Relative Importance of Thermodynamic Variables to the Worldwide Variability of Thunderstorms

A 16-year Tropical Rainfall Measuring Mission (TRMM) Convective Feature (CF) dataset and ERA-Interim reanalysis data are used to examine the relationships between thermodynamic environments and thunderstorm convective intensity. Two statistical models are built to reconstruct the global distribution of thunderstorms based on their environmental variables. Using the probability functions of intense thunderstorms, the first model suggests that four variables, including Convective Available Potential Energy (CAPE), Convective Inhibition (CIN), low-level shear, and warm cloud depth, may be used to derive a geographical distribution of intense thunderstorms that is close to the observations. The second approach utilizes a random forest model to test the relative importance of these four variables for a convective cloud having lightning globally, as well as regionally. The strong land-ocean contrast of frequency of thunderstorms and some hotspot regions can be closely reproduced with the random forest model only based on the four variables from the reanalysis data. This suggests that the land-ocean contrast in thunderstorms can be largely interpreted by the fundamental differences between the thermodynamic conditions over land and ocean. The investigation of the relative importance of the four environmental variables over different regions suggests that CAPE plays an important role in the lightning over central America, central Africa, and the Maritime Continents. CIN is relatively more important over Amazon and a few coastal areas. Low-level wind shear is more important over Argentina, south-central United States, and the eastern Sahel. Warm cloud depth is shown with greater importance than the other three variables near mountains.