During the winter the snow cover will change according to accumulation, wind transportation, and melting. In order to understand these processes and their influence on the vegetation and the hydrology, there is a need for frequent maps showing the snow cover distribution and how it changes during the season. This service has been requested from the Norwegian user NVE.

 

 

The service consists of two parts. Firstly an empirical snow distribution model is established for a given area of interest, e.g. a drainage area. The model represents the temporal and the spatial aspects of the snow retreat (melting) in a typical melting season. This model is first generated, and later also updated, from snow cover classification of Landsat-like images, acquired at various parts of the melting season. The model can be considered as a likelihood function defining the sequence of the pixels becoming snow-free.

 

The operational part of the service is based on this empirical model and medium-resolution satellite images, like MODIS. The algorithm for the snow distribution pattern goes like this: Retrieve the snow covered fraction for the local area by means of the MODIS image. Then produce a corresponding snow fraction from the empirical model by applying a threshold of the likelihood function. The mask that is produced by this threshold defines the estimated snow distribution pattern for the current day. Cloud detection is required in the pre-processing of MODIS data.