92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Tuesday, 24 January 2012: 9:30 AM
Spatiotemporal Data Mining of High Resolution Simulations of Tornadoes
Room 242 (New Orleans Convention Center )
Amy McGovern, Univ. of Oklahoma, Norman, OK; and R. Kimes, B. Pirtle, and R. A. Brown

Tornado prediction is a very challenging task for a variety of reasons, including the lack of data available from current sensing platforms to the lack of knowledge of the exact processes that lead to tornadogenesis. In this project, we are studying the formation of tornadoes using spatiotemporal data mining. We have created a unique dataset of over 50 high resolution simulations of supercell thunderstorms. These simulations were created using the Advanced Regional Prediction System (ARPS) and have a horizontal resolution of 75 m. By creating a large set of realistic simulations, we have a unique data set that will enable us to better understand why seemingly similar severe storms become tornadic or stay non-tornadic. To accomplish this, we have developed a set of novel spatiotemporal relational data mining methods, the Spatiotemporal Relational Probability Tree and the Spatiotemporal Relational Random Forest. We present preliminary results of our data mining effort.

Supplementary URL: