Non-linear stochastic optimization for agricultural planning under future surface-water scarcity in the area downstream of the Karkheh Dam, Iran

Water resources are increasingly vulnerable to climate variability and land use change, challenging agricultural planning in arid and semi-arid regions. This study developed a stochastic non-linear optimization framework for the area downstream of the Karkheh Dam, Iran, under projected surface-water scarcity. Published Soil and Water Assessment Tool (SWAT)-based projections defined 12 representative water-availability states, while an multi-layer perceptron (MLP) inflow–outflow model estimated seasonal downstream availability. Three management scenarios were evaluated in GAMS. Scenario 1 retained the current cropping pattern and irrigation efficiency. Scenario 2 imposed a wheat-only strategy, increased allowable cultivated area by 10%, and raised overall irrigation efficiency by 10 percentage points. Scenario 3 adopted a water-conserving strategy limited to barley and open-field vegetables, with the same efficiency improvement. Compared with the wheat-centred strategy, Scenario 3 reduced agricultural water requirements and improved physical water productivity. The framework demonstrates how hydrological projections, reservoir-outflow estimation, and optimization can support adaptive agricultural planning under constrained surface-water availability.