Impact of uncertainty in behavioural factors on irrigation demands

In this paper, analyses of uncertainty in the results of modelling irrigation demands in two irrigation areas are presented. The Next GENeration IRRigation (NGenIrr) demand model has been used for the analyses. This model incorporates behavioural and biophysical irrigation demand factors and associated uncertainties. On-farm decision making is represented using compromise programming, where tradeoffs between conflicting objectives are modelled, e.g. when deciding on crop areas at the start of an irrigation season, the model chooses crop mixes that achieve a compromise between maximising gross margins and minimizing risk of suffering a water shortage during the season. In the NGenIrr model, behavioural factors affect crop areas and irrigation scheduling. Biophysical factors affect crop water usage, soil water balances and on-farm storage volumes. Several irrigation seasons with varying climatic and water availability conditions were simulated for Shepparton and Coleambally Irrigation Districts. After establishing satisfactory performance of the model for the two districts, the uncertainties in behavioural and biophysical parameters were each varied systematically. The measure used to determine the effects of these changes was the number of observed data points remaining outside the 90% confidence interval of the model outputs. The modelling results show that the contribution of uncertainty in demand estimates from uncertainty in behavioural and biophysical parameters are likely to vary from region to region. In annual cropping regions, uncertainty in human behavioural factors can play a more important role than uncertainty in biophysical factors. Thus for the Coleambally Irrigation District, where annual cropping dominates, the uncertainty in behavioural factors contributed more to the uncertainty in demand estimates. However, in the Shepparton Irrigation District, which is dominated by perennial and horticultural crops, uncertainty in biophysical factors can account for most of the uncertainty in irrigation demand estimates. The results for both districts showed that with minimal modelling uncertainty the number of points outside the 90% confidence interval was quite high: 43 and 47 for Shepparton and Coleambally respectively (out of 60 points). For Shepparton, the number of outlying points decreases asymptotically to 19 as more and more uncertainty is added to the model. For Coleambally, the number of outlying points decreases to 5.