Comparison of hail detection techniques using dual-pol radar observations in Belgium

In the frame of a Belgian project aiming at improving the hydrometeors classification near the ground through the use of radar observations, two dual-pol classification schemes and a single-pol one including hail are tested and compared with crowd-sourced observations. The first scheme is based on the work of Keenan (2003) and is a supervised method making use of fuzzy logic while the second one is a semi-supervised method designed by Besic et al. (2016) that includes clustering as well as fuzzy logic. The single-pol scheme follows the Waldvogel method and is the current operational method used at the Royal Meteorological Institute of Belgium (RMIB). The crowd-sourced observations are collected through the smartphone application of the RMIB. This application was released in 2019 and allows every user to report the real-time weather, including the occurrence of hail and its size. We will present the validation scores and analyse the behaviour of the three methods for several hail storms. The uncertainties will be discussed, not only related to the three classification schemes but also to the users’ reporting behaviour and the scarcity of the hail events. Finally, the short term perspectives of further improvement of the hydrometeors classification will be given.