2.2 SP data
The SP data was derived by using the snow cover phenology data of the northern middle and high latitudes (2001–2014) (Chen et al.,2016). That original data is an improved merge of five widely used snow datasets, including the reanalyzed dataset of daily snow depths published by the Canada Meteorological Center, the binary daily snow cover mask dataset derived from both the interactive multi-sensor snow and ice mapping system and the Northern Hemisphere weekly snow cover and sea ice extent, the 8-day Level 3 snow cover fraction product derived from the moderate resolution imaging spectroradiometer satellite MODIS data, and the snow water equivalent data derived from the near-real-time ice and snow extent dataset (Chen et al., 2016). Chen et al. (2016) employed a multi-data method in developing the combined snow cover phenology matrix (with the geo-grids unified into 0.5°×0.5°) that can integrate snow cover phenology information derived from multiple sources of snow observations. The snow cover accumulation season was defined to be from the November of the last year to the February of any given year, and the snow melting season is from the March to June of any given year (Chen et al., 2016). The snow-onset SP date was defined as the beginning date of the first five consecutive days on which snow is observed to cover the ground surface in its accumulation season, and the snow-end SP date was defined as the beginning date of the last five consecutive days when snow cover can still be noticed in its melting season (Chen et al., 2016). Specifically for the derivations of the snow-onset and -end SP datasets required in this study, the NH-related terrestrial area was simply segmented out by restricting the latitude to be larger than 45° in the Northern Hemisphere.