Evaluating Snow in EURO-CORDEX Regional Climate Models with Observations for the European Alps: Biases and Their Relationship to Orography, Temperature, and Precipitation Mismatches

Abstract

Climate models are important tools to assess current and future climate. While they have been extensively used for studying temperature and precipitation, only recently regional climate models (RCMs) arrived at horizontal resolutions that allow studies of snow in complex mountain terrain. Here, we present an evaluation of the snow variables in the World Climate Research Program Coordinated Regional Downscaling Experiment (EURO-CORDEX) RCMs with gridded observations of snow cover (from MODIS remote sensing) and temperature and precipitation (E-OBS), as well as with point (station) observations of snow depth and temperature for the European Alps. Large scale snow cover dynamics were reproduced well with some over-and under-estimations depending on month and RCM. The orography, temperature, and precipitation mismatches could on average explain 31% of the variability in snow cover bias across grid-cells, and even more than 50% in the winter period November-April. Biases in average monthly snow depth were remarkably low for reanalysis driven RCMs (<approx. 30 cm), and large for the GCM driven ones (up to 200 cm), when averaged over all stations within 400 m of altitude difference with RCM orography. Some RCMs indicated low snow cover biases and at the same time high snow depth biases, and vice versa. In summary, RCMs showed good skills in reproducing alpine snow cover conditions with regard to their limited horizontal resolution. Detected shortcomings in the models depended on the considered snow variable, season and individual RCM.Atmosphere 2020, 11, 46 2 of 20 resolution and coverage. It should be noted that RCMs have been run also at higher resolutions (2.2 km with a convection permitting scheme [2]) and employed for snow cover studies [3]. Due to high computational requirements, such simulations are still rare (i.e., not available as model ensemble) and do not cover longer temporal ranges (e.g., like the full 21st century).Large-scale and high-resolution validation of snow in RCMs is hampered, because there are only few observational data sets, especially for snow water equivalent (SWE). Nevertheless, there have been regional attempts. Salzmann and Mearns compared multiple RCMs at 50 km resolution with station observations and reanalysis in the Upper Colorado River basin in the USA and found that RCMs simulated too little SWE with a too-late start and too-early end of snow cover, most likely because of unresolved topography and model schemes [4]. Steger et al. compared RCMs at 25 km with a gridded SWE product for Switzerland and found that, in general, RCMs reproduced spatial and seasonal variability but underestimated SWE at low elevation and overestimated SWE at high elevations [5]. Da Ronco et al. compared snow cover in the RCM COSMO-CLM with remote sensing observations from MODIS for the Po river basin in Italy and found a good agreement regarding duration, extent, and interannual variability, but a too-early snow melt in the RCM. Additionally, they also highlighted the importance of horizontal resolution [6]...

Evaluation of Current Trends of Climatic Actions in Europe Based on Observations and Regional Reanalysis

“…Although specific studies regarding effects of climate change on temperature and precipitation are widespread [53,55], a limited number of studies focus on snow [7,8]. In fact, on the one hand, the outputs of global climate models (GCMs) have too coarse a spatial resolution; on the other, regional climate models (RCMs) have only recently reached reasonable horizontal resolutions to be suitable to evaluate changes in snow parameters, greatly affected by local scale phenomena [56]. A first evaluation of the snow variables in the EURO-CORDEX initiative [52,53] can be found in [56], while an original procedure to derive ground snow load from available climate projections is described in [7,8].…”

“…In fact, on the one hand, the outputs of global climate models (GCMs) have too coarse a spatial resolution; on the other, regional climate models (RCMs) have only recently reached reasonable horizontal resolutions to be suitable to evaluate changes in snow parameters, greatly affected by local scale phenomena [56]. A first evaluation of the snow variables in the EURO-CORDEX initiative [52,53] can be found in [56], while an original procedure to derive ground snow load from available climate projections is described in [7,8]. A procedure for the estimation of future trends of climatic actions based on high-resolution RCMs projection was proposed by the authors in [9].…”

Extreme Ground Snow Loads in Europe from 1951 to 2100

“…Regarding climate projections of snow variables, some inconsistencies were already identified in previous studies for high altitude cells [14], usually above 1500 m. For example, the snow accumulation process may lead to cells characterized by constant snow cover, and unrealistic snow accumulation. This result is not surprising since the resolution and the refinement of the climate model does not allow to capture huge local variations; for this reason, the grid cells affected by unrealistic snow accumulation, higher than 70 kN/m 2 according to [14], were disregarded in the elaboration of RCMs outputs.…”

“…Aiming at evaluating the evolution of snow load on ground in Europe, the climate model data produced by the EURO-CORDEX initiative [25,26], which are provided by the Earth System Grid Federation (ESGF), are considered in the following, focusing on snow variables [14]. Daily snow water equivalent data were collected from the combination of nine global and regional climate models as described in Table 1.…”

“…Impacts of climate changes on temperature and precipitation are commonly assessed using climate projections [2][3][4][5], while little has yet been done for snow [14], although it represents a significant climate variable. In addition to the complexity of the snow accumulation phenomena, it must be remarked that, on the one hand, changes in snow parameters cannot be appreciated by using only general or global circulation models (GCMs), having low grid resolution (>100 km) [14]. On the other hand, regional climate models (RCMs) with adequate grid resolutions (10-50 km) have been developed only in recent time.…”