Relative Importance of the Different Rainfall Statistics in the Calibration of Stochastic Rainfall Generation Models

Abstract

Stochastic rainfall generators are used in hydrologic analysis because they can provide precipitation input to models whenever data are not available, and their parameters are calculated so that the long-term statistics of the synthetic rainfall time series match those of the rainfall records. However, although mentioned in the literature, the relative importance of each rainfall statistic on the watershed response has not been addressed yet, and no guidance on how to account for it has been provided. In this paper, this relative importance is estimated and used to ponder each statistic differently in the calibration of rainfall generators so that it better reflects the watershed hydrology. Rainfall records of 1,249 rain gauges throughout the contiguous United States were used in the study. It was found that when synthetic rainfall time series are generated by weighting the precipitation statistics according to their relative importance, predicted runoff depths and peak flows are underestimated by 4 and 3%, respectively, whereas when they are generated by giving the same weight to all statistics, the underestimation is by 20 and 14%, respectively. These results, based on a significant number of rain gauges, confirm the benefit of weighing the statistics differently for watershed analysis.