Satellite Observations of Precipitation - the Link between Products and Processes

Operational satellite precipitation products, particularly those from Infra-red and passive microwave sensors, are severely under-constrained – meaning that the observations are not unique but can correspond to a number of different surface rainfall conditions. Modern inversions can account for this non-uniqueness and ascribe an uncertainty to the final product that stems from this non-uniqueness. The frequency of occurrence of possible solutions for a given radiance signal, however, is rarely random, but instead varies with the meteorological conditions and the cloud processes that create the precipitation. This leads to consistent errors, that are neither random nor systematic, but are dependent on meteorological regimes. These errors are impossible to predict from the radiance signal alone and require a deeper understanding of the cloud processes to develop accurate QPE products. Some progress has been made to link observed precipitation structures to regional biases, and precipitation processes to the observed cloud structures. Together, this provides a first step make robust global predictions estimates along with their random and regime dependent structures.