Stochastic Modelling of Daily Rainfall

Stochastic generation of rainfall data offers an alternative to the use of observed records. This paper presents an evaluation of daily rainfall generation models at 21 stations across Australia. The models are the Transition Probability Matrix (TPM) method, the Daily and Monthly Mixed (DMM) algorithm, and a variation of each model. A goal of stochastic modelling is to generate synthetic data that are representative of the statistical characteristics of the historical data. One hundred replicates of length equal to the historical data were generated using the above models. Preliminary assessment of the models suggests that DMM models were unable to preserve the amounts of rainfall on solitary wet days and the TPM model needed to be modified by the empirical adjustment factor to preserve the annual variability.