Validation and uncertainty analysis of satellite rainfall algorithms

Satellites, while offering excellent spatial coverage, determine precipitation indirectly, using algorithms that transform satellite-sensed radiance (either emitted or scattered) from clouds or raindrops into precipitation. A large uncertainty is associated with satellite precipitation estimates, stemming from unknown variation in space and rime of the physical and statistical relationships between precipitation and satellite-sensed radiance. To mitigate this, satellite algorithms must be calibrated and verified using surface precipitation sampled from different climate regimes and seasons. Recently developed statistical techniques have been used effectively to reduce spatial sampling error associated with sparsely distributed raingages and thereby improve our understanding of satellite algorithm quality. This paper provides an example of satellite precipitation validation, including a description of the types of satellite data used to estimate precipitation, as well as the results from a major project (the Global Precipitation Climatology Project [GPCP]), to estimate global precipitation through a combination of satellite and raingage products. In addition, a recently developed procedure to investigate spatial averaging, scaling, and uncertainty analysis will be used to examine the GPCP product. Specifically, uncertainty analysis applied to comparisons between satellite monthly rainfall estimates and rainfall estimates constructed from Pacific atoll-sited raingauge sites will be discussed.