Prediction of rainfall from WSR-88D radar using support vector regression

The objective of this paper is to utilize Support Vector Regression (SVR), and compare that to other methods, to facilitate rainfall estimation. Ground truth rainfall data are necessary to apply intelligent systems techniques. With the advent of a national network of advanced radars, massive archived data sets are available for data mining. The application of SVR is new for estimation of rainfall by radar. The radar data base contains measurements of reflectivity, wind velocity and spectrum width. Current rainfall detection algorithms make use of only the reflectivity variable, leaving the other two to be exploited. The focus of the research is to capitalize on these additional radar variables. Rainfall totals from the Oklahoma Mesonet are utilized for the ground truth in training and verification data. Results indicate that modeling with reflectivity alone is most accurate for SVR with MSE equivalent to optimized ANNs for reflectivity and spectrum width inputs.