The Italian territory is characterized by relevant mountain chains (Alps and Apennines) in the proximity of coastal areas that lift the airflow, favouring condensation which is a key factor in the production of heavy precipitation. The complex orography in combination with steep, urbanized, small catchments, makes this area one of the most exposed Mediterranean areas to hydrogeological risks. About 90% of Italian municipalities are susceptible to floods, which have caused 466 deaths between 1990 and 2006 alone and over EUR 19 billion in damages.
Liguria region, in particular, is a coastal area prone to severe floods, especially during autumn. Flood episodes are associated with a number of topographical (relatively steep though not high slopes, a concave arc-shaped coastline, and short torrents/rivers having very small catchments, of the order of 102 km2 or even smaller) and meteorological–climatological factors, namely a coastal slope exposed to moist southerly flows having the character of “warm conveyor belts” or “atmospheric rivers”, favourable condition for the development of convective systems. In this framework, an Observational Campaign, will take place in the autumn in Liguria region. In order to enhance our understanding of back-building Mesoscale Convective Systems that frequently affect the study area producing heavy damages, the existing observations will be supplement with a X-band weather radar, a Meteodrone ad a network of low-cost temperature sensors. To increase the forecast reliability of extreme weather events, an innovative early warning system will be used. The system combines a cloud resolving NWP model at high spatial and temporal resolution, using a rapid update cycling 3DVAR assimilation technique, with a radar-based nowcasting system through a blending technique.
To reduce the uncertainty related to the initial state of the atmosphere at small spatio-temporal scales, a twofold aspect has been considered:

• Assimilation of heterogenous datasets in a rapid update cycle (every 1-3 hours). In details, GNSS data to gain a deeper understanding of integrated water vapour content status at high temporal frequency; weather radar reflectivity to improve the reconstruction of the 3D cloud field; lightning to better capture the development of convective structure. The early warning system will also integrate the observations acquired during the Observational Campaign to exploit the untapped potential of assimilating X-band radar reflectivity as well as temperatures and wind measurements from meteodrones and low-cost sensors.
• An innovative post-processing algorithm named Score-Weighted Improved NowcastinG (SWING) to reduce timely and spatial uncertainty in the convective field simulation.
• Application of post-processing techniques (e.g., PyDDA algorithms developed at NSSL) to retrieve wind kinematics in precipitation storm systems from Doppler weather Radar/Lidar acquisitions using 3D data assimilation techniques. The forecast products will be produced at least at hourly temporal resolution and further processed through AI techniques aiming at further improving forecast skills for highly localized phenomena (heavy rainfall, hailstorms, severe wind etc) via hybridization or post-processing with machine learning and deep learning techniques.