Searching for systematic location errors of quantitative precipitation forecasts over the Calabria region

This article statistically analyses the location errors of the precipitation patterns forecast by three limited area models, namely the Fifth-Generation NCAR/Penn State Mesoscale Model (MM5), the QUADRICS BOlogna Limited Area Model (QBOLAM) and the Regional Atmospheric Modelling System (RAMS), over the Calabria region (Italy) for the period October 2000 - May 2002. Contiguous rain area (CRA) analysis is the diagnostic tool used to assess and quantify the position errors of the precipitation forecasts with respect to the observed precipitation patterns. Observation gridded analyses were obtained by means of the Barnes algorithm on the available rain gauge observations. Moreover, an approach to measure the quality of precipitation forecasts routinely by means of a global indicator called CRA Mean Shift (CMS) that summarizes the CRA verification outcomes is proposed. The CMS index would represent a statistical indicator of model quality in forecasting the correct positions of precipitation patterns. The model's tendency to misplace the forecast precipitation patterns towards a particular direction was tested by using a bootstrap procedure. All models seem to show statistically poor abilities in forecasting the correct precipitation pattern position over the verification domain considered. As far as the tendency towards a particular direction is concerned, only the RAMS model seems to show a systematic horizontal misplacement of precipitation patterns towards a particular direction.