Real-Time Model for Optimal Water Allocation in Irrigation Systems during Droughts

Abstract

The agricultural sector is the main victim of drought and efficient water planning, and management is the best strategy to reduce drought pressures on this sector. One operational approach is the application of real-time models to help water managers decide on mitigation measures, such as deficient irrigation or reducing cropped areas using the new incoming information (for example, recorded inflows and precipitation). The present paper introduces a real-time modeling approach for optimal water allocation during a drought. The model includes two main components: forecasting and optimization modules. The forecasting module uses a recurrent neural network technique to forecast annual inflows that is updated as monthly climate and hydrological data are introduced to the model. The optimization module allocates water among the irrigation units and their crops by considering growing stage, sensitivity to water stress at different stages, available/forecasted water, and previous decisions about water release. The model was tested for the 1999 drought of the Zayandeh Rud irrigation system. Traditional operating procedures were shown to produce 42% loss whereas the proposed method would have reduced loss to 12%.