WRF data assimilation of weather stations and lightning data for a convective event in northern Italy

The present work shows the relevance of assimilating mesoscale observations and lightning data in the Weather Research and Forecasting (WRF) model, to simulate a strong convective event in northern Italy, poorly forecasted by available weather models even a few hours before the event itself. The data assimilation was conducted by testing the 3D-VAR and 4D-VAR assimilation algorithms implemented in the WRF data assimilation (WRFDA) suite, with different configurations and different assimilation windows. An extensive sensibility test has been operated to properly analyze the effect that the assimilation of a single station has on the model outcomes. Input data were taken from two networks of more than 1000 citizen-science meteorological stations, available in northern Italy, and from lightning flashes derived from Earth Networks Total Lightning Network, assimilated using the atmospheric water vapor as a proxy variable. Rain forecasts over an area in the north of Milan were compared to the station’s measurements in the same area; POD, FAR, and CSI categorical statistics have been calculated. Results showed a positive improvement in the forecasted rain amounts with the ingestion of mesoscale weather data into 3D-VAR and 4D-VAR algorithms, more pronounced using 4D-VAR with a more frequent input data integration. A few improvements were reported by the 3D-VAR, with the lightning data assimilation, probably caused by the absence of the model’s spin-up time with this configuration. An ideal simulation, which increased the water vapor of the air mass 2 h before the convective event, reported a positive enhancement of the rain amounts. The tests conducted on a single convective event are nevertheless encouraging, because they show a positive improvement of forecast with the assimilation of near-ground weather data and tropospheric water vapor 1 or 2 h before the beginning of the convection activity.