Snow cover over the Iberian mountains in km-scale global climate simulations: evaluation and projected changes

TL;DR

This study evaluates the Iberian snow climatology simulated by a km-scale global model (IFS-FESOM) within nextGEMS, assessing its ability to reproduce historical snow patterns and to project future changes under SSP3-7.0. Using multiple high-resolution references (ERA5-Land, CERRA, IPE-CSIC, AEMET, MODIS), the work shows the model captures key elevation-dependent features and seasonal cycles, albeit with a positive bias in snow-cover days at lower elevations. Under SSP3-7.0, the simulations project a substantial reduction in snow-cover duration across Iberia, most pronounced at low-to-mid elevations and in southern ranges, driven primarily by fewer snowfall days due to warming and reduced precipitation, with regional variations in the drivers. The results demonstrate that storm-resolving, km-scale global simulations can provide valuable local-scale snow information and near-term climate projections without regional downscaling, supporting adaptation planning in complex terrain.

Abstract

Mountains represent a crucial natural resource for many regions worldwide, particularly in Mediterranean areas such as the Iberian Peninsula, where climate change can have profound consequences. However, due to their coarse resolution, traditional climate models cannot properly represent the complexity of orographic processes. Storm-resolving models, with horizontal resolutions finer than 10 km, offer a path forward to provide high-quality information for local adaptation even in complex terrain. In this work, we assess the performance of one such model, the IFS-FESOM developed within the EU-NextGEMS project, in simulating the historical seasonal snow cover in the main Iberian mountain ranges and projecting climate change effects. It is evaluated against four high-resolution reanalysis-based products, satellite, and in situ observations. Despite a positive bias in the length of the snow cover season and the number of snowfall days, the model can represent features of the snow climatology, such as its elevation dependency, seasonal cycle, and mean covered area, successfully within the range of uncertainty of the reference data. In addition, the projection under the SSP3-7.0 scenario shows a marked reduction of the snow season across almost all of Iberia, with low- to mid-elevations and the southernmost ranges showing the greatest reductions. These changes are linked to reduced snowfall days, caused primarily by rising temperatures but also by decreasing precipitation, particularly in southern and Mediterranean sectors. Our results suggest that storm-resolving simulations can provide local climate projections of snow information in complex terrain without regional downscaling.

Figures (12)

• Figure 1: Comparison between historical simulations (1991-2014) of CMIP6 model MIROC6 (a,c) and nextGEMS IFS-FESOM model (b,d). Mean extended winter (NDJFMA) 2-meter temperature (a,b) and number of days with a snow cover over 1 cm (c,d). Note that c and d feature different color-bar scales.
• Figure 2: IFS-FESOM model orography interpolated to a regular 0.125º x 0.125º latitude-longitude grid. The four mountain regions that are studied in this work are highlighted, enclosed in rectangular areas that are later considered for analysis.
• Figure 3: Mean number of days with a snow cover over 1 cm during NDJFMA versus grid point altitude. Each dot corresponds to a different grid point of the common 0.125º lat-lon grid. Each subplot corresponds to the grid points enclosed in the rectangular areas around each of the main mountain ranges, as shown in Figure \ref{['fig: orog_ifs_zones']}.
• Figure 4: Maps of the mean number of days with a snow cover over 1 cm during NDJFMA for the reference datasets (a-d) and mean bias of the IFS historical simulation with respect to each reference dataset (e-h). The period considered is 1990 to 2019 (except for the IPE dataset which ends in 2014). Statistical insignificance ($\alpha_{FDR}=0.05$) of the bias is signaled by line hatching.
• Figure 5: Spatial distribution of the mean NDJFMA bias of snowfall (sf), total precipitation (tp), minimum 2-meters temperature (Tmin), and 2-meters relative humidity (RH) for the main Iberian mountain ranges as defined in Figure \ref{['fig: orog_ifs_zones']}. The mean bias is calculated for each grid point with respect to the CERRA reanalysis and the results are shown as a boxplot. Only grid points over 700 m of altitude are considered.