Agent-based simulation of equity-oriented water reallocation under drought: A case study of the Geum River Basin

Study region : Geum River Basin, South Korea, a structurally heterogeneous watershed with 14 agricultural, 12 domestic/industrial intake, and 3 hydropower stakeholders. The basin has the highest agricultural water demand among South Korea’s six major river basins, with precipitation concentrated in the summer monsoon season, making dam operations critical for balancing competing water demands during irrigation. Study focus : This study develops an integrated framework to improve equitable water allocation during droughts by coupling a basin-scale allocation model with an Agent-Based Model (ABM). While centralized optimization overlooks decentralized social dynamics, autonomous stakeholder agents make voluntary diversion decisions based on locally perceived reliability deficits, redistributing water to the most vulnerable peers via hypothetical conveyance infrastructure. Streamflow is represented by SWAT-derived daily inflows spanning 39 water years (October 1981–September 2020). New hydrological insights and conclusions : First, agricultural water supply vulnerability in the Geum River Basin is governed less by absolute demand magnitude and more by basin topology and infrastructure configuration. Tributary-dominated subbasins (Agr 7, Agr 11, and Agr 13), which are structurally isolated from the regulated mainstem, exhibited the lowest supply reliability during the severe historical droughts of 1994–1997 and 2014–2017, identifying these locations as priority targets for adaptive drought management in the basin. Second, agent-based simulation of voluntary water donation among Geum River stakeholders demonstrates that equity improvements can emerge as system-level outcomes from repeated decentralized interactions. During drought period, the agricultural Gini coefficient decreased from 0.152 (no ABM) to 0.097–0.117 across three behavioral scenarios, with the largest reliability gains concentrated at the most structurally vulnerable tributary nodes. Third, sensitivity analysis across nine supply–demand combinations confirmed that the simulated equity benefit within the Geum River Basin is robust to variations in inflow magnitude and irrigation demand scaling, with gains increasing proportionally under higher demand pressure. However, the Pro-social scenario yielded diminishing equity returns relative to the Bounded Rational scenario, as excessively frequent donations reduced the reliability of intake donor stakeholders, suggesting that moderate rather than maximum cooperation produces the most balanced outcomes across stakeholder types.