Tests of a space-time model of daily rainfall in southwestern Australia based on nonhomogeneous random cascades

A space-time rainfall model is constructed to generate synthetic fields of spacetime rainfall, with a daily time step. The model has two components: a temporal model based on a first-order, four-state discrete Markov chain wh ich generates a daily time series of the regionally averaged rainfall and a spatial model based on a nonhomogeneous random cascade process which disag gregates the above regionally averaged rainfall amounts to produce spatial patterns of daily rainfall across the region. The cascade generators used t o disaggregate the rainfall spatially are a product of stochastic and deter ministic factors; the latter enable the model to capture systematic spatial gradients exhibited by measured data. The model is applied to a 400 km x 4 00 km region encompassing the Swan-Avon River Basin in the southwest of Wes tern Australia. The model parameters are estimated on the basis of 11 years of observed daily rainfall data from 490 rain gauges located in the region . A detailed testing of the model is undertaken on the basis of a compariso n of the statistical characteristics of the spatial and temporal variabilit y of rainfall between the rainfall fields obtained from the rain gauge netw ork and those generated by the simulation model. It was found that the spat ial and temporal characteristics of the simulated rainfall field are in ver y good agreement with those of observed rainfall. In particular, the model preserves the observed systematic spatial variations of daily, monthly, and annual rainfall across the region and the observed seasonal variations. St atistical characteristics of storm durations and interstorm periods at indi vidual gauge locations are not reproduced that well, which suggests general izations need to be made to include space-time correlations.