A collaborative team of meteorologists and data mining experts conducted a case study to detect and classify mesocyclone signatures in WSR-88D radar data using mining techniques. Radar data for May 6th, 1994 and May 11th, 1992 from Norman and Tulsa, Oklahoma was used in this case study. Mesocyclone Detection Algorithm (MDA) was developed for this case study and is similar to the algorithm used in National Severe Storm Laboratory (NSSL). The primary difference between the two algorithms is in the technique used to segment the two dimensional (2D) mesocyclone signatures. This detection algorithm uses a region growing technique; therefore the shape of the feature is no longer a restriction. Truth/false labels were assigned to the mesocyclone features generated from the MDA by comparing with a truth set derived by an expert using the NSSL algorithm. This labeled feature data set was then used in a series of analysis experiments. The objective of these experiments was to: Evaluate the performance of different classifiers in their ability to distinguish between a true mesocyclone signature versus an artifact or noise; Optimize the classification results by selecting key parameters from the feature data set using different feature reduction techniques; Explore the patterns in the feature data set using various clustering algorithms. This paper will describe the results from these experiments.