Alternative Approach for Estimation of Precipitation Using Doppler Weather Radar Data

Precipitation estimates from Doppler weather radar (DWR) provide much better spatial resolution as compared to rain gauges and are therefore becoming more popular in hydrological applications. However, traditional estimates of precipitation from radar-measured reflectivity (e.g.,Z=aRb) are deterministic and thus do not offer any information about the uncertainty associated with the estimate. However, the radar scans may contain significant errors that propagate to the rainfall estimates. This gives rise to the need for the probabilistic estimates of rainfall. This paper proposes a copula-based approach to obtain the joint cumulative distribution function (CDF) of reflectivity (Z) and precipitation (R) from which the conditional CDF of precipitation is determined. Three copulas are implemented, and the temporal and spatial transferability of each model is evaluated using different measures of performance. It is established that the precipitation estimates are better than those obtained from the four different existing methods, including the traditional approach. In addition, uncertainty estimates are also available from the proposed approach. Though the mean (50th quantile) of the probabilistic estimates does not correspond well to the gauge rainfall data, availability of the estimates at various confidence levels makes the proposed copula-based approach suitable for different applications.