Investigating 3D and 4D variational rapid-update-cycling assimilation of weather radar reflectivity for a heavy rain event in central Italy

Abstract

Forecasting precipitation over the Mediterranean basin is still a challenge because of the complex orographic region that amplifies the need for local observation to correctly initialize the forecast. In this context, data assimilation techniques play a key role in improving the initial conditions and consequently the timing and position of the precipitation forecast. For the first time, the ability of a cycling 4D-Var to reproduce a heavy rain event in central Italy, as well as to provide a comparison with the largely used cycling 3D-Var, is evaluated in this study. The radar reflectivity measured by the Italian ground radar network is assimilated in the Weather Research and Forecasting (WRF) model to simulate an event that occurred on 3 May 2018 in central Italy. In order to evaluate the impact of data assimilation, several simulations are objectively compared by means of a fraction skill score (FSS), which is calculated for several threshold values, and a receiver operating characteristic (ROC) curve. The results suggest that both assimilation methods in the cycling mode improve the 1-, 3- and 6-hourly quantitative precipitation estimation. More specifically, the cycling 4D-Var with a warm start initialization shows the highest FSS values in the first hours of the simulation both with light and heavy precipitation. Finally, the ROC curve confirms the benefit of 4D-Var: the area under the curve is 0.91 compared to 0.88 for the control experiment without data assimilation.