Global Perspectives on Convective Storm Frequency from Multi-Reanalysis Climatology

The global distribution of severe convection has become a topic of increasing interest in recent decades with the arrival of the 4th and 5th generation reanalysis products at horizontal and temporal resolutions that have a good capacity to represent convective environments. Intercomparison and evaluation of reanalysis relative to observed profiles globally reveals inconsistencies in how these best-guess analyses represent convective environments. This suggests to capture global severe convective environments a multi-reanalysis ‘ensemble’ is necessary, as no single product can be considered to be equally performative across different regions. Traditionally, modelers have focused on regionally invariant relationships, or calibrated approaches to balance performance across different domains, however, this presents a significant challenge when the well observed regions are typically confined to mid-to-high latitudes. Like the single reanalysis approach, single models or parameter combinations have proven of limited utility in depicting the wide range of environments that can produce these hazards across different regions.This poster will illustrate the evidence for variable reanalysis performance, a multi-reanalysis climatology of convective environments, and introduce a hierarchical random forest modeling to represent the regional and seasonal variations in potential environments. This novel approach will be compared to existing models that use linear or fixed threshold approaches to approximate environment frequency, and highlight regions where the hierarchical approach shows considerable differences. Finally, spatial maps and distributional spaces will be used to illustrate how these environments relate to traditional phase spaces, and regions of common convective environment characteristics.