Abstract

Convective Available Potential Energy (CAPE), the integral of buoyancy of air parcels arising from the surface, is closely linked to extreme weather and an important parameter that models should capture. Recent studies have been interested in showing projections of CAPE into future as a climatological index. However, to date, limited use has been made of the decades of observational record of CAPE from radiosondes to evaluate model performance or to attempt to confirm trends. We compare CAPE values over North America from the observational radiosonde network (IGRA), 12.5 km reanalysis (ERA-I; ERA5), and 4 km convection-resolving regional WRF simulations. We show that surface-based CAPE in both reanalysis and convection-resolving ERA-I reanalysis is systematically biased low, primarily driven by low humidity at the high tail, corresponding to the most impacts-relevant conditions. The “missing tail” implies that models may underestimate the intensity of convective processes and extreme precipitation. Reproducing observed distributions of CAPE is an important diagnostic for ensuring confidence in future projections.