Data from the first Algorithm Intercomparison Project(AIP/1) collected over Japan and surrounding waters in June, July, and August 1989 are used in this study to assess the importance of visible data in satellite rain estimation techniques. The purpose of the project was to compare different methods for estimating rainfall using satellite measurements. Radar and surface gauge data provided the validation set.
RAINSAT, an estimation technique using both visible (VIS) and infrared (IR) data, achieved the highest correlation with the validation data. In this paper rainfall estimates from RAINSAT (VIS+IR) am compared with two IR-only techniques to deduce the effectiveness of VIS data. Some estimates are also made using a VIS-only technique. Comparisons am made on both a spatial and diurnal basis.
Cloud climatologies for a subset of the AIP/1 data and the southern Ontario data on which RAINSAT was trained showed a marked similarity. It is found that the total volume of rain as a function of albedo is very similar for both Japanese and Ontario data.
The VIS data generally produced higher correlations with the validation data than did the IR data. This was especially the case when rain fell from warm, orogaphically induced rainfall. When rain fell from cold bright clouds. especially over the ocean, the correlations of the two types of data with the validation data were similar.
It is also shown that normalization of VIS data by the cosine of solar zenith data was inadequate to remove diurnal variations in apparent brightness.