Methods for Verifying Satellite Precipitation Estimates

Satellite precipitation estimates are widely used to measure global rainfall on monthly timescales for climate studies (e.g., Huffman et al. 1997). Near realtime satellite precipitation estimates are becoming increasingly available to the wider community. These precipitation estimates are potentially very useful for applications such as numerical weather prediction (NWP) data assimilation, now-casting and flash flood warning, tropical rainfall potential, and water resources monitoring, to name a few. As with any observational data, it is important to understand their accuracy and limitations. This is done by verifying the satellite estimates against independent data from rain gauges and radars. The terms validate and verify are used here to mean the same thing, “to determine or test the truth or accuracy of”. The former is preferred by the satellite community, while the latter is preferred by the NWP modeling community. Standard measures such as bias, correlation, and root mean square error (RMSE) error are very useful in quantifying the errors in climate-scale satellite precipitation estimates. Users of near real-time precipitation estimates often require more specific information on expected errors in rain location, type, mean and maximum intensities. Diagnostic validation approaches can reveal additional information about the nature of the errors. This paper discusses validation methods that give useful information to (a) help algorithm developers to improve their products, and (b) help users of satellite precipitation estimates to understand the accuracy and limitations of those products.