Stripping back the Modern to reveal Cretaceous climate and temperature gradient 1 underneath 2

13 During past geological times, the Earth suffered several intervals of global warmth but their driving 14 factors remain equivocal. A careful appraisal of the main processes involved in those past events is essential to 15 evaluate how they can inform future climates, and thus to provide decision makers with a clear understanding of 16 the processes at play in a warmer world. In this context, the greenhouse Earth of the Cretaceous era, specifically 17 the Cenomanian-Turonian (~94 Ma), is of particular interest, as it corresponds to a thermal maximum. Here we 18 use the IPSL-CM5A2 Earth System Model to unravel the forcing parameters of the Cenomanian-Turonian greenhouse climate. We perform six simulations with an incremental change in five major boundary conditions in 20 order to isolate their respective role on climate change between the Cretaceous and the preindustrial. Starting 21 with a preindustrial simulation, we implement: (1) the absence of polar ice sheets, (2) the increase in 22 atmospheric p CO 2 to 1120 ppm, (3) the change of vegetation and soil parameters, (4) the 1% decrease in the 23 Cenomanian-Turonian value of the solar constant and (5) the Cenomanian-Turonian paleogeography. Between 24 the first (preindustrial) simulation and the last (Cretaceous) simulation, the model simulates a global warming of 25 more than 11°C. Most of this warming is driven by the increase in atmospheric pCO 2 to 1120 ppm. 26 Paleogeographic changes represent the second major contributor to the global warming while the reduction in 27 the solar constant counteracts most of the geographically-driven global warming. We also demonstrate that the 28 implementation of Cretaceous boundary conditions flattens the temperature gradients compared to the 29 piControl simulation. Interestingly, we show that paleogeography is the major driver of the flattening in the low- 30 to mid-latitudes whereas the p CO 2 rise and polar ice sheet retreat dominate the high-latitudes response. Two major observations can be made from these results. First, paleogeography has a strong impact on the low-latitudes SST gradient because it widens the latitudinal band of relatively homogeneous warm tropical SST). As explained before (See Results – D paleogeography), this is due to the opening of 329 equatorial gateways. Second, the SST gradient beyond 40° of latitude is flattened in two steps with paleogeography being the major contributor followed by atmospheric p CO 2 increase. we incrementally 493 implement changes in boundary conditions on a pre-industrial simulation to obtain in the end a 494 simulation of the Cenomanian-Turonian stage of the Cretaceous. This study confirms the primary 495 control exerted by atmospheric p CO 2 on atmospheric temperatures, with a contribution of 61% to the 496 total absolute global warming. At the global scale, paleogeographic and solar constant changes have 497 opposite effects, canceling each other, while polar ice cap retreat and vegetation and soil parameter 498 changes have only minor impact. Atmospheric p CO 2 still explains the majority of the global SST 499 warming (49%) but the amount of change explained by paleogeography increases compared to the 500 atmospheric temperature change and thus represents a major contribution (30%). The study of 501 temperature gradients reveals that the reduction of the meridional SST gradients between the 502 preindustrial and the Cretaceous is mainly due to the paleogeographic changes and to a lesser extent 503 to the increase of p CO 2 . The atmospheric gradient response is more complex because its flattening is 504


INTRODUCTION 32
The Cretaceous era is of particular interest to understand the drivers of past greenhouse climates

49
The main factor generally considered as responsible for the Cretaceous global warmth is the higher level and the NEMO ocean model (Madec, 2012) including the LIM2 sea-ice model (Fichefet and Maqueda, 1997) and  Table 1. The soil parameters, i.e., the mean color and texture (rugosity), are reduced to its Cretaceous value (Gough, 1981

179
The simulated changes between the preindustrial simulation (piControl) and the Cretaceous     Fig 4c), via contrasted regional responses (Dice or Dpaleo -Fig 4b and 4f). In the next section,

207
NOICE-and 4X-NOICE simulations. The whole surface is warmer with an amplification located over the 208 Arctic and Austral oceans and which is generally larger over continents than over oceans (Fig 4c). The

209
warming is due to a general decrease of planetary albedo and of the atmosphere's emissivity. The 210 decrease in atmosphere's emissivity is directly driven by the increase of CO2, and thus greenhouse 211 trapping in the atmosphere, but it is also amplified by an increase in high-altitude cloudiness over the 212 Antarctic continent (Fig 5a,b). The decrease in planetary albedo is due to (1) a decrease of sea ice and  decrease. Nevertheless, its impact is minor given the strong atmospheric temperature warming 225 observed over the Arctic (Fig 4c).

226
The contrast in the response of the atmosphere over continents and oceans is due to the

242
The change in soil parameters and the implementation of theoretical zonal PFTs, in simulation 243 4X-NOICE-PFT-SOIL, drive a warming of 0.8 °C. This warming is essentially located above arid areas, 244 such as the Sahara, Australia, or the Middle-East, and polar latitudes (Antarctica/Greenland) (Fig 4d).

245
The warming above arid areas is mostly caused by the implementation of a mean uniform soil color,

267
(3) Increase of oceanic area in the North Hemisphere (Fig 2)

268
(4) Decrease of oceanic area in the South Hemisphere (Fig 2) 269 270 The opening of equatorial gateway creates a zonal connection between the Pacific, Atlantic

288
The atmosphere response to paleogeographic change in the mid-and high-latitudes is 289 different in the Southern and Northern Hemispheres as the oceanic areas decrease or increase 290 between the CT configuration and the modern. In the Southern Hemisphere, the reduced ocean 291 surface area (Fig 2) limits evaporation and moisture injection into the atmosphere, which in turn leads 292 to a decrease in relative humidity and low-altitude cloudiness ( Supplementary Fig 1) and an associated 293 year-round warming due to a reduced planetary albedo. In the Northern Hemisphere, the oceanic 294 area increases (Fig 2) and results in a strong increase of evaporation and moisture injection into the 295 atmosphere. Low-altitude cloudiness and hence the albedo, both increase and lead to the cooling 296 during the summer as discussed above (Fig 6). During winter, on the other hand, an increase of high-297 altitude cloudiness leads to an enhanced greenhouse effect and counteracts the larger albedo. This high-altitude cloudiness increase is due to the extratropical increase in OHT (Fig. 8)

307
This warming is slightly weaker than the mean annual global atmospheric temperature at 2m 308 discussed above, and most likely occurs because of evaporation processes due to the weaker 309 atmospheric warming above oceans compared to that above continents. As for atmospheric 310 temperatures, pCO2 appears as the major controlling parameter of the ocean warming (49% of the 311 absolute temperature change), followed by paleogeography (30%) and solar constant (16%), although 312 the latter again drives cooling rather than warming. PFT and soil parameter changes and polar ice cap 313 retreat instead have a minor impact at the global scale (4% and 0% respectively). It is interesting to 314 note the increased contribution of paleogeography in the simulated sea surface warming compared to 315 the atmospheric warming, which is probably driven by the major changes simulated in the surface 316 circulation (Fig. 7).

339
Hemisphere. This yields latitudinal temperature gradients of 0.49°C/°latitude and 0.54°C/°latitude, 340 respectively. As for the SST gradients, we normalized the curves so that temperatures at the Equator 341 are equal for each simulation (Fig 9d). This shows that the different mechanisms responsible for the 342 flattening of the gradients are different for each hemisphere. In the south, at high-latitudes, three 343 parameters contribute to reducing the equator-to-pole temperature gradient in the following order of   The results predicted by our CT simulation were compared to the reconstructed atmospheric 369 and oceanic paleotemperatures from proxy data (Fig 10a,b). The SST data compilation is essentially

385
In the mid-latitudes (30-60°) the proxy records show a wide range of SST, ranging from 10°C to 386 more than 30°C. We observe that this trend can be reproduced in our simulation when considering 387 the local monthly maximum and minimum temperatures (grey shaded areas , Fig 10a), suggesting a 388 reasonable model-data agreement. Simulated atmospheric temperatures for these latitudes in the 389 Southern hemisphere also show reasonable agreement, whereas the Northern Hemisphere mean 390 zonal temperatures in our model are slightly warmer than that inferred from proxies (Fig 10b).

391
There are unfortunately only a few high-latitudes SST data points available, which renders the 392 model-data comparison difficult. This is further aggravated by both the proxy-based and simulated SST       (Hay et al., 2019). Finally, from a proxy 432 perspective, it was suggested that a sampling bias could exist, with a better record of temperatures 433 during the warm season at high latitudes and during the cold season in low latitudes (Huber, 2012). temperatures (Fig 10b), as high-latitude reconstructed temperatures are more consistent with 436 simulated summer temperatures whereas the consistency is better with simulated winter 437 temperatures in the mid-to low-latitudes, but more work is required to unambiguously demonstrate 438 the existence of these biases.

467
It has been suggested that high latitude warming, and an associated reduced meridional SST

521
The authors declare that they do not have competing interests.

523
We express our thanks to Total E&P for funding the project and granting permission to publish. We