Stripping back the Modern to reveal the Cenomanian-Turonian climate and temperature gradient underneath 2

12 During past geological times, the Earth experienced several intervals of global warmth, but their driving 13 factors remain equivocal. A careful appraisal of the main processes controlling past warm events is essential to 14 inform future climates and ultimately provide decision makers with a clear understanding of the processes at 15 play in a warmer world. In this context, intervals of greenhouse climates, such as the thermal maximum of the 16 Cenomanian-Turonian (~94 Ma) during the Cretaceous period, are of particular interest. Here we use the IPSL- 17 CM5A2 Earth System Model to unravel the forcing parameters of the Cenomanian-Turonian greenhouse climate. 18 We perform six simulations with an incremental change in five major boundary conditions in order to isolate 19 their respective role on climate change between the Cenomanian-Turonian and the preindustrial. Starting with a 20 preindustrial simulation, we implement the following changes in boundary conditions: (1) the absence of polar 21 ice sheets, (2) the increase in atmospheric p CO 2 to 1120 ppm, (3) the change of vegetation and soil parameters, (4) the 1% decrease in the Cenomanian-Turonian value of the solar constant and (5) the Cenomanian-Turonian 23 paleogeography. Between the preindustrial simulation and the Cretaceous simulation, the model simulates a global warming of more than 11°C. Most of this warming is driven by the increase in atmospheric pCO 2 to 1120 25 ppm. Paleogeographic changes represent the second major contributor to global warming, whereas the 26 reduction in the solar constant counteracts most of geographically-driven warming. We further demonstrate that the implementation of Cenomanian-Turonian boundary conditions flattens meridional temperature gradients 28 compared to the preindustrial simulation. Interestingly, we show that paleogeography is the major driver of the flattening in the low- to mid-latitudes, whereas p CO 2 rise and polar ice sheet retreat dominate the high-latitude response.


102
In this study, we investigate the forcing parameters of CT greenhouse climate by using a set of   Pierre-Simon Laplace) within the CMIP5 framework (Dufresne et al., 2013). It is a fully-coupled Earth System 117 Model, which simulates the interactions between atmosphere, ocean, sea ice, and land surface. The model Its former version, IPSL-CM5A-LR, has a rich history of applications, including present-day and future climates 120 (Aumont and Bopp, 2006;Swingedouw et al., 2017) as well as preindustrial (Gastineau et    Six simulations were performed for this study: one preindustrial control simulation, named piControl, and 146 five simulations for which the boundary conditions were changed one at a time to progressively reconstruct the Functional Types (PFTs) and mean parameters for soil), 4X-NOICE-PFT-SOIL-SOLAR (previous changes + reduction 150 of the solar constant) and 4X-CRETACEOUS (previous changes + CT paleogeography). The piControl simulation 151 has been run for 1800 years and the five others for 2000 years in order to reach near-surface equilibrium (see

161
In the 4X simulations (i.e., all except piControl and 1X-NOICE), pCO2 is fixed to 1120 ppm (4 P.A.L), a value 162 reasonably close to the mean suggested by a recent compilation of CT pCO2 reconstructions (Wang et al., 2014).

163
In the 4X-NOICE-PFT-SOIL simulation, the distribution of the 13 standard PFTs defined in ORCHIDEE is 164 uniformly reassigned along latitudinal bands, based on a rough comparison with the preindustrial distribution of 165 vegetation, in order to obtain a theoretical latitudinal distribution usable for any geological period. The list of 166 PFTs and associated latitudinal distribution and fractions are described in Supplementary Table 1. Mean soil 167 parameters, i.e., mean soil color and texture (rugosity), are calculated from preindustrial maps (Zobler, 1999; 168 Wilson and Henderson-sellers, 2003) and uniformly prescribed on all continents. The impact of these idealized 169 PFTs and mean parameters is discussed in the results.

170
The 4X-NOICE-PFT-SOIL-SOLAR simulation is initialized from the same conditions as 4X-NOICE-PFT-SOIL 171 except that the solar constant is reduced to its CT value (Gough, 1981). We use here the value of 1353.36 W/m 2 172 (98.9% of the modern solar luminosity, calculated for an age of 90 My).

173
The 4X-CRETACEOUS simulation, finally, incorporates the previous modifications plus the implementation 174 of the CT paleogeography. The land-sea configuration used here is that proposed by Sewall (2007), in which we    warming or cooling associated to each change. The computational costs would be too high for this 215 study to explore this further here; it is an interesting problem that we leave for a future investigation.

216
The results presented in the following are averages calculated over the last 100 simulated years.

GLOBAL CHANGES 219
The progressive change of parameters made to reconstruct the CT climate induces a general 220 global warming (Table 2

225
Paleogeographic changes also represent a major contributor to the warming, leading to an increase in 226 T2M of 2.6°C. In contrast, the decrease in solar constant leads to a cooling of 1.8°C at the global scale.

238
3.2 The major contributor to global warming -DCO 2

239
As mentioned above, the fourfold increase in pCO2 leads to a global warming of 9°C (Table 3,

252
The contrast in the atmospheric response over continents and oceans is due to the impact of 253 the evapo-transpiration feedback. Oceanic warming drives an increase in evaporation, which acts as a 254 negative feedback and moderates the warming by consuming more latent heat at the ocean surface.

255
In contrast, high temperatures resulting from continental warming tend to inhibit vegetation 256 development, which acts as positive feedback and enhances the warming due to reduced 257 transpiration and reduced latent heat consumption.

294
The albedo and emissivity changes are linked to atmospheric and oceanic circulation (3) Increase in oceanic area in the North Hemisphere (Fig 2)

299
(4) Decrease in oceanic area in the South Hemisphere (Fig 2) 300 301 In the CT simulation, we observe an intensification of the meridional surface circulation and Toggweiler, 2003) and that leads to the formation of a strong circumglobal equatorial current (Fig 7b) .

308
This connection permits the existence of stronger easterly winds that enhance equatorial upwelling

312
Circumpolar Current (ACC) during the Cretaceous (Fig 7c-d). Notwithstanding, the observed increase 313 in southward OHT between 40° and 60°S (Fig 8c) is explained by the absence of significant zonal

323
The atmosphere's response to the paleogeographic changes in the mid-and high-latitudes is 324 different in the Southern and Northern Hemispheres because the ocean to land ratio varies between 325 the CT configuration and the modern. In the Southern Hemisphere, the reduced ocean surface area in 326 the CT simulation (Fig 2) limits evaporation and moisture injection into the atmosphere, which in turn 327 leads to a decrease in relative humidity and low-altitude cloudiness (Supplementary Fig S8) and 328 associated year-round warming due to reduced planetary albedo. In the Northern Hemisphere, 329 oceanic surface area increases (Fig 2) and results in a strong increase in evaporation and moisture 330 injection into the atmosphere. Low-altitude cloudiness and planetary albedo increase and lead to increase is consistent with the simulated increase in extratropical OHT (Fig. 8). The mean annual global SST increases as much as 9.8°C (from 17.9°C to 27.7 °C) across the 343 simulations. The SST warming is slightly weaker than that of the mean annual global atmospheric 344 temperature at 2m discussed above, and most likely occurs because of evaporation processes due to 345 the weaker atmospheric warming simulated above oceans compared to that simulated above 346 continents. Unsurprisingly, as for the atmospheric temperatures, pCO2 is the major controlling 347 parameter of the ocean warming (7°C), followed by paleogeography (4.5°C) and changes in the solar 348 constant (2.3°C), although the latter induces a cooling rather than a warming. PFT and soil parameter 349 changes and the removal of polar ice sheets have a minor impact at the global SST (0.6 °C and 0°C 350 respectively). It is interesting to note the increased contribution of paleogeography to the simulated 351 SST warming compared to its contribution to the simulated atmospheric warming, which is probably 352 driven by the major changes simulated in surface ocean circulation (Fig. 7).

353
Mean annual SST in the preindustrial simulation reach ~ 26°C in the tropics (calculated as the 354 zonal average between 30°S and 30°N) and ~ -1.5°C at the poles (beyond 70° N -Fig 9a). In this work,

361
The progressive flattening of the SST gradient can be visualized by superimposing the zonal 362 mean temperatures of the different simulations and by adjusting them at the Equator (Fig 9b). Two 363 major observations can be drawn from these results. First, paleogeography has a strong impact on the

382
As for the SST gradients, we plot atmospheric meridional gradients by adjusting temperature 383 values so that temperatures at the Equator are equal for each simulation (Fig 9d). This normalization   The results predicted by our CT simulation can be compared to reconstructions of 397 atmospheric and oceanic paleotemperatures inferred from proxy data (Fig 10a, b). Our SST data paleobotanical and paleosoil studies (see Supplementary data for the complete database and 401 references).

402
The Cretaceous equatorial and tropical SST have long been believed to be similar or even

431
Hemisphere reveal reasonable agreement with data whereas Northern Hemisphere mean zonal 432 temperatures in our model are slightly warmer than that inferred from proxies (Fig 10b). At high-433 latitudes, the same trend is observed for atmospheric temperatures as it is for SST with data indicating

RECONSTRUCTED LATITUDINAL TEMPERATURE GRADIENTS 438
The simulated northern hemisphere latitudinal SST gradient of (~0.45°C/°latitude) is in good 439 agreement with those reconstructed from paleoceanographic data in the Northern Hemisphere 440 (~0.42°C/°latitude) whereas it is much larger in the Southern Hemisphere (~0.39°C/°latitude vs 441 ~0.3°C/°latitude) (Fig 11). This overestimate of the latitudinal gradient holds for the atmosphere as   (Hay et al., 2019). Finally, from a proxy perspective, it has been suggested that a 470 sampling bias could exist, with a better record of temperatures during the warm season at high 471 latitudes and during the cold season in low latitudes (Huber, 2012). Such possible biases would help 472 reduce the model-data discrepancy, in particular for atmospheric temperatures (Fig 10b), as high-473 latitude reconstructed temperatures are more consistent with simulated summer temperatures 474 whereas the consistency is better with simulated winter temperatures in the mid-to low-latitudes, but 475 further work is required to unambiguously demonstrate the existence of these biases.

576
The authors declare that they do not have competing interests.

AKNOWLEDGMENTS 578
We express our thanks to Total E&P for funding the project and granting permission to publish. We