Model-dependent latitudinal temperature gradient drives Late Ordovician climate stability

Among the five great extinction events of the Phanerozoic, the Late Ordovician stands out as it isconcomitant with a massive glacial event under high atmospheric pCO2. This apparent climateparadox was addressed in numerous climate modeling studies. In particular, [1] showed that underthe specific palaeogeographical conditions of the Hirnantian (445 Ma), with an ocean-dominatedNorthern Hemisphere, the climate system may undergo a “tipping point” where a small pCO2variation leads to either glacial or ice-free warm equilibrium state.Those results were obtained with the intermediate complexity Fast Ocean Atmosphere Model(FOAM). We have conducted new simulations using the state-of-the-art coupled IPSL-CM5A2-LREarth System Model [2], spanning a wide range of pCO2 for the Hirnantian. We find that the climatetipping point is entirely absent, and that the equilibrium climate sensitivity is strikingly linear in thisset of simulations.We conducted a detailed model intercomparison and we have identified major differences betweenthe models in the representation of the radiative transfer, cloud cycle and oceanic eddy dynamicswhich contribute to the qualitatively different model behaviors, enhanced under high atmosphericpCO2 content. Specifically, the FOAM tipping point corresponds to an abrupt transition from a sharpNorthern latitudinal temperature gradient at low pCO2 (cold state) to a flattened gradient with warmpolar latitudes (ice-free warm state). In contrast, the IPSL-CM5A2 temperature gradient is relativelyconstant across pCO2, with year-long sea ice confined in the Northern latitudes even under 15Xpreindustrial pCO2 level (4200 ppm).We propose a physical mechanism to link the warm FOAM flattened latitudinal temperature gradientto the dramatic sea-ice albedo feedback sensitivity via the increased stratification of the superficialocean. Since this mechanism is independent of the physical parameterizations and relativecomplexity of the models, and comparing our results with other scarce published climate simulationsof the Hirnantian [3,4], we propose that the latitudinal temperature gradient, seen as a model-dependent emerging feature, may be the main driver of the previously unveiled sea-ice albedoclimate tipping point.References:[1] Pohl et al. (2014), Climate of the Past, 10, 6[2] Sepulchre et al. (2020), Geoscientific Model Development, 13,7[3] Pohl et al. (2017), Paleoceanography, 32, 4[4] Valdes et al. (2021), Climate of the Past, 17, 4