Laboratory for Atmospheres. NASA/Goddard Space Flight Center, Greenbelt, Maryland
Science Systems & Applications, Inc., Lanham, Maryland
A multichannel physical approach for retrieving rainfall and its vertical structure from SSM/I observations is examined. While a companion paper was devoted exclusively to the description of the algorithm, its strengths, and its limitations, the main focus of this paper is to report on the results, applicability, and expected accuracies from this algorithm. Some examples are given that compare retrieved results with ground-based radar data from different geographical regions to illustrate the performance and utility of the algorithm under distinct rainfall conditions. Move quantitative validation is accomplished using two months of radar data from Darwin, Australia, and the radar network over Japan. Instantaneous comparisons at Darwin indicate that root-mean-square errors for 1.25° areas over water are 0.09 mm h1 compared to the mean rainfall value of 0.224 mm h1 while the correlation exceeds 0.9. Similar results are obtained over the Japanese validation site with rms errors of 0.6 1 5 mm h1 compared to the mean of 0.880 mm h1 and a correlation of 0.9. Results are less encouraging over land with root-mean-square errors somewhat larger than the mean rain rates and correlations of only 0.71 and 0.62 for Darwin and Japan, respectively. These validation studies are further used in combination with the theoretical treatment of expected accuracies developed in the companion paper to define error estimates on a broader scale than individual radar sites from which the errors may be analyzed. Comparisons with simpler techniques that are based on either emission or scattering measurements are used to illustrate the fact that the current algorithm, while better correlated with the emission methods over water, cannot be reduced to either of these simpler methods.