Intercomparison of products for climate applications

Abstract

Observational datasets of global precipitation are widely used for a range of climate applications. The precipitation products, however, are not strictly a “true” representation of nature, but have their own uncertainties related to issues such as sampling errors and algorithmic assumptions. We present here an intercomparison of 11 global precipitation datasets. Major conclusions are: • While the overall geographical pattern of precipitation is coherent among products, the magnitude varies from one dataset to the other. The agreement is poor particularly at high latitudes, since light and/or solid precipitation typical of high latitudes is difficult to estimate accurately from satellite microwave radiometry. • A systematic bias is present between gridded gauge products, which is presumably partially responsible for the spread in merged multi-satellite datasets adjusted to the gauge products. • The bias characteristics in the annual/monthly mean precipitation are a poor predictor of those in extreme precipitation.