The ARPAL atmospheric operational modeling chain and its applications: description and validation

The paper describes the model chain operational at the Meteo-Hydrological Center of the Liguria Region (ARPAL) based on the CNR-ISAC models BOLAM (BOlogna Limited Area Model) and MOLOCH (MOdello LOCale in Hybrid coordinates). Some of the chain applications and a statistical verification of its most recent implementation are also shown. The first operational run of the BOLAM numerical weather prediction (NWP) model at ARPAL was launched in September 1999, in correspondence with the beginning of the Mesoscale Alpine Programme Special Observing Period field campaign. Since then, the collaboration between CNR-ISAC and ARPAL has allowed to maintain and update the operational chain, also thanks to a continuous upgrade of the computational resources available at ARPAL. Since 2005, the non-hydrostatic model MOLOCH has been added to the forecasting chain. In the present operational setup, BOLAM is run over a European domain at the horizontal resolution of about 8 km, using ECMWF-IFS analysis and forecasts as initial and boundary conditions. MOLOCH model is nested in BOLAM, and it runs over a domain including the entire Italian territory, on a 1.5-km horizontal resolution grid. The forecast products, consisting in 72-h BOLAM and 48-h MOLOCH predictions, are available in 1.5 h after the reception of the ECMWF-IFS data. Four runs every day are performed. Other modeling systems in cascade are driven by NWP models. For several years, wind fields obtained from the atmospheric modeling chain have been used to force the wave model WAVEWATCH III, implemented at different resolutions over the Mediterranean basin. Other applications in cascade consist in hydrological prediction for the basins of the Liguria Region, wildfire, and ocean circulation modeling. Moreover, the different ini- tializations of the BOLAM and MOLOCH models provide an important contribution to the ARPAL operational Poor Man’s Ensemble prediction system. A verification based on 3 years of data available from the ARPAL ground observing net- work shows a general capability of the high-resolution models in better forecasting heavy precipitation events, due to a better description of convective phenomena, while a less marked improvement with respect to large-scale models is shown for low precipitation thresholds. Results also show an improvement of model performance for all monitored variables, precipitation, 2-m temperature, and 10-m winds, when resolution increases and when model domains are enlarged.