Comparison of four radar profiling algorithms in the S‐band

Abstract

AbstractRadar profiling algorithms are widely used for precipitation‐induced path‐attenuation correction. In this study, we investigate the impact of the four well‐known profiling algorithms on the efficacy of quantitative precipitation estimation retrievals. The first method, named the “simple PhiDP‐based method”, distributes precipitation‐induced path attenuation, or PhiDP (differential phase shift), of a rain segment to each radar gate according to radar reflectivity (Zh). The second one is a linear version of the ZPHI method derived from an approximate linear relationship between specific attenuation (Ah) and specific differential phase (KDP). The third one is a nonlinear version of the ZPHI that respects the nonlinearity of the Ah–KDP relationship. The fourth method is inspired from the self‐consistent method of Bringi. The two‐moment normalized raindrop size distribution (DSD) with two variables, scaled drop concentration and mean drop diameter , is used to represent natural variability of DSD. Two configurations are considered: a constant over a segment and a variable over a segment. The Ah, KDP, and rain intensity are estimated by the four profiling algorithms. The sensitivity of each algorithm to radar measurements uncertainty is also investigated. The evaluation of the estimates shows the following. First, for the range‐constant situation, the nonlinear ZPHI and the self‐consistent method yield the best estimates if the measurements errors are not considered. However, the nonlinear ZPHI and the self‐consistent method are significantly impacted by the Zh bias and the measurements noises respectively. Second, for the range‐variable situation, the self‐consistent method does not have a good performance even when radar measurements are accurate. The performance of the nonlinear ZPHI method is similar to the simple PhiDP‐based method, whereas the latter is immune to the Zh calibration error. These results suggest that the benefit of introducing in quantitative precipitation estimation algorithms can be significantly reduced if is not constant in reality and/or if the radar measurements are biased and noisy.