Prediction of Rainfall for Short Term Irrigation Planning and Scheduling—Case Study in Victoria, Australia

Abstract

In the short-term planning (7–14 days) and operation of complex irrigation systems, an estimate of irrigation water demand (IWD) is of a fundamental concern. To predict the IWD, a reliable estimate of the expected rainfall during any irrigation period is of fundamental importance. Rainfall is generally predicted with a certain probability of exceedance. However, the standard flood flow-frequency distributions cannot be used for prediction of rainfall of such short durations because these rainfall series in general consist of zero values. Two methods, the total probability theorem (TPT) with three normalizing transformations (i.e., power, log, and square root), and the leaky law (LL) were used to predict the rainfall of short durations (7–14 days, depending upon the number of irrigations per season) in the Goulburn irrigation area (GIA) of Victoria, Australia. Investigations indicated that the TPT using the power transformation (TPTP) was more effective in modeling the short-term data series than the log (TPTL) and square-root transformations (TPTS). Although the overall fitting of the short-term rainfall data series by the LL method was significantly (99%) better than the TPT method, some series could not be fitted by the LL method. This revealed that the LL method could not model all short-term rainfall data series. Results showed that although both the TPT and LL were quite satisfactory in predicting short-term seasonal rainfall of short durations in the study area, none of them was individually able to model all the short-term rainfall series. Hence, the joint use of TPT and LL methods was recommended for short-term rainfall prediction of the study area.