Climatology of Mesoscale Convective System-Related Severe Hazards – Classification Algorithm Development and Representation in Convection-Permitting Climate Models

This study focuses on the severe weather hazards produced by mesoscale convective systems (MCSs) in the central and eastern United States. The goals of this study are to develop an algorithm for identifying severe hazards produced by MCSs, reconstruct an MCS-related hazard climatology, evaluate the performance of convection-permitting climate model (CPCM) in representing these events and investigate the potential future changes in these hazards due to climate change. To achieve these goals, an MCS tracking algorithm based on satellite brightness temperature and radar reflectivity is applied to observations and for current and future scenarios from CPCMs. To develop the hazard classification algorithm, MCSs that produced severe hazards, including flash flood, hail, tornado, and wind, are identified based on storm reports. Object-based classification algorithms are developed using the random forest method. For model training and testing, statistical values–such as means, standard deviations, minima, and maxima–are derived from storm- and environment-related parameters from both non-event and event-producing MCS segments during the warm season between 2004 and 2016. These variables are direct output from the tracking algorithm and/or are computed from atmospheric reanalysis data. Three individual models are built for flash flood, severe, and significant severe events for hail, wind, and tornadoes. For flash flood and severe classification, the model can produce promising results with more than 90% accuracy, while it struggles for significant severe event classification. Given that storm reports heavily rely on human input, the development of these algorithms is beneficial in creating a more comprehensive and objective-based climatology of MCS hazards. Subsequently, classification algorithms are applied to model-simulated MCSs for both historical and future events, addressing the model’s ability in reproducing MCS hazard climatology and to anticipate the potential future changes in MCS properties and hazards.