Data Sources and Methodology
This map was computed from the 18-year (1979 to 1997) daily snow-depth analysis of Brown et al. (2003). The analysis combined all available surface observations of snow depth with snow-depth estimates from a simplified physical snowpack model in areas of sparse data (mountains and high latitudes). This dataset is considered to be one of the best available representations of snow cover over North American middle latitudes. The quality of the dataset is less certain, however, in mountainous areas and over high latitudes, where there are few snow-depth observations.
The map was generated by computing the median date of snow-cover loss at each point in the 0.25 degree latitude-longitude grid over 18 snow seasons (season of 1979 to season of 1996). Loss date was defined as the last date with 14 consecutive days of snow cover greater than 2 centimetres in depth. A minimum of nine loss dates was required over the 18-year period in order to compute a median. The results were then adjusted with 3 degrees by 3 degrees grid averaging to yield contours for display. The grid resolution was insufficient to resolve detailed topography of areas such as Vancouver Island. In this area, contours were generated manually by taking into account available surface observations, satellite observations of the snow line, and local topography. Manual editing of contours was also carried out over the Canadian Arctic Archipelago to resolve some discrepancies related to the inability of the gridded dataset to resolve some of the smaller land masses. |
