Interannual variability and extreme-value characteristics of several stochastic daily precipitation models

The effects of different formulations for the occurrence- and amounts-components of stochastic daily precipitation (i.e., `weather generator') models are investigated with respect to overall goodness of fit, and their capacity to represent observed interannual precipitation variability, extreme daily precipitation amounts, and long runs of consecutive wet and dry days. Daily precipitation data from 30 locations across the US, representing a wide range of precipitation climates, are considered. The conventional first-order Markov model for daily precipitation occurrence is found to be generally adequate for the central- and eastern US stations investigated, but is inferior to a number of more complex alternatives for the western stations. The commonly used Gamma distribution model for nonzero daily precipitation amounts is clearly inferior to the Mixed Exponential distribution for all locations. Practical implications of these findings for choosing a weather generator formulation are discussed.