Lightning data assimilation in the WRF-ARW model for short-term rainfall forecasts of three severe storm cases in Italy

Abstract

This study analyses the impact of total lightning data assimilation on cloud-resolving short-term (3 and 6 h) precipitation forecasts of three heavy rainfall events that occurred recently in Italy by providing an evaluation of forecast skill using statistical scores for 3-hourly thresholds against observational data from a dense rain gauge network. The experiments are performed with two initial and boundary conditions datasets as a sensitivity test. The three rainfall events have been chosen to better represent the convective regime spectrum: from a short-lived and localised thunderstorm to a long-lived and widespread event, along with a case that had elements of both.This analysis illustrates the ability of the lightning data assimilation (LDA) to notably improve the short-term rainfall forecasts with respect to control simulations without LDA. The assimilation of lightning enhances the representation of convection in the model and translates into a better spatiotemporal positioning of the storm systems. The results of the statistical scores reveal that simulations with LDA always improve the probability of detection, particularly for rainfall thresholds exceeding 40 mm/3 h. The false alarm ratio also improves but appears to be more sensitive to the model initial and boundary conditions. Overall, these results show a systematic advantage of the LDA with a 3-h forecast range over 6-h.