Rainfall persistence: Detection,modelling, costs and value of probability information

The basic hypothesis discussed in this paper is that the presence of persistence or bunchiness in rainfall sequences reduces farm profitability. After a review of past investigations into rainfall persistence, statistical tests are shown to indicate positive inter-temporal correlations in excess of seasonality for weekly and monthly precipitation data from three recording stations in the northern Brigalow region of Queensland. A relatively simple procedure is advanced for generating artificial rainfall sequences containing persistence. Simulation experiments with a grazing-property model incorporating this synthesis procedure are used to test the basic hypothesis. Results indicate substantial reductions in both income and the rate of asset accumulation due to persistence in weekly rainfall. Finally, provision of meteorological information on persistence in the form of conditional rainfall probabilities is proposed, and a simulation and decision theory approach is suggested for determining the potential value of such information.