ARPA-Piemonte's special weather observations network in the Alps and Torino plains is used for model initialization within a running start data assimilation strategy, and for verification of the resulting 24-h forecasts. Mesoscale models are run with and without data assimilation to determine the added value of various meteorological input data. The observational data were also used for dynamic analysis where they are continuously assimilated within a mesoscale model throughout the 24-h periods to produce a more complete and dynamically consistent meteorological analysis than that provided by the observations alone.
The meteorological model forecasts for six cases representing the range of weather conditions observed over the study region during February 2006 generally show improved predictive skill when using data assimilation and increasing model resolution, especially at the surface and in the boundary layer. Statistical differences were relatively small between the 4-km and 1.3-km grids, although subjective analysis revealed greater mesoscale details using 1.3-km resolution. This result poses important questions regarding the input meteorological data requirements for AT&D models and the allocation of computational resources for numerical weather prediction.