Nowcasting Techniques for Short-Term Weather Forecasts Using Radar Data

Short-term weather forecasts using radar measurements predict precipitation within minutes to a few hours, aiding various fields such as severe weather monitoring, airport operations, and planning of sporting events. Radar forecasts are especially crucial during severe weather conditions like storms, as they provide real-time data that enhances weather alert systems and emergency preparedness. In airports, accurate precipitation forecasting is vital for air traffic management, flight safety, and scheduling. Similarly, for outdoor sports events, reliable forecasts ensure the safety of participants and spectators by enabling informed decisions about schedule adjustments. This study implemented a precipitation nowcasting procedure in the Tuscany region to provide accurate forecasts over various areas. The method chosen aims to be compatible with cost-effective hardware without sacrificing performance. While recent advancements in machine learning, especially deep neural networks, have improved nowcasting, their computational demands are high. Therefore, a deterministic nowcasting method was selected for this study, as it balances quality and computational efficiency. The chosen method, described by Ayzel et al. [1], was applied to data from the radar network of the Italian Civil Protection in Tuscany, covering all precipitation phenomena from January 1, 2022, to December 31, 2023. The study aimed to determine the impact of area size on forecast quality, assessed through Critical Success Index (CSI) and Mean Absolute Error (MAE). The analysis revealed that nowcasting quality depends on the persistence of rainfall intensity and is generally better for stratiform than convective phenomena due to the latter's variability. The paper is structured as follows: Section 2 describes the nowcasting technique and chosen procedure, introduces evaluation indices, and details the radar network. Section 3 presents the dataset. Section 4 presents the results and analysis of the nowcasting procedure. Section 5 concludes the study.