Use of Lightning and Storms Life Cycle Information in Radar Rainfall Estimation

Radar rainfall estimates derived from conversion of reflectivity are known to contain systematic and random errors (bias) that limit the quantitative use of radar rainfall in various applications. Enhancement of radar rainfall estimates is normally accomplished through gauge-adjustment procedures; however, those procedures require long integration periods to moderate sampling differences between the two sensors. To improve the efficiency of radar-gauge based adjustment techniques, classification based on storm microphysical information is needed. In this study, use of two auxiliary data derived from cloud-to-ground (CG) lightning measurements and a storm tracking algorithm was performed. This information was used to classify storms into thunderstorms (storms associated with CGs) vs. showers (storms without lightning) and according to the storm's maturity stage (i.e., growing, mature or decay stage). The radar rainfall data from South Florida Water Management District and for a period of twenty months were used. The radar rainfall estimates at 2-km resolution and 15-min time intervals, and corresponding rain gauge measurements from 120 gauges, and CG occurrences from the National Lightning Detection Network were used. Tracking was applied to the radar rainfall data to identify the storm families used in this analysis. The radar error analysis for different storm types and storm stages indicates that precipitation microphysical information is critical for improving radar rainfall estimation. It has shown that radar rain estimates tend to give stronger biases in storms of strong convective nature (both in terms of lightning occurrences and in terms of the growing stage of the storm), while showers and storms at mature stage tend to be better represented by the fixed climatological Z-R relation. Results from this study demonstrated that CG lightning and storm's maturity stage information could help reduce the variability in the Z-R conversion. Such classification may have a consequential effect on improving the efficiency of a mean-field bias-adjustment algorithm applied at higher temporal scales of aggregation (hourly, daily, etc.) for the radar rainfall estimate.