Snow depth, a very significant factor in agriculture and climate research, is one of the
most important parameters for snow amount calculation. It is proved there is a good linear
relationship between snow depth and snow surface reflectance in visible to short-infrared window
channels when snow has a depth within 30cm, which makes it possible to retrieve snow depth
using AVHRR or MODIS data and station-measured snow-depth data.
This paper mainly introduces the principle theory and process to establish a snow-depth
retrieval model within 30cm using EOS/MODIS visible to short-infrared window channels' data
and station-measured data, considering snow characteristics in different physical states and
various complex underneath conditions including DEM, land cover such as grassland, forest,
cropland and so on. Based on snow characteristics and underneath conditions, snow is devided
into many types: old dry snow in flat grassland, new dry snow in flat grassland, old dry snow in
mountainous grassland, old dry snow in flat cropland and so on. Fourteen kinds of snow have been
modeled respectively in this retrieval model.
Through 4 years validation in XinJiang Province of China since 2002, the precision of
snow-depth retrieval model using MODIS visible to short-infrared channels' data can reach more
than 80%. In flat area with single underneath condition, where wind power can be ignored, the
model can always get a better precision. On the contrary, in mountainous forests, the precision of
the model is not that good.