Quantification of uncertainty in irrigation scheduling and crop growth simulations using neural network based inflow forecast models

Crop growth simulation models play major roles in an efficient irrigation water management. The effectiveness of such models depends on reasonable and accurate estimate of irrigation water requirement and crop yield. This requires an optimization framework which integrates inflow forecast and reservoir operation models with crop growth simulation models. Generally lack of knowledge about the physics of the system makes the simulation models to have assumption and simplifications, which in turn induces varying levels of complexity in the model. Consequently, the complex and stochastic nature of systems to be modeled introduces lot of uncertainty in the simulations. Therefore, in order to make effective decisions, it is apparent that one need to quantify the uncertainty associated with the model simulations. In this study uncertainty of irrigation schedules developed using simulation models is established so that an effective decision on better water management can be performed. An artificial neural network based hydrologic model is employed for the reservoir inflow forecast with quantification of predictive uncertainty. The quantified level of uncertainty in forecasted inflow is then propagated through reservoir simulation, crop growth simulation models, and the associated uncertainty in the reservoir releases and the crop yield are estimated. The proposed framework is applied to a real world case study of a command area in southern India. Some of the preliminary results are presented in Table 1 and Figure 1. It is observed from Table 1 that there is 5% and 9% uncertainty respectively in total crop yield and amount of irrigation volume with respect to the mean value simulation. The Fig 1 depicts the irrigation schedules suggested by the simulation model considering the upper and lower uncertainty bounds of the forecasted inflow. The results suggest that the reasonable quantification of uncertainty helps in effective decision making of irrigation scheduling which is in turn improves the confidence of irrigation system. Fig 1 Uncertainties estimated in irrigation during Kariff season (2005) Table 1 Summary statistics of estimated uncertainty