Climate change and rapid urbanization intensify uncertainties in watershed water resource systems, challenging the effectiveness of integrated green infrastructure (GI) and water resource management strategies. This study investigates the socio-hydrologic dynamics of uncertain watershed systems under various water allocation schemes in the arid Colorado River Lower Basin, where GIs are implemented to enhance local water availability through rainwater harvesting. In this study, a bi-level multiagent framework (BL-MAUS) is developed that integrates fuzzy random variables with a bi-level multiagent architecture to model these complex interactions, integrating agent-based models that simulate the decision-making of watershed and urban water managers with hydrologic models that capture both city- and watershed-scale dynamics under social and hydrologic uncertainties. This coupled system enables the prediction of uncertain watershed behavior by modeling multi-agent interactions driven by specific water allocation schemes and GI implementation. The results show that a Water Trading Scheme (WTS) effectively minimizes water use costs but is highly sensitive to hydrologic uncertainties. In contrast, Water Quota (WQS) and Water Tariff schemes (WTaS) are found to improve the equity of water allocation. The implementation of GIs mitigates the adverse impacts of social uncertainty on costs and equity, notably reducing costs in WTS and WQS and enhancing equity in WTaS, although their overall impact remains limited-an outcome attributable to high construction costs, land constraints in dense urban areas, and the inherently local scale of GI benefits, though their effectiveness is strongly policy-dependent. This framework provides a robust tool for assessing watershed water resource systems under uncertainty, offering valuable governance insights for arid regions and advancing adaptive water management strategies globally.