Land-atmosphere interactions in sub-polar and alpine climates in the CORDEX FPS 1 LUCAS models: I. Evaluation of the snow-albedo effect 2

1. CICERO Center for International Climate Research, Oslo, Norway 7 2. NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway 8 3. International Center for Theoretical Physics, Trieste, Italy 9 4. Climate Service Center Germany, Helmholtz-Zentrum Hereon, Hamburg, Germany 10 5. Wyss Academy for Nature, Climate and Environmental Physics, Oeschger Center for Climate Change Research, 11 University of Bern, Bern, Switzerland 12 6. Department of Meteorology and Climatology, School of Geology, Aristotle University of Thessaloniki, 13 Thessaloniki, Greece 14 7. Laboratoire des Sciences du Climat et de l’environnement, Paris, France 15 8. Department of Atmospheric Physics, Faculty of Mathematics and Physics, Charles University, Prague, Czech 16 Republic 17 9. Institute for Meteorology and Climate Research, Karlsruhe Institute of Technology, Karlsruhe, Germany 18 10. Instituto Dom Luiz, Faculdade de Ciências da Universidade de Lisboa, 1749-016 Lisboa, Portugal 19 11. Swedish Meteorological and Hydrological Institute, Norrkoping, Sweden 20 12. Center for Environmental Systems Research, University of Kassel, Germany 21 13. Department of Geography, Ludwig-Maximilians-Universität, Munich, Germany. 22 23 24 Corresponding author: Anne Sophie Daloz (anne.sophie.daloz@cicero.oslo.no) 25


Abstract 28
In the Northern Hemisphere, the seasonal snow cover plays a major role in the climate system via its 29 effect on surface albedo and fluxes. The parameterization of snow-atmosphere interactions in climate 30 models remains a source of uncertainty and biases in the representation of the local and global climate.

31
Here, we evaluate the ability of an ensemble of regional climate models (RCMs) coupled to different 32 land surface models to simulate the snow albedo effect over Europe, in winter and spring. We use a 33 previously defined index, the Snow Albedo Sensitivity Index (SASI), to quantify the radiative forcing

52
In particular, snow can have a strong impact on climate due to its high albedo, primarily because of the 53 contrast in the surface energy balance between snow-covered and snow-free land surfaces (

79
Here, we investigate the ability of an ensemble of RCMs to represent snow cover and the 80 radiative forcing from the snow albedo effect (SASI) over Europe, including a comparison between mid-81 and high-latitude regions. We derive SASI using radiative fluxes and snow cover from satellites, 82 reanalysis and model outputs. Building on findings by Xu Jacob et al., 2020) and it enables us to perform a broader 91 assessment of several RCMs within a consistent framework. Our assessment is carried out in two parts 92 and published in companion articles. In Part I, we investigate the ability of these RCMs to represent the examined in the companion paper (Part II) while here, we use ten models from the EVAL experiment 98 only, which employ their standard land use and land cover maps.

99
Section 2 introduces the modeling and observational datasets used in this study as well as the 100 derivation of SASI, while Section 3 examines and discusses the ability of climate models to represent 101 SASI compared with satellite observations and reanalyses, focusing on the strength and timing of the 102 signal. Further, the origin of the differences between the models are explored by evaluating potential 103 common biases in the ensemble of simulations as well as individual model biases. The analysis also 104 explores the differences in SASI between mid-and high-latitude regions, opening the discussion on the 105 impacts of different land cover for the simulation of SASI, which will be further explored in Part II.

106
Finally, Section 4 the last sections offer some concluding remarks.

177
For MODIS data, the following processing steps are applied: 1. Data are masked according to the prevailing cloud cover since high cloud cover prevents a 179 correct estimation of snow cover. We apply two different thresholds (20% and 50%) to the 180 percent of clouds in each cell. 181 2. Only data flagged as "best", "good", and "ok" are used while all other data are masked.

187
Only grid points with more than 50% land fraction are included.

188
The masking for MODIS data implies that single grid points can contribute differently to the average 189 over one region. To make the models and reanalyses comparable, each grid point is weighted by the

276
The LUCAS simulations also show a pronounced peak in SASI in all regions (Fig. 3)

Inter-model differences in SASI 307
To better understand the origin of the differences in SASI across RCMs, we explore the 308 relationship between SASI and its components, surface snow cover and shortwave radiation, during the 309 accumulation and ablation periods.

374
The results show that climate models are able to reproduce some of the SASI characteristics 375 (e.g. existence of a peak, amplitude of the peak) compared to reanalysis and satellite observations 376 (Section 3.1), even if large differences appear between the RCMs. The climate models' ability to 377 represent SASI is highly related to their representation of snow cover (Section 3.3), which can be 378 difficult to represent for climate models (Matiu et al., 2020). Our results also suggest that the models' 379 capability highly differs between the accumulation and ablation periods. Most models have much lower 380 agreement with reanalyses and satellite observations in the ablation period, with some exceptions (e.g. 381 CCLM-CLM5.0 over East Europe), indicating a common bias regarding snow cover in spring, pointing 382 towards a bias from LSMs. This bias seems to be common to most LSMs even if they are based on 383 different assumptions and parameterizations (see Section 2.3). It is also interesting that even though 384 CCLM-TERRA is not as advanced in terms of snow modeling compared to the other models (e.g.

385
Section 2.1.3), it still manages to represent SASI reasonably well over Europe. In addition, the 386 representation of the sub-grid scale surface heterogeneity (Table 1;