The European Alps stretch over a range of climate zones which affect the spatial distribution of snow. Previous analyses of station observations of snow were confined to regional analyses. Here, we present an Alpine-wide analysis of snow depth from six Alpine countries – Austria, France, Germany, Italy, Slovenia, and Switzerland – including altogether more than 2000 stations of which more than 800 were used for the trend assessment. Using a principal component analysis and k-means clustering, we identified five main modes of variability and five regions which match the climatic forcing zones: north and high Alpine, north-east, north-west, south-east, and south and high Alpine. Linear trends of monthly mean snow depth between 1971 and 2019 showed decreases in snow depth for most stations from November to May. The average trend among all stations for seasonal (November to May) mean snow depth was −8.4 % per decade, for seasonal maximum snow depth −5.6 % per decade, and for seasonal snow cover duration −5.6 % per decade. Stronger and more significant trends were observed for periods and elevations where the transition from snow to snow-free occurs, which is consistent with an enhanced albedo feedback. Additionally, regional trends differed substantially at the same elevation, which challenges the notion of generalizing results from one region to another or to the whole Alps. This study presents an analysis of station snow depth series with the most comprehensive spatial coverage in the European Alps to date.

We present a novel approach to simulate and reconstruct annual glacier-wide surface mass balance (SMB) series based on a deep artificial neural network (ANN; i.e. deep learning). This method has been included as the SMB component of an open-source regional glacier evolution model. While most glacier models tend to incorporate more and more physical processes, here we take an alternative approach by creating a parameterized model based on data science. Annual glacier-wide SMBs can be simulated from topo-climatic predictors using either deep learning or Lasso (least absolute shrinkage and selection operator; regularized multilinear regression), whereas the glacier geometry is updated using a glacier-specific parameterization. We compare and cross-validate our nonlinear deep learning SMB model against other standard linear statistical methods on a dataset of 32 French Alpine glaciers. Deep learning is found to outperform linear methods, with improved explained variance (up to +64 % in space and +108 % in time) and accuracy (up to +47 % in space and +58 % in time), resulting in an estimated r² of 0.77 and a root-mean-square error (RMSE) of 0.51 m w.e. Substantial nonlinear structures are captured by deep learning, with around 35 % of nonlinear behaviour in the temporal dimension. For the glacier geometry evolution, the main uncertainties come from the ice thickness data used to initialize the model. These results should encourage the use of deep learning in glacier modelling as a powerful nonlinear tool, capable of capturing the nonlinearities of the climate and glacier systems, that can serve to reconstruct or simulate SMB time series for individual glaciers in a whole region for past and future climates.

Most permafrost is located in the Arctic, where frozen organic carbon makes it an important component of the global climate system. Despite the fact that the Arctic climate changes more rapidly than the rest of the globe, observational data density in the region is low. Permafrost thaw and carbon release to the atmosphere are a positive feedback mechanism that can exacerbate global warming. This positive feedback functions via changing land–atmosphere energy and mass exchanges. There is thus a great need to understand links between the energy balance, which can vary rapidly over hourly to annual timescales, and permafrost, which changes slowly over long time periods. This understanding thus mandates long-term observational data sets. Such a data set is available from the Bayelva site at Ny-Ålesund, Svalbard, where meteorology, energy balance components and subsurface observations have been made for the last 20 years. Additional data include a high-resolution digital elevation model (DEM) that can be used together with the snow physical information for snowpack modeling and a panchromatic image. This paper presents the data set produced so far, explains instrumentation, calibration, processing and data quality control, as well as the sources for various resulting data sets. The resulting data set is unique in the Arctic and serves as a baseline for future studies. The mean permafrost temperature is −2.8 °C, with a zero-amplitude depth at 5.5 m (2009–2017). Since the data provide observations of temporally variable parameters that mitigate energy fluxes between permafrost and atmosphere, such as snow depth and soil moisture content, they are suitable for use in integrating, calibrating and testing permafrost as a component in earth system models.The presented data are available in the Supplement for this paper (time series) and through the PANGAEA and Zenodo data portals: time series (https://doi.org/10.1594/PANGAEA.880120, https://zenodo.org/record/1139714) and HRSC-AX data products (https://doi.org/10.1594/PANGAEA.884730, https://zenodo.org/record/1145373).

The shortage of information on snow properties in high latitudes places a major limitation on permafrost and more generally climate modelling. A dedicated field program was therefore carried out to investigate snow properties and their spatial variability at a polygonal tundra permafrost site. Notably, snow samples were analysed for surface-normal thermal conductivity (K_{eff − z}) based on X-ray microtomography. Also, the detailed snow model SNOWPACK was adapted to these Arctic conditions to enable relevant simulations of the ground thermal regime. Finally, the sensitivity of soil temperatures to snow spatial variability was analysed. Within a typical tundra snowpack composed of depth hoar overlain by wind slabs, depth hoar samples were found more conductive (K_{eff − z} = 0.22±0.05 W m⁻¹ K⁻¹) than in most previously published studies, which could be explained by their high density and microstructural anisotropy. Spatial variations in the thermal properties of the snowpack were well explained by the microtopography and ground surface conditions of the polygonal tundra, which control depth hoar growth and snow accumulation. Our adaptations to SNOWPACK, phenomenologically taking into account the effects of wind compaction, basal vegetation, and water vapour flux, yielded realistic density and K_{eff − z} profiles that greatly improved simulations of the ground thermal regime. Also, a density- and anisotropy-based parameterization for K_{eff − z} lead to further slight improvements. Soil temperatures were found to be particularly sensitive to snow conditions during the early winter and polar night, highlighting the need for improved snow characterization and modelling over this period.

The main objective of this study was to monitor wet snow conditions from Sentinel-1 over a season, to examine its variation over time by cross-checking wet snow with independent snow and weather estimates, and to study its distribution taking into account terrain characteristics such as elevation, orientation, and slope. One of our motivations was to derive useful representations of daily or seasonal snow changes that would help to easily identify wet snow elevations and determine melt-out days in an area of interest. In this work, a well-known approach in the literature is used to estimate the extent of wet snow cover continuously over a season and an analysis of the influence of complex mountain topography on snow distribution is proposed taking into account altitude, slope, and aspect of the terrain. The Sentinel-1 wet snow extent product was compared with Sentinel-2 snow products for cloud free scenes. We show that while there are good agreements between the two satellite products, differences exist, especially in areas of forests and glaciers where snow is underestimated. This underestimation must be considered alongside the areas of geometric distortion that were excluded from our study. We analysed retrievals at the scale of our study area by examining wet snow Altitude–Orientation diagrams for different classes of slopes and also wet snow Altitude–Time diagrams for different classes of orientations. We have shown that this type of representation is very useful to get an overview of the snow distribution as it allows to identify very easily wet snow lines for different orientations. For an orientation of interest, the Altitude–Time diagrams can be used to track the evolution of snow to locate altitudes and dates of snow loss. We also show that ascending/descending Sentinel-1 image time series are complementary to monitor wet snow over the French alpine areas to highlight wet snow altitude ranges and identify melt-out days. Links have also been made between Sentinel-1 responses (wet snow) and snow/meteorological events carefully listed over the entire 2017–2018 season.

A realistic simulation of snow cover and its thermal properties are important for accurate modelling of permafrost. We analyse simulated relationships between air and near-surface (20 cm) soil temperatures in the Northern Hemisphere permafrost region during winter, with a particular focus on snow insulation effects in nine land surface models, and compare them with observations from 268 Russian stations. There are large cross-model differences in the simulated differences between near-surface soil and air temperatures (ΔT; 3 to 14 °C), in the sensitivity of soil-to-air temperature (0.13 to 0.96 °C °C⁻¹), and in the relationship between ΔT and snow depth. The observed relationship between ΔT and snow depth can be used as a metric to evaluate the effects of each model's representation of snow insulation, hence guide improvements to the model's conceptual structure and process parameterisations. Models with better performance apply multilayer snow schemes and consider complex snow processes. Some models show poor performance in representing snow insulation due to underestimation of snow depth and/or overestimation of snow conductivity. Generally, models identified as most acceptable with respect to snow insulation simulate reasonable areas of near-surface permafrost (13.19 to 15.77 million km²). However, there is not a simple relationship between the sophistication of the snow insulation in the acceptable models and the simulated area of Northern Hemisphere near-surface permafrost, because several other factors, such as soil depth used in the models, the treatment of soil organic matter content, hydrology and vegetation cover, also affect the simulated permafrost distribution.