21 years of Global SnowPack – Findings and application of the new NRT product

In terms of area, snow makes up the largest proportion of the cryosphere, but it is also the most short-lived with the greatest seasonality and variability. The use of remote sensing to detect snow has long been dependent on either passive microwave sensors or multispectral systems such as AVHRR, MODIS, or Landsat. While the former provide data on a daily basis and also allow insights into the snowpack (e.g. snow-water equivalent), their geometric resolution is insufficient for a closer look at the snowpack dynamics. Sensors such as Landsat offered a good geometric resolution, but the repetition rate was inadequate. The MODIS (Moderate-Resolution Imaging Spectroradiometer) sensor filled exactly this gap and has been providing data since 2000 on board the Terra satellite and since 2002 on board Aqua. For this period, the National Snow & Ice Data Center (NSIDC) offers daily snow cover as a level 3 product. The daily snow product MOD10A1 (Terra) or MYD10A1 (Aqua) has a nominal resolution of 500 m and is in sinusoidal projection. The detection of snow is based on the Normalized Difference Snow Index (NDSI), which makes use of the different reflection of snow in the visible spectral range (VIS) and the short-wave infrared (SWIR). Since snow reflects almost complete in the VIS, but almost none in the SWIR, the NDSI adapts a high value for snow cover. In addition, the normalized Difference Vegetation Index (NDVI) is used for snow under thick vegetation cover. The MODIS product now contains the values between 0-100 for NDSI (only positive values are assigned to land, multiplied by 100) and other values for different classes. The daily MODIS snow information forms the data basis for the Global SnowPack (GSP) processor. There, data gaps (e.g. through clouds or polar night) are filled in four steps. First, the Terra and Aqua data are combined and then filled with the day before and after. In the next step, a digital elevation model is used to determine the height from which there are only snow pixels and those from which there are only snow-free pixels; all pixel heights below or above are filled accordingly. The last step is the seasonal filter, in which all remaining data gaps are filled by gradually going backwards in the time series. Based on these "days until cloudfree", the elevation of the pixel and the day of the year, individual accuracy estimates are made and supplied as an accuracy layer. Since the MODIS snow data is available with a delay of two days and an additional day is necessary for the application of the 3-day interpolation, the near real-time product of the GSP is available after three days and can be downloaded from the GeoService of the Earth Observation Center. The reference period for the analysis of the snow cover is the hydrological year, which runs from the beginning of the meteorological autumn to the end of the meteorological summer. This period is further subdivided into an early and late snow season, the time of separation being mid-winter. For these seasons, the cumulative snow cover for each pixel is calculated and stored as early and late snow cover durations. The analysis of the variability of these snow cover durations for certain spatial areas (e.g. river catchment areas) enables trends and significant developments to be found. Recently, this data on snow cover has been coupled with hydrological models in order to better understand and predict extreme hydrological events. The Global SnowPack time series now comprises 21 hydrological years and the developments identified so far will be presented