Lagrangian Characteristics of Machine Learning-Based Drop Size Distributions in Convective Cells: A Case Study

Understanding the variability of raindrop size distributions (DSDs) and their evolution is crucial for unraveling the microphysical processes within convective storms. This study utilized X-band polarimetric radar and a machine learning-based DSDs retrieval algorithm to investigate the spatiotemporal evolution of DSDs within convective cells. A well-developed typical convective cell was intensively observed using a unique scanning strategy, where range-height indicator (RHI) scan directions were adaptively adjusted in real-time and provided detailed observations of the cell lifecycle from initiation to dissipation. The retrieved DSD parameters demonstrated that the developing stage was characterized by active condensational growth, the mature stage by a dominant collision-coalescence process, and the dissipating stage by a similar DSDs pattern with the mature stage, but with a smaller mean diameter, weakened collision-coalescence process, and the potential evaporation. These findings highlight the potential of dual-polarimetric radar and machine learning-based retrieval techniques in providing deeper insights into the spatial and temporal structures of DSDs.