Numerical assimilation of the national radar mosaic reflectivity for a moderate-heavy precipitation event in Central Italy

In this work, we show the application of a methodology to assimilate radar reflectivity in the Regional Atmospheric Modeling System (RAMS). The methodology is derived from Caumont et al. (2010) and consists of a 1D retrieval followed by a 3D-Var assimilation. In the 1D retrieval a model-driven relative humidity profile is derived in correspondence with radar reflectivity observations. In particular, a Bayesian method is used to define a relative humidity profile weighting the relative humidity profiles simulated by the model in the nearby points. The weight assigned to each profile depends on the difference between observed and simulated reflectivity. The greater the agreement between the observed and simulated radar reflectivity, the larger is the weight assigned. Then, the 1D retrieval of relative humidity is used as pseudo-observation in the 3D-Var analysis. The method is applied to a moderate-heavy precipitation case occurred in Central Italy on 16 September 2017, which was underestimated by the operational configuration of the model, not assimilating radar reflectivity. The precipitation occurred mainly in two different phases, at the start of the day (00-06 UTC) and at the end of the day (18-00 UTC) and it was observed by the National radar mosaic managed by the department of civil protection. For this case, the RAMS is run at 4 km horizontal resolution over the whole Italy, and analyses assimilating radar reflectivity are available hourly for the whole day. The results for the daily precipitation show that the assimilation of radar reflectivity over the whole Italian domain has a fundamental impact on the rainfall simulation. In particular, performing a simulation that assimilates continuously the radar reflectivity gives a good representation of the precipitation field that is largely underestimated by the configuration without radar assimilation. Performing very short term forecast of the periods 00-06 UTC and 18-00 UTC shows also that the assimilation of radar reflectivity is very important to simulate the intense precipitation for the analyzed case. Each simulation period is divided into two tri-hour forecast intervals (for example the period 00-06 UTC is divided in 00-03 UTC and 03-06 UTC) preceded by a 6h simulation that assimilates radar reflectivity. Finally, the forecast of the event for the two phases of heavy precipitation (00-06 UTC and 18-00 UTC) using assimilation of radar reflectivity is compared with a similar forecast assimilating lightning, following the methodology of Federico et al. (2017). References Caumont et al. : 1D+3DVar assimilation of radar reflectivity data: a proof of concept, Tellus (2010), 62A, 173-187 Federico, S., Petracca, M., Panegrossi, G., and Dietrich, S.: Improvement of RAMS precipitation forecast at the short-range through lightning data assimilation, Nat. Hazards Earth Syst. Sci., 17, 61-76, doi:10.5194/nhess-17-61-2017, 2017.