5 year radar-based rainfall statistics: disturbances analysis and development of a post-correction scheme for the German radar composite

Abstract. A radar-based rainfall statistic demands high quality data that provide realistic precipitation amounts in space and time. Instead of correcting single radar images, we developed a post-correction scheme for long-term composite radar data that corrects corrupted areas, but preserves the original precipitation patterns.

The post-correction scheme is based on a 5 year statistical analysis of radar composite data and its constituents. The accumulation of radar images reveals artificial effects that are not visible in the individual radar images. Some of them are already inherent to single radar data such as the effect of increasing beam height, beam blockage or clutter remnants. More artificial effects are introduced in the process of compositing such as sharp gradients at the boundaries of overlapping areas due to different beam heights and resolution. The cause of these disturbances, their behaviour with respect to reflectivity level, season or altitude is analysed based on time-series of two radar products: the single radar reflectivity product PX for each of the 16 radar systems of the German Meteorological Service (DWD) for the time span 2000 to 2006 and the radar composite product RX of DWD from 2005 through to 2009. These statistics result in additional quality information on radar data that is not available elsewhere.

The resulting robust characteristics of disturbances, e.g. the dependency of the frequencies of occurrence of radar reflectivities on beam height, are then used as a basis for the post-correction algorithm. The scheme comprises corrections for shading effects and speckles, such as clutter remnants or overfiltering, as well as for systematic differences in frequencies of occurrence of radar reflectivities between the near and the far ranges of individual radar sites. An adjustment to rain gauges is also included.

Applying this correction, the Root-Mean-Square-Error for the comparison of radar derived annual rain amounts with rain gauge data decreases from 1181 to 171 mm a⁻¹ (application period: 2005, 2006 and 2009) and from 317 to 178 mm a⁻¹ (validation period: 2007, 2008). The entire correction scheme is applicable on an annual scale. The correction of mean annual rain amounts derived from radar composite data for the whole period leads to an almost undisturbed and homogeneous distribution of rain amounts for Germany.