Gap-filled daily high-resolution snow-cover maps around Alpine eddy-covariance stations derived with SnowFLAKES

Dataset description This dataset provides gap-filled daily, high-resolution binary snow cover maps for 12 eddy-covariance stations within the FLUXNET, ICOS, and i-Box networks distributed across the French, Swiss, Austrian, and Italian Alps. The maps were generated to characterize surface snow conditions within the source areas of eddy-covariance measurements for the updated analysis of Scheidt et al. (2026). The dataset was specifically used to determine whether the area enclosed by a station’s 80% flux-footprint contour was snow covered and to calculate the snow-cover fraction (SCF) within that footprint. Footprint-scale SCF can be calculated from the maps as the proportion of valid pixels within the footprint classified as snow covered. The data cover the following eddy-covariance sites for the hydrological years (1st October - 30th September) from 2013/14 to 2023/24: Alp Weissenstein (CH-Aws), Switzerland Col du Lautaret (FR-CLt), France Früebüel (CH-Fru), Switzerland Monte Bondone (IT-MBo), Italy Mazia Pasture (IT-MtP), Italy Neustift (AT-Neu), Austria Nivolet (IT-NiG), Italy Torgnon (IT-Tor), Italy The i-Box observational network in the Inn Valley, Austria, comprising the Eggen (CS-SF1), Kolsass (CS-VF0), Hochhäuser (CS-NF27), and Weerberg (CS-NF10) stations For each station outside the i-Box network, the NetCDF files cover a 15 km × 15 km square centered on the station. Because the four i-Box stations are located close to one another, they are represented by a common 20 km × 20 km domain centered on the station cluster. The maps have a spatial resolution of 50 m and are provided in the WGS 84 / UTM zone 32N coordinate reference system (EPSG:32632). Files are organized by site and hydrological year. The binary snow cover condition variable represents daily snow-cover conditions, while the reconstruction_method variable distinguishes pixels obtained from high-resolution satellite observations from pixels reconstructed by the gap-filling procedure. High-resolution snow cover fraction maps were derived from Sentinel-2 MSI and Landsat 8 and Landsat 9 OLI imagery using SnowFLAKES (Snow extent from applying a Flexible Learning Algorithm using KErnel-based Spectral unmixing) at a resolution of 50m. SnowFLAKES is an unsupervised, image-adaptive snow-mapping approach developed for mountainous terrain and challenging observation conditions, including topographic shadows and variable illumination. The processing follows the approach described by Barella et al. (2022). Daily, spatially complete maps were generated by combining the available high-resolution snow observations with daily, lower-resolution MODIS SCF from the MOD10A1 product (Hall and Riggs, 2021). Gap filling and spatial downscaling were performed using an implementation adapted from the hierarchical multiresolution data-fusion method proposed by Premier et al. (2021). The approach uses spatial patterns learned from available optical observations to reconstruct snow-cover conditions on days without usable high-resolution satellite imagery. The resulting product represents daily binary snow presence at high spatial resolution. Raw Sentinel-2, Landsat, and MODIS imagery, eddy-covariance measurements, and flux footprint polygons are not included in this archive. The archive also includes an interactive Python viewer for the daily *_SC_GAM.nc snow-cover products. It can display snow-cover classes and reconstruction provenance over time, with station locations shown for context. Setup instructions and examples are provided in the README. NetCDF structure and value conventions For each site and hydrological year a NetCDF product is provided. Files ending in _SC_GAM.nc contain the 50 m high-resolution product. The main variable, binary_snow_cover_condition (time, y, x), has units of percent and represents binary snow presence: 0 denotes snow-free ground and 100 denotes snow-covered ground. The product uses the dimensions time, y, and x. Projected x and y coordinates are expressed in meters, time identifies the daily observations, reference_time supplies the time reference, and spatial_ref contains the coordinate-reference-system definition. In the binary snow cover files, 255 represents pixels outside the valid analysis area or pixels without data. The value 205 denotes a cloud-contaminated or otherwise unresolved observation. Only finite values in the inclusive range 0–100 should therefore be treated as valid snow-cover data; 205, and 255 must be excluded from snow-cover calculations. The unsigned 8-bit variable reconstruction_method(time, y, x) in the _SC_GAM.nc files records pixel provenance. Its flag values are 0 for not applicable, 1 for a high-resolution satellite observation, and 4 for a pixel reconstructed using the generalized additive model (GAM) gap-filling procedure. This variable must not be interpreted as snow-cover fraction. EC-station information NetCDF code Eddy-covariance site Station abbreviation Country Latitude Longitude Elevation Aws Alp Weissenstein CH-Aws Switzerland 46.583056 9.790639 1,978 m CLt Col du Lautaret FR-CLt France 45.041400 6.410530 2,050 m Fru Früebüel CH-Fru Switzerland 47.115833 8.537778 982 m MBo Monte Bondone IT-MBo Italy 46.014680 11.045831 1,550 m MtP Mazia Pasture IT-MtP Italy 46.684285 10.585237 1,550 m Neu Neustift AT-Neu Austria 47.116318 11.320229 970 m NiG Nivolet IT-NiG Italy 45.519191 7.171434 2,555 m Tor Torgnon IT-Tor Italy 45.844440 7.578055 2,168 m i-Box Eggen CS-SF1 Austria 47.316150 11.616500 829 m i-Box Kolsass CS-VF0 Austria 47.305341 11.622190 545 m i-Box Hochhäuser CS-NF27 Austria 47.287550 11.631220 1,009 m i-Box Weerberg CS-NF10 Austria 47.299750 11.672960 930 m