Validation of oceanic rainfall retrievals is a challenge because of the lack of suitable ground truth and the impracticality of obtaining such measurements in sufficient number. Thus we must take a different approach to validation, to wit, making a model of the measurement uncertainties. In the last few years, reasonable estimates of the sizes of various contributors to the error have been developed making a total error estimate possible. The contributions all depend on various parameters so that a single estimate of the error is not possible. Rather, the error estimate must be integrated with the retrieval so that each retrieved value reported has an associated computed uncertainty.

Examination of the retrievals on a pixel-by-pixel basis permits some insights. At high rain rates the uncertainty in the Drop Size Distribution (DSD) dominates due to its large impact at 10.7 GHz. At low rain rates, instrument calibration dominates. At some intermediate rain rates the Beam Filling Correction is largest by a small margin.

Going to a Level 3 product (Space/Time average) alters the uncertainty picture somewhat. The random parts of the error can be directly estimated by partitioning the data into even and odd days of the month. Also, a determination of the zero rain rate offset is both possible and necessary. This determination absorbs many of the error sources such as calibration and surface wind. Generally sampling dominates the error in Level 3 products. In a percentage sense, the errors are largest where it rains least due primarily to this sampling error. In rainy areas such as the ITCZ, the sampling error is more modest but physics-based uncertainties are still significant.