Generalisation of a hybrid model for point rainfall

Abstract

A generalised hybrid model to generate point rainfall for a wide range of aggregation levels is presented in this paper. The rainfall process is expressed as a product of a binary chain model, which preserves the dry and wet sequences as well as the mean, and a correlated jitter used to improve the deficiencies in the second-order properties of the binary chain. Analytical derivations of the moments of a binary chain are presented. As the jitter model the exponential of a second-order autoregressive Gaussian process is selected. Two possible binary chain models are analysed, a non-randomised Bartlett–Lewis model and a Markov chain. Although both binary chain models perform equally well, the Bartlett–Lewis model is preferred for reasons of parameter parsimony.