Quantifying the uncertainty of spatial precipitation analyses with radar-gauge observation ensembles

Sound quantitative precipitation estimates (QPE) are a key component of many hydrological and meteorological applications. The small-scale variability of precipitation and the limited coverage and accuracy of observations pose a major challenge to QPE precision. This study investigates a geostatistical method that combines radar and rain gauge data, to (a) generate best estimate precipitation fields and to (b) simulate ensembles of random precipitation fields that are consistent with the observations and represent the inherent analysis uncertainty. Daily precipitation data for 2008 covering the territory of Switzerland are used. Data transformation is applied to improve the compliance with model assumptions. The accuracy of the point estimates and the reliability of the probabilistic estimates from kriging with external drift are evaluated. A technical verification and plausibility experiments comparing the spatial uncertainty of the generated observation ensembles are performed.