Comparison of WPMM versus regression for evaluating Z-R relationships

The accuracy of the estimation of Z-R relationships is evaluated for the Window Probability Matching Method (WPMM) and regression methods. The evaluation is based on experiments of random subsampling of disdrometer-obtained 1-min reflectivity Z and rain-rate R pairs. The simulation of the disparity between the radar and the rain gauge measurement volumes was done by 3-min time averaging of the reflectivity data. Geometrical mismatch and synchronization inaccuracies between the radar and rain gauges are simulated by desynchronization of dt minutes, that is, shifting the R and Z time series with respect to each other by dt minutes. The WPMM and bias-corrected regression methods have similar skill in estimating rainfall accumulation even when geometrical and synchronization errors are introduced. However, the WPMM has significant advantage in estimating the rain intensities when geometrical and synchronization errors are introduced to the radar-gauge-measured Z-R pairs for simulating real-world radar and rain gauge comparisons. Regression-based Z-R relationships tend to overestimate the low rain intensities and underestimate the high rain intensities with the crossover at the estimated median rain volume intensity. This trend becomes more severe with the increased desynchronization. This reduction of the dynamic range of R does not occur when using WPMM. Although rain gauge bias correction may render the overall rain accumulation insensitive to the power of the Z-R law, its appropriate selection has a major effect on the partition of rainfall amounts between weak and strong intensities or the partition between convective and stratiform rainfall.