Revisiting Environment-Based Hail Size Classification: A Direct Comparison of Composite Indices and Environmental Parameters from SPC Mesoanalysis

With an emerging focus on conditional peak hazard intensity forecasting at the Storm Prediction Center (SPC), there is a need for accurate, reliable guidance for severe hail size classification. Numerous composite indices, including Large Hail Parameter (LHP); Significant Hail Parameter (SHIP); and the Sounding Analog Retrieval System (SARS), seek to distinguish severe (defined now as >=1 inch diameter) hail from significant-severe (>=2 inch diameter) hail using a combination of environmental parameters. These indices are incorporated into SPC Mesoanalysis, which provides hourly best-guess assessments of the environment from the gridded 40-km RUC/RAP-based surface objective analysis (sfcOA) scheme and are utilized by forecasters for mesoscale discussion and watch issuance. However, no work exists that directly compares and verifies these indices. Therefore, this study comparatively evaluates composite indices and environmental parameters from SPC Mesoanalysis, along with additional parameters and indices not found in sfcOA, for conditional hail size classification using analysis data gathered for every initialization hour and nearest grid point containing Storm Data hail reports from January 2011 to December 2023.

While all sfcOA indices demonstrate some statistically significant skill in distinguishing hail intensity, SARS has the greatest separation between each hail size bin. Still, these differences are minimal, and the predictability of conditional hail size by existing indices remains limited. Moreover, SARS demonstrates variability in the accuracy of predictions across different regions and observed hail size bins. Environmental parameters not used in existing indices, such as effective bulk shear and entrainment-adjusted CAPE, demonstrate greater individual skill in size classification than the components of existing indices. Tree-based machine learning models incorporating these parameters were also developed during this project and evaluated according to the aforementioned methods. This approach demonstrates a substantial statistical improvement in performance over existing indices and provides a promising avenue for future advancements in environment-based hail size forecasting.