Application of the LEPS technique for Quantitative Precipitation Forecasting (QPF) in Southern Italy: a preliminary study

This paper reports preliminary results for a Lim- ited area model Ensemble Prediction System (LEPS), based on RAMS (Regional Atmospheric Modelling System), for eight case studies of moderate-intense precipitation over Cal- abria, the southernmost tip of the Italian peninsula. LEPS aims to transfer the benefits of a probabilistic forecast from global to regional scales in countries where local orographic forcing is a key factor to force convection. To accomplish this task and to limit computational time in an operational implementation of LEPS, we perform a cluster analysis of ECMWF-EPS runs. Starting from the 51 members that form the ECMWF-EPS we generate five clusters. For each cluster a representative member is selected and used to provide initial and dynamic boundary conditions to RAMS, whose integrations generate LEPS. RAMS runs have 12-km horizontal resolution. To analyze the impact of enhanced horizontal resolution on quantitative precipitation forecasts, LEPS forecasts are compared to a full Brute Force (BF) ensemble. This ensem- ble is based on RAMS, has 36 km horizontal resolution and is generated by 51 members, nested in each ECMWF-EPS member. LEPS and BF results are compared subjectively and by objective scores. Subjective analysis is based on precipita- tion and probability maps of case studies whereas objective analysis is made by deterministic and probabilistic scores. Scores and maps are calculated by comparing ensemble pre- cipitation forecasts against reports from the Calabria regional raingauge network. Results show that LEPS provided better rainfall predic- tions than BF for all case studies selected. This strongly suggests the importance of the enhanced horizontal resolu- tion, compared to ensemble population, for Calabria for these cases. To further explore the impact of local physiographic fea- tures on QPF (Quantitative Precipitation Forecasting), LEPS results are also compared with a 6-km horizontal resolution deterministic forecast. Due to local and mesoscale forcing, the high resolution forecast (Hi-Res) has better performance compared to the ensemble mean for rainfall thresholds larger than 10mm but it tends to overestimate precipitation for lower amounts. This yields larger false alarms that have a detrimental effect on objective scores for lower thresholds. To exploit the advantages of a probabilistic forecast com- pared to a deterministic one, the relation between the ECMWF-EPS 700 hPa geopotential height spread and LEPS performance is analyzed. Results are promising even if addi- tional studies are required.