Testing the skill of an analytical expression for the probability of occurrence of tornadoes

Tornadoes represent major meteorological hazards, in terms of damages to buildings, vehicles and structures and casualties. Because of their small space scale (order of 1km or less), duration (order of 1000s), strongly nonlinear and chaotic dynamics, tornadoes cannot be reproduced in operational weather prediction and climate models. It is important to develop approaches overcoming this limitation and capable of delivering reliable early warnings by civil protection services and estimating whether frequency and strength of tornadoes will change because of anthropogenic climate change. Recently, a probabilistic approach has been developed that resulted in analytical expressions of the probability of tornadoes occurrence based on meteorological parameters that can be extracted from weather prediction and climate models, such as WMAX (updraft maximum parcel vertical velocity, linked to the Convective Available Potential Energy CAPE), WS700 (the wind shear in the lower troposphere), LCL (the lifting condensation level), SRH900 (low-level storm relative helicity). An example is the formula log10(P)=-6.6+WMAX/(3.1+5.2 · WMAX/WS700), which is meant to describe dependence of probability P of occurrence of a tornadoes  on the surrounding environmental conditions and to distinguish among conditions with low and high probability. In this study this and similar formulas are applied to hindcasting the probability of tornadoes using ERA5 data. The purpose is to assess the skill of the method for operational prediction and explore its validity for climate change studies. The methodology supporting this formula is extensively described in Ingrosso, R., Lionello, P., Miglietta, M. M., and Salvadori, G.: Brief communication: Towards a universal formula for the probability of tornadoes, Nat. Hazards Earth Syst. Sci., 23, 2443–2448, https://doi.org/10.5194/nhess-23-2443-2023, 2023.