Assignment of rainfall confidence values using multispectral satellite data at mid-latitudes: first results

The authors propose a new method for the assignment of rainfall confidences on a pixel basis using cloud properties derived from optical satellite data during daytime. This approach is based on the concept model that the probability for precipitation is a function of the liquid water path, which in turn can be computed using the satellite-retrieved cloud optical thickness and the cloud effective droplet radius. In order to evaluate the principal potential of this idea, scenes from the Terra-MODIS sensor during the severe European summer floods in 2002 have been analysed in order to derive a corresponding regression function that interlinks the liquid water path with the rainfall probability or better with the confidence that a pixel which is classified as raining does actually rain. A first evaluation against ground-based radar data during March 2004 shows good skill of this new method.