005; Kutzbach et al., 1990; Smith et al., 2004]. Conjointly with higher oxygen solubility rates in cold waters [Weiss, 1970], this mechanism likely ensured an efficient water mass oxygenation during the Late Pliensbachian cooling event, which preceded the massive release of greenhouse gas [Suan et al., 2010]. Indeed, a good ventilation of bottom seawater is supported in the NW Tethyan basins by low organic matter accumulations and abundant bioturbation features in shallow paleoenvironments during the spinatum ammonite zone [Röhl et al., 2001; Schmid-Röhl et al., 2002] and also by high diversification rates of macro- and microorganisms inhabiting different parts of the water column [Dera et al., 2010; Hallam, 1987; van de Schootbrugge et al., 2005a]. Moreover, Hori  interpreted the presence of hematite-bearing bedded cherts in Japanese outcrops as a proof of oxidizing conditions in deep oceanic waters of Panthalassa before the T-OAE.
 During the Early Toarcian warming event, the freshening of surface Arctic seawaters drove major changes in the global ocean dynamics marked by a reorganization of surface circulation patterns (Figure 7). Similarly to processes suggested in the Arctic seas during the PETM or the ongoing warming event [Aagaard and Carmack, 1989; Cope and Winguth, 2011], regional changes in water mass density stopped the northward meridional convection and reduced the sinking of salty subtropical Panthalassan currents in the Arctic area, retrospectively reinforcing the surface water freshening processes and warming of deep waters. As previously suggested by Schmidt and Mysak  as well as Bjerrum et al. , this regional change in deep-water formation was partly compensated by higher mixing rates in the southern hemisphere, and to a lesser extent in the southern Asian areas of the subtropical Tethyan domain (Figure 7b). More surprising, we note that for the 4× and 6× experiments, the mixed layer depth appears slightly deeper in the European basins. This would suggest a good ventilation of the water column during the T-OAE, which seems quite contrary to the setting of anoxic events in the NW Tethys. However, it is imperative that we note that this local simulated pattern should be regarded with caution. This is primarily because our simulations rely on a homogeneous epicontinental bathymetry of −200 m, which neglects the presence and the role of shallow silled Euro-boreal basins promoting water mass restriction [McArthur et al., 2008]. Nevertheless, Trabucho-Alexandre et al.  recently demonstrated that the conditions of organic matter accumulations would have been markedly variable in the NW Tethys, and regionally monitored by very high productivity rates in contexts of highly dynamic energetic conditions.
 At global scale, the polar disturbances in water mass conditions caused a substantial reduction in the global ocean meridional overturning circulation characterized a progressive disruption in the symmetrical mixing patterns between the two hemispheres (Figure 8). Importantly, this sluggish thermohaline circulation is in agreement with radiolarian orientations measured in Jurassic cherts from Japan, which suggest a long-term decrease in the deep Panthalassan current velocity from the Pliensbachian to the Toarcian [Dozen and Ishiga, 1997]. Also, these results are very coherent with pCO2 sensitivity experiments performed for the previous Late Triassic warming event [Huynh and Poulsen, 2005]. Nevertheless, it is worth noting that for other periods like the Late Permian, GCM simulations show that massive rises in atmospheric CO2 concentrations may drive opposite mechanisms, with more vigorous and more symmetrical deep-sea circulations under warmer climates [Winguth et al., 2002; Winguth and Maier-Reimer, 2005]. This implies that simple extrapolations between high pCO2 levels, ocean dynamics, and OAE may be avoided, because the paleogeographical context may have a paramount influence.
3.4. Consequences on the NW Tethyan Seas
 As discussed above, our results show that SSTs increase by +2 to +3.5°C in most epicontinental basins of the NW Tethys (Figure 9b). The only exception concerns a rise of +5°C in deeper waters (−100 to −200 m) of southwestern seas approximately corresponding to the present Lusitanian Basin. As revealed by circulation patterns (Figures 7a, 9c, and 9d), this thermal anomaly is probably related to a strengthening of warm equatorial Tethyan currents drifting along the northern Gondwanan margins. Furthermore, the drift of these currents through westernmost areas and their subsequent clockwise rotation due to southward-directed boreal flows is coherent with neodymium isotope data [Dera et al., 2009b], as well as ammonite and nannofossil migration routes evidenced for the Early Toarcian [Dera et al., 2011b; Reggiani et al., 2010]. Based on 6× experiment results (Figure 9d), it is also likely that more vigorous NW Tethyan currents would account for major disruptions in faunal provincialism and the celebrated northward expansion of marine Mediterranean faunas at the beginning of the Toarcian [Arias and Whatley, 2005; Dera et al., 2011b; Macchioni and Cecca, 2002; Vörös, 2002].
 However, the spatial distribution of rises in SST simulated during the Toarcian warming event is in disagreement with available δ18O data. Our synthesis of available geochemical data show that shifts in oxygen isotopes, recorded at the onset of the T-OAE (Figure 1), decrease less significantly in sedimentary basins that are geographically outside of the western Boreal domain. This is especially obvious in the Lusitanian and Grand Causses basins, which are adjacent to the Mediterranean domain. Our results provide further evidence suggesting that these sorts of oceanic changes may greatly complicate the paleoclimatic interpretation of δ18O data and refute an exclusive and straightforward thermal interpretation of isotopic records [Bailey et al., 2003; Dera et al., 2009b; Saelen et al., 1996; van de Schootbrugge et al., 2005a]. Conversely, the observed oxygen isotope gradient is more consistent with regional influxes of brackish seawaters from Arctic seas through the Viking Corridor (Figure 9a). According to our simulations, these southward-directed currents were probably driven by strong water density gradients and local basinal configurations, which may have reduced the salinity of adjacent seas by −1 to −5 psu at the surface. Our global modeling approach is consistent with the hypothesis of Hesselbo et al.  and the regional simulations of Bjerrum et al. , and further contradicts the direct influence of strong freshwater influxes from boreal landmasses at this time [Bailey et al., 2003; Röhl et al., 2001; Saelen et al., 1996].
 As our model does not use isotope fractionation modules, it remains difficult to directly evaluate the oxygen isotope composition of Arctic waters and their impact on δ18O values recorded by the Tethyan fossils. Nevertheless, several modeling studies have shown that during periods of high atmospheric pCO2 levels, offsets of 2 to 6‰ may occur between polar and subtropical seawater δ18O values [Roche et al., 2006; Tindall et al., 2010]. As these estimates correspond to the range of observed δ18O variations in the Early Toarcian, discharges of brackish Arctic waters with lower δ18O values appears as a plausible scenario for explaining both the strong amplitude and the latitudinal gradient of δ18O changes recorded by fossils in the NW Tethys. Of course, this hypothesis should be tested with an appropriate model in a next step.
 Our coupled ocean–atmosphere modeling approach provides a reliable scenario for explaining the synchronous setting of bottom anoxic conditions in both deep oceanic and epicontinental seas, as well as their spatial differences in intensity. Our model shows that the Early Toarcian warming event and the resulting changes in polar water mass density and temperature triggered a general slowdown in the thermohaline circulation. It is likely that this global disturbance accounted for a strong decrease in the water column mixing, resulting in oxygen depletions in most bottom paleoenvironments. However, additional factors to water mass stratification such as global increases in nutrient supply, lower oxygen solubility in warmer waters, and basinal restrictions cannot be excluded and it is likely they acted in combination during past OAEs [Jenkyns, 2010; McArthur et al., 2008; Meyer and Kump, 2008]. For example, the conjunction of additional adverse circumstances could explain the regional severity of anoxic events in the NW Tethys. According to our results and high-resolution simulations of Bjerrum et al. , we suggest that recurrent discharges of colder, nutrient-rich, and freshened Arctic seawater through the Viking Corridor could have facilitated local water mass stratifications and increased productivity rates in the Euro-boreal basins. Compared to shallow or open marine paleoenvironments of the Mediterranean area, this regional severity in anoxic to euxinic conditions is actually reflected in higher organic contents [Baudin et al., 1990] as well as stronger extinction rates in the north-European basins [Dera et al., 2010]. Also, it is clear that a strong restriction of northern basins was a major parameter for enhancing these disturbances [McArthur et al., 2008].
 Our simulations show that massive rises in atmospheric CO2 concentrations triggered major ecological disturbances during the T-OAE. Conjointly with an average global warming event of +4.5°C, the demise of polar sea ice and stronger high-latitude continental runoff rates could have freshened the Arctic surface seawaters, resulting in a thermohaline circulation collapse. This disruption was likely accountable for global bottom anoxia processes both in epicontinental seaways and deep oceanic settings. Influxes of boreal brackish seawaters through the Viking Corridor probably influenced the spatial disparity of δ18O variations recorded in the NW Tethyan domain and drastically strengthened regional productivity rates, water mass stratification, and anoxic conditions. Finally, our coupled modeling approach show that the rise in SST was heterogeneous across the globe and would have not exceeded +2 to +5°C in subtropical seas, attesting that changes inferred from NW Tethyan δ18O variations may be highly biased by changes in salinity and circulation patterns.
 We acknowledge V. Lefebvre, A. C. Chaboureau, and P. Sepulchre for the valuable technical assistance during the modeling work. We thank C. Bjerrum and B. van de Schootbrugge for their constructive and helpful reviews, as well as A. Caruthers for fruitful discussions and comments on the manuscript. The ANR Project Phylospace funded this work.