Simulation of Daily Precipitation from a Nonparametric Renewal Model

Wet/dry spell characteristics of daily precipitation are of interest for a number of hydrologic applications (e.g., flood forecasting or assessment of erosion potential). Here, we examine issues related to designing an appropriate nonparametric scheme that focuses on spell characteristics for resampling historical daily precipitation data. A subset of the nonparametric wet/dry spell model presented in Lall et al (1993) is tested with synthetic data to justify the strategy proposed for applications. An application of the nonparametric wet/dry spell model to a Utah data set follows. Performance is judged on a set of statistical measures. Numerical comparisons of model performance with a parametric alternating renewal model are also offered for this data set. Our presentation stresses data exploratory aspects rather than formal hypothesis testing.