Systematic and random radar rainfall error and its impact on rainfall-runoff prediction accuracy

The radar rainfall uncertainty is decomposed conceptually into mean-field and range-dependent errors. Due to the natural variability of rainfall and geometry of radar beam both errors are expected to have a systematic and random component. This study aim at evaluating statistical parameterizations for the above errors associated with widespread events occurring over mid-latitude mountainous regions. Furthermore, the study aims at understanding the relative impact of the two error components in runoff prediction accuracy for varying basin characteristics, including runoff generation mechanism, size, and shape. Finally, we attempt to assess the improvement on runoff through correction of the systematic component of radar rainfall error using existing techniques. The database is comprised of nine flood inducing storms measured by ground-based radar and a dense raingauge network in north-eastern Italian Alps, and three medium size (77-116 sqkm) watersheds with varying characteristics. One watershed is fully mountainous, while the other two vary from hilly to flat. Radar raingauge comparisons at close ranges (< 40 km) are used to evaluate the statistical radar error models. Multiple data-based simulations are performed using the evaluated radar error models, time series of surface radar rainfall fields, and calibrated watershed models to assess rainfall and runoff error statistics for both corrected and uncorrected radar rainfall estimates.