The role of meteorological and hydrological uncertainties in the performance of optimal water allocation approaches

Efficient reservoir operation and irrigation scheduling are important for the mitigation of water shortages in Iran. For more accuracy, the hydrological and meteorological uncertainties associated with reservoirs and farm levels should be considered. The major contribution of the current paper is to evaluate the uncertainties of evapotranspiration (ET) and inflow, and the issue of constant/variable agricultural demand (CAD/VAD) for optimal irrigation scheduling and reservoir operation. Some optimization approaches were employed and compared during a drought episode in the Zayandeh-Rud agricultural system. Approaches include: (i) DP-CAD: dynamic programming (DP), considering CAD and no inflow uncertainty; (ii) SSDP-CAD: sampling stochastic DP (SSDP) with CAD and inflow uncertainty; (iii) LP-NLP-VAD: implementing linear (LP) and non-linear programming (NLP) modelling for crop types, growing stages, and irrigation systems under deterministic conditions; (iv) SDP-NLP-VAD: similar to the third approach, but considers ET uncertainties using a stochastic DP (SDP) rather than an LP model, and uses stochastic crop yield functions in the NLP formulation. DP-CAD and SDP-NLP-VAD were the simplest and most complicated modelling processes, respectively. SDP-NLP-VAD was the most time-consuming to reach a steady state and a global optimal solution. The LP-NLP-VAD and SDP-NLP-VAD approaches, which account for variability in crop water requirements, conservatively consider water shortages and reservoir release.