Verification of Quantitative Precipitation Forecasts via Stochastic Downscaling

Abstract

The use of dense networks of rain gauges to verify the skill of quantitative numerical precipitation forecasts requires bridging the scale gap between the finite resolution of the forecast fields and the point measurements provided by each gauge. This is usually achieved either by interpolating the numerical forecasts to the rain gauge positions, or by upscaling the rain gauge measurements by averaging techniques. Both approaches are affected by uncertainties and sampling errors due to the limited density of most rain gauge networks and to the high spatiotemporal variability of precipitation. For this reason, an estimate of the sampling errors is crucial for obtaining a meaningful comparison. This work presents the application of a stochastic rainfall downscaling technique that allows a quantitative comparison between numerical forecasts and rain gauge measurements, in both downscaling and upscaling approaches, and allows a quantitative assessment of the significance of the results of the verification procedure.