An agent-based-nash modeling framework for sustainable groundwater management

Abstract

An agent-based-Nash modeling framework has been developed to find a sustainable solution for groundwater management in Daryan Aquifer, Fars Province, Iran. This framework also includes a MODFLOW simulation model, an Artificial Neural Network (ANN), and a Non-dominated Sorting Genetic Algorithm-II (NSGA-II) optimization model. Groundwater state was simulated using MODFLOW and it was calibrated based on the measured data provided by Regional Water Organization (RWO) of Fars Province. In order to reduce the computational time, an ANN was trained and validated based on the input-output data of the MODFLOW model to estimate groundwater level. The validated ANN was linked to a nonhomogeneous elitist NSGA-II multi-objective optimization model to find a Pareto optimal front among the three objectives of reducing irrigation water deficit, increasing equity in water allocation, and reducing groundwater drawdown, as the objectives of the three main groundwater resource stakeholders; farmers, the government executive sector, and the environmental protection institutes. The Nash bargaining model was applied to the optimal solutions in order to find a compromise among the stakeholders. Social influential factors in the study environment, and policy mechanisms to encourage agents to cooperate with the management decisions were implemented in the agent-based model. These factors include training, incentives, penalties, and social norming (neighbors' impacts), as well as considering the executive and judicial systems. After application of the agent-based model, computed optimum solutions were modified according to social conditions. Finally, the Nash bargaining model was used again to find a compromise among modified optimal objectives of the stakeholders. Implementation of this solution led to 58.3% less water extraction and approximately 3 m water level uplift.