Large Sample Nonparametric Modeling of the Uncertainties in Radar Rainfall Products

The high level of uncertainties in radar rainfall (RR) estimates is a broadly acknowledged problem. However, comprehensive quantitative information about their characteristics is not available. To fill this gap, we have been developing an empirically-based approach to the quantification of the functional-statistical error structure in RR products. Our goal is to deliver a realistic mathematical model describing the dependence of the frequency distribution of the errors on RR magnitude under variable conditions. At present, the considered conditions include different distances from the radar, seasons, time-scales and rainfall estimation parameters. Later, when relevant data samples become available, we will analyze the dependences of the uncertainty model on different spatial resolutions, geographic locations, climatic regimes and RR estimation algorithms.

The results presented here are based on the 6-year-long sample of the Level II reflectivity data from the Oklahoma City station (KTLX) of the NEXRAD system. Several sets of rainfall products are generated uniformly with the Built 4 of the PPS algorithm using different setups of its parameters. These radar estimates are completed with the corresponding raingauge data from the Oklahoma Mesonet and the ARS Micronet covering the radar umbrella. The raingauge data are used as a ground reference (GR) to estimate the uncertainty characteristics using the nonparametric functional estimation methods. In this presentation, we describe the selected results of this large-sample uncertainty analysis of the RR products. We also discuss the application of the error modeling to the probabilistic quantitative precipitation estimation (PQPE).