ABSTRACT
This chapter describes the current approach to performing large-scale data mining on the supercell simulations. It transforms understanding and prediction of tornadogenesis through the development and application of spatiotemporal data mining techniques to simulations of tornadic and nontornadic thunderstorms. The advent of mobile radars and sophisticated storm-scale data assimilation techniques have recently enabled the blending of observations and simulations into highly descriptive analyses of real-world tornadoes. Simulations of complex events such as severe weather require extensive high-performance computing experience and a significant involvement from domain scientists to ensure that the data are realistic. The chapter examines individual forests for physical insights into the causes of the tornadogenesis in the simulations. It develops the spatiotemporal relational probability tree (SRPT) and its related spatiotemporal relational random forest (SRRF) techniques. The mesocyclone and tornado objects are used to label each storm for the SRRF for training.
