Our goal is obtaining quantitative characterizations of the accuracy of averages of satellite estimates of rainfall. It will first focus on error descriptions for averages from Tropical Rainfall Measuring Mission (TRMM) instruments, the passive microwave (TMI) and radar (PR), but then extend the approach to data from other satellites with passive microwave radiometers in order to evaluate GPM-type multi-satellite datasets. The kinds of averages studied will be those that are required for climatological studies of precipitation trends and in describing the diurnal cycle of rainfall, as well as for validation efforts designed to detect large space- and time-scale biases in the satellite estimates. The proposed approach complements other parallel efforts to describe the quasi-random retrieval errors made on the smaller scales of individual instrument fields of view. The research will build on the extensive experience of the P.I. and Co-I. in this field. A much simpler method of making error estimates is proposed that is capable of providing, routinely, maps of the rms random error not only of space/time averages but also for diurnal cycle statistics. Such error maps are essential in using satellite data for climate change research, precipitation trend analysis, for comparisons with climate models, and for validating the satellite estimates against surface data. The method requires extracting the space/time covariances of TRMM (or other satellite) data, and, importantly, the estimates are mostly based on local statistics rather than on global parameterizations. The method appears to work well in test cases, and suggests that the error characteristics of the TMI and PR data differ more than was first assumed. It is likely that they will differ for other satellites as well, and knowledge of these differences will help inform efforts to merge the various satellite data as envisioned in GPM.
