Sampling-Induced Conditional Biases in Satellite Climate-Scale Rainfall Estimates

Abstract

The effect of temporal sampling error in satellite estimates of climate-scale rainfall is to produce a ?conditional? bias where the algorithm overestimates high rainfall and underestimates low rainfall. Thus, the bias is conditional on the value of the estimate. This paper illustrates the problem using satellite infrared rainfall estimates together with a well-known satellite algorithm and shows it to be a function of the averaging scale, the sampling rate, and the temporal autocorrelation structure of the satellite estimates. Using realistic sampling rates, it is shown that significant biases exist in satellite rainfall estimates if polar-orbiting data are used in their construction. A simple correction for this bias based upon the estimated autocorrelation structure is given.