As a contribution to the Mesoscale Verification Inter-Comparison over Complex Terrain (MesoVICT) project, the present work investigates how the variation of the model domain size and native resolution can affect the application of the contiguous rain area (CRA) analysis on quantitative precipitation forecast (QPF) verification. The investigation is based on the analysis of two of the six MesoVICT core events monitored during the forecast demonstration phase of the Mesoscale Alpine Programme (referred to as MAP D‑PHASE), and the Convective and Orographically-induced Precipitation Study (COPS), which occurred respectively on 20–22 June and 25–28 September 2007, plus one additional high impact event that took place on 22–24 November 2007 at the end of the MAP D‑PHASE Operations Period (DOP). The MesoVICT test-bed area is centred over Central Europe and it covers a complex terrain region characterized by the presence of the Alps (i.e., complex orography) and the Mediterranean Sea (i.e., lack of observations, coastlines). Specifically, the CRA analysis is applied to 24‑h QPF fields provided by four different configurations of the hydrostatic BOlogna Limited Area Model (BOLAM), with horizontal grid size ranging between 0.07° and 0.1° and model domain extending from only the Alpine area to the entire MesoVICT test-bed area, and by an ad‑hoc MesoVICT customized configuration of the convection-permitting MOLOCH model, with a horizontal grid size of 0.0225° and model domain covering the entire MesoVICT test-bed area. Rainfall measurements collected during DOP are objectively analysed on the verification grid using a two-pass Barnes scheme in order to be used for the spatial verification analysis. The results confirm that the correlation-based pattern matching criterion is a more suitable basis for the applicability of the CRA approach than the Mean Square Error (MSE) criterion. The limits of the latter criterion are also discussed. The extent of the model domain affects the CRA application in a way comparable to the effect of data-void areas in the observational analysis fields and in the QPFs. Sensitivity to the increase of the model (native) resolution is found as well, especially when a complex spatial structure characterizes the rainfall event under investigation. A quality control of the CRA outcomes is then recommended to accurately evaluate the diagnosed spatial forecast errors and to distinguish reliable results from suspicious ones. The results suggest the need for future studies that apply the CRA analysis at shorter rainfall integration times and on sub-structures of the rainfall fields.
Full Text Finder