On the skill of high frequency precipitation analyses

Abstract

The ability of radar-rain gauge merging algorithms to precisely analyse convective precipitation patterns is of high interest for many applications, e.g. hydrological modelling. However, due to drawbacks of methods like cross-validation and due to the limited availability of reference datasets on high temporal and spatial scale, an adequate validation is usually hardly possible, especially on an operational basis. The present study evaluates the skill of very high resolution and frequently updated precipitation analyses (rapid-INCA) by means of a very dense station network (WegenerNet), operated in a limited domain of the south-eastern parts of Austria (Styria). Based on case studies and a longer term validation over the convective season 2011, a general underestimation of the rapid-INCA precipitation amounts is shown, although the temporal and spatial variability of the errors is – by convective nature – high. The contribution of the rain gauge measurements to the analysis skill is crucial. However, the capability of the analyses to precisely assess the convective precipitation distribution predominantly depends on the representativeness of the stations under the prevalent convective condition.