Modeling Farmers’ Irrigation Behaviors under Risk Using a Distributed Hydrological-Agent Model

Abstract

The effective use of precipitation forecast enables farmers to optimize rainfall utilization, significantly supporting irrigation management. However, previous studies have seldom accounted for farmers’ irrigation behaviors under risk, often leading to inaccurate estimations of forecast value. To address this issue, this study proposes a distributed hydrological-agent model that couples an agent-based decision model for irrigation and the SWAT–MODFLOW hydrological model. The decision model applies prospect theory to construct a utility function that quantifies farmers’ satisfaction with irrigation under risk. This study focuses on revealing how forecast skill influences farmers’ irrigation decisions and hydrological responses. Results show that forecast skill has an increasing marginal benefit on increasing farmers’ trust levels and utility. Specifically, an increase in the forecast skill score (FSS) from 0–1 corresponds to a 7% reduction in irrigation water use and a threefold increase in utility. Whereas utilizing forecast information for irrigation has minimal impact on crop yield, it leads to a 14%–22% reduction in return flow and a 0–0.012 m annual decrease in groundwater levels. Sensitivity analysis shows that farmers’ irrigation behavior is price-sensitive only when FSS is between 0.3 and 0.7. High water-saving rewards enhance farmers’ trust levels, whereas high crop prices lead to risk aversion and excessive irrigation. These findings provide valuable insights for managers to guide farmers effectively utilize forecast information, thereby supporting agricultural water management.