Abstract : 2E.5
Verification and intercomparison of precipitation fields modelled by LAMs in the alpine area: two FORALPS case studies

Stefano Mariani, Marco Casaioli, Christophe Accadia, Nazario Tartaglione
stefano.mariani@apat.it
Agenzia per la Protezione dell'Ambiente e per i Servizi Tecnici (APAT)

Sharing knowledge and experience about NWP verification among forecasting centres, environmental agencies and research centres in the alpine area is one of the main goals of the EU FORALPS project - INTERREG IIIB Alpine Space. In this framework, employed state-of-the-art verification techniques have been shared and applied to numerical simulations of specific case studies of rainfall events on the north-eastern alpine area. These techniques will be also applied in the frame of the APAT participation to MAP D-PHASE. In this work, simulations of two events (17-19/11/2002 and 9/9/2005) with three LAMs are discussed and verified with traditional and more innovative techniques. Models include two high-resolution non-hydrostatic models – WRF running at the Regional Meteorological Observatory (OSMER) of Friuli Venezia Giulia (Italy) and ALADIN operational at the Environmental Agency of the Republic of Slovenia (EARS) – and the hydrostatic 10-km QBOLAM operational over the Mediterranean basin at APAT. All models are forced by ECMWF analyses and forecasts, used as a basis for verification together with rain gauges and satellite. Verification techniques include traditional eyeball methods and non-parametric skill scores along with up-to-date object-oriented techniques as the Continuous Rain Area (CRA) analysis. Besides, simulation of satellite observed fields (METEOSAT-7 WV channel) are also employed. Both events are connected with the passage of small Mediterranean cyclones. The quality of the ECMWF forecast of these cyclones’ evolutions (trajectory and details) seems to depend on the model initialisation. In addition, space-time shift in the LAMs' predicted precipitation is found. Such a displacement varies among models. Object-oriented techniques – as CRA – are employed to provide a quantitative basis to these results. This is done with the aim of studying how the model error propagates from global model to LAMs, in connection with the configuration differences among LAMs.