Multi-sensor analysis of hailstorms in the Mediterranean Region

Hailstorms are among the most impactful manifestations of thunderstorm development and evolution and represent a major meteorological risk in the Mediterranean region, where they cause significant socio-economic losses. Despite their relevance, hail events are still not accurately and comprehensively captured due to their local-scale extent, the lack of appropriate observing systems and the limited availability of high-resolution modelling studies. In the context of climate warming and the expected increase in frequency and intensity of extreme weather events across Europe, improving the detection and understanding of hailstorms has become crucial. This PhD research presents a multi-sensor analysis of hailstorms in the Mediterranean Basin, integrating satellite, lightning and radar data to investigate their formation, development and detection. The first part of the work focuses on the “Multi-sensor Approach for Satellite Hail Advection” (MASHA) Method, a satellite-based technique that exploits the physical interaction between ice hydrometeors (ice crystals, graupel and hailstones) and microwave and infrared electromagnetic waves to estimate hail probability (HP) every five minutes over a wide Mediterranean domain. Strengths and limitations of the MASHA Method were assessed, highlighting issues related to the calibration of the MWCC-H algorithm and to the misinterpretation of overshooting top (OT) structures in severe hailstorms. To overcome these limitations, two complementary approaches were explored. First, a neural network-based method relying on a ResNet-50 Convolutional Neural Network (CNN) was trained using infrared satellite images to recognize hailstorms. Preliminary results demonstrated promising performance in distinguishing hail-producing storms from ordinary thunderstorms, opening new perspectives for satellite-based monitoring. Second, lightning activity was investigated as a proxy for hail detection. A new hail probability (NHP), combining MASHA-derived HP and lightning-based indicators, significantly improved hail report classification and reduced underestimation, particularly for deep OT hailstorms producing large and very large hailstones. Further exploratory analyses examined the relationship between lightning variables and hail occurrence through the development of an innovative mathematical approach for lightning jump (LJ) detection. Results showed that LJ indicators, lead time (LT) and jump intensity are strongly associated with hail occurrence and severity, with a positive correlation between lightning jump intensity and maximum hail size. The comparison between the LJ method and a radar-based hail detection algorithm demonstrated comparable performance, while their combination maximized detection capability. Overall, this research highlights the potential of a multi-sensor framework for improving hailstorm identification and monitoring in the Mediterranean region, emphasizing the role of satellite and lightning data for real-time applications and nowcasting systems, and providing new insights into the relationship between electrification processes and hail severity.