MYRORSS Storm Cluster Dataset for Tornado Detection and Climatology (2005–2011)

Data Description This dataset contains object-based storm clusters derived from the Multi-Year Reanalysis of Remotely Sensed Storms (MYRORSS; Williams., 2020) over the contiguous United States for 2006–2011. Scale-1 storm clusters are identified from radar composite reflectivity using a segmentation algorithm and are tracked at 5-minute intervals with a minimum area of approximately 200 km². Each cluster is data mined with radar-derived and near-storm environmental parameters from NEXRAD WSR-88D observations and Rapid Update Cycle (RUC) analyses. Clusters are collocated with the NOAA Storm Events Database to assign event labels (tornado, hail, and wind). In addition to the storm cluster dataset, this repository includes products developed: A curated feature set of radar and NSE predictors used for model training Preprocessed input datasets used to train and evaluate machine learning models Tuned and calibrated CatBoost and Random Forest models for tornado detection and severity classification Reference Williams, S. S., Ortega, K. L., Smith, T. M., & Reinhart, A. E. (2022). Comprehensive Radar Data for the Contiguous United States: Multi-Year Reanalysis of Remotely Sensed Storms. Bulletin of the American Meteorological Society, 103(3), E838-E854. https://doi.org/10.1175/BAMS-D-20-0316.1