A first-order Markov model for assessing rainfall discontinuity in Central Australia

For six stations in central Australia daily rainfalls were combined over five-day intervals. The resulting pentads were classed as either wet or dry according to whether a total of at least 0.1 inch of rain had occurred. A first order Markov model and a simple random model were constructed for the distribution of runs of wet and dry pentads. A χ2 test showed that the random model gives a poor fit to the observed frequencies. With the Markov model, however, good agreement was obtained with the exception of a few cases. Although the Markov model could be further improved it seems to be a practical statistical tool for assessing the long-term incidence of runs of wet or dry weather in central Australia.