Fine-scale rainfall nowcasting

Comparison was made of a number of traditional radar adjustment or radar – rain gauge merging methods, including the Mean Field Bias, Range-dependent and Brandes adjustment methods; kriging based rain gauge interpolation to radar methods (e.g. Kriging with External Drift: KED) as well as co-kriging, Kalman filter based and Bayesian merging methods. Also, more advanced methods based on dynamic adjustment and a newly proposed approach based on quantile mapping were tested.

General conclusion from that comparison was that radar based rainfall estimates need adjustment to or merging with rain gauges in order to remove the bias. Most methods perform well, but Bayesian merging methods appear to be the better ones. One general shortcoming identified for the existing methods, is that they often spatially smooth out the peak rainfall intensities during extreme thunderstorms. Given that these peak intensities are of utmost importance for urban hydrological applications, a new advanced merging method has been developed. The method identifies the extreme rain cells in radar images as local singularities and explicitly addresses these singularities in the merging process.

Click on the links below to access information (scientific papers) on these methods and comparisons.