A low-dimensional model of the thunderstorm lifecycle and its verification in remote-sensing observations

Low-dimensional representations of the governing thunderstorm dynamics are essential for parametrisations and short-term predictions of severe weather. These are important prerequisites for safe and efficient operations of various sociotechnical systems (e.g. aviation, coupled energy systems). The prototype of the convection-scale thunderstorm evolution and elementary building block of organised thunderstorm systems is the cell lifecycle. In this presentation, we propose an analytical, low-dimensional model of the thunderstorm lifecycle. The very concept of a lifecycle implicitly assumes that thunderstorms are definite and recurrent coherent structures. This suggests a modelling approach in terms of macroscopic variables that characterise thunderstorms integrally. Inspired by Doswell et al.’s ingredients-based method and past experience with expert systems, we adopt a rule-based approach. This way, we formalise the general features of the lifecycle in a set of reaction rules, which are equivalent to a system of coupled, nonlinear differential equations. By design, this model is qualitatively consistent with the known features of the thunderstorm lifecycle. In order to verify that our model quantitatively captures the essential features of actual thunderstorm manifestations in the atmosphere, we compare our model against remote-sensing observations. In particular, we use satellite imagery from the SEVIRI instrument onboard the Meteosat Second Generation satellite and radar data from the WX composite over Germany, operated by the German Meteorological Service (DWD). Our model is intended to capture the most fundamental, repeatable features of the thunderstorm lifecycle. We therefore present an elementary post processing of the remote-sensing data to single out the relevant observational signatures associated with single-cell thunderstorm occurrence. Eventually, we demonstrate that our model correctly captures this temporal signature of post-processed remote-sensing observations quantitatively. Ultimately, our thunderstorm-lifecycle model may improve parametrisations as well as traditional nowcasting or physics-informed machine-learning approaches.