Dynamics and Model Representation of Two Contrasting Extreme Precipitation Events in the Sahel

TL;DR

This study investigates two extreme flood events in Mali, analyzing rainfall structures, dynamical forcings, and the ICON model's ability to simulate their evolution, revealing case-dependent model performance and convective organization issues.

Contribution

It provides a detailed comparison of convective parameterization effects in the ICON model for extreme Sahelian rainfall events, highlighting model strengths and deficiencies.

Findings

Both events were driven by organized convective systems with cyclonic vortices.

EXPLC better captures rainfall in San but fails in Kenieba; PARAM performs better in Kenieba.

FSS shows model skill varies by case; SAL indicates EXPLC produces too many scattered rainfall systems.

Abstract

Two extreme flood-inducing precipitation events in two cities in Mali, on 08 August 2012 in San (127 mm) and on 25 August 2019 in Kenieba (126 mm), are investigated with respect to rainfall structures, dynamical forcings, and the ability of the ICOsahedral Nonhydrostatic (ICON) model to represent their evolution. Two sets of experiments with convective parameterization enabled (PARAM) and disabled (EXPLC), both at 6.5 km grid spacing, are conducted for each case. While the (thermo)dynamical fields of the simulations are compared with ERA5 reanalysis data, the rainfall fields are tested against the satellite-based precipitation dataset IMERG by applying the spatial verification methods Fractions Skill Score (FSS) and the Structure-Amplitude-Location (SAL) score. In addition, a spectral filtering of tropical waves is applied to investigate their impact on the extreme events. The most prominent results are: (1) Both cases were caused by organized convective systems associated with a westward propagating cyclonic vortex, but differ in their environmental setting. Although both cases featured an east African wave (AEW), the San case involved convective enhancement along dry Saharan airmasses, whereas the Kenieba case occurred within an unusual widespread wet environment extending deep into the Sahel. (2) Although EXPLC captures the rainfall distribution in the San case better than PARAM, it fails to organize convection in the moisture-laden Kenieba case, which PARAM is capable of simulating. (3) The FSS confirms the case-dependency of the ICON skill. The SAL method hints towards a systematic deficiency of EXPLC to represent the convective organization by producing too many scattered and weak rainfall systems, while PARAM is more effective in converting abundant moisture into excessive rainfall.