A nonparametric model for stochastic generation of daily rainfall occurrence

This paper presents a nonparametric model for generating single-site daily rainfall occurrence. The model is formulated to reproduce longer-term variability and low-frequency features such as drought and sustained wet periods, while still reproducing characteristics at daily time scales. Parsimony is achieved within a Markovian framework by using "aggregate'' predictor variables that describe how wet it has been over a period of time. The use of a moving window to give a seasonally representative sample at any given time of year ensures an accurate representation of the seasonal variations present in the rainfall time series. Actual simulation proceeds by resampling from the historical record of rainfall occurrence, conditional to the current values of the associated predictors. The model is applied using historical daily rainfall occurrence from Sydney, Australia. We find that the use of multiple predictors specified at a daily level and at seasonal, annual, and multiyear aggregation levels leads to sequences that more closely reproduce the longer-term variability present in the historic records, compared to sequences produced by models incorporating only one or two predictors.