A machine-learning-based perspective on deep convective clouds and their organisation in 3D : Part 1: Influence of deep convective cores on the cloud life cycle

Abstract

In this two-part study, we examine spatio-temporal patterns of convective clouds, their properties, and organisation. We use a machine-learning-based method to extrapolate a contiguous 3D cloud field of 2D satellite data. The predicted data are used to simultaneously track both the horizontal and vertical development of clouds. Our research focuses on West Africa, a region known for frequent convective events and severe weather. In Part 1, this study compares cloud and core properties and the cloud life cycle over land and ocean during a 6-month period from March to August 2019. Our analysis reveals that 65 % of tracked cloud systems contain only a single core and persist for less than 3 h. Despite their shorter lifespan compared to multi-core clusters, single-core clouds exhibit stronger changes in the radar reflectivity and a higher vertical growth. In contrast, multi-core clouds show greater horizontal growth, encompassing larger cloud and core areas, higher cloud-top heights (CTHs), and higher average reflectivity at 10 km altitude. We also find that, in systems with more cores, both the maximum number of cores and the peak core area occur later during the cloud life cycle. Notably, the differences in cloud characteristics between land and ocean are smaller than those associated with the number of convective cores. However, the results may not fully capture climatological differences. Further research using longer time series is needed to quantify the observed variability of tropical convection.