A process-guided hybrid Bayesian belief network to bridge watershed modeling and BMP planning

Planning of Best Management Practices (BMPs) is increasingly dependent on Process-based Watershed Models (PWMs) and suffers from large uncertainties. The large complexity and long runtimes of such models make prudential watershed management a difficult undertaking. In this study, we aimed to bridge the gaps between PWMs and uncertainty-based BMP planning with Bayesian Belief Network (BBNs). A new modeling framework of Process-guided Hybrid BBNs (PH-BBNs) was developed to represent the probabilistic cascade of critical modules in Soil & Water Assessment Tool (SWAT), a widely-used PWM, across external stressors (e.g., weather and various BMPs), parametric uncertainties, and watershed predictors. The PH-BBN modeling framework was used for decision support of Nonpoint Source (NPS) pollution mitigation in an intensively-cultivated area adjacent to Lake Dianchi, one of the three most eutrophic lakes in China. Our findings suggest that PH-BBNs can capture the critical pathways of water, sediment, and Total Phosphorus (TP) loss. Watershed projections are subject to large uncertainties to varying degrees in different landscapes. According to variance-based sensitivity analysis, pre-cipitation accounts for >80% of the projection variability, which underlines watershed vulnerability to climate change. As the effectiveness of parallel terraces, filter strips, and fertilization management would degrade with increasing rainfall intensity, they should be conservatively designed for BMP sustainability. Implementation of parallel terraces and filter strips is recommended as they are projected to reduce 80% of TP loads with >90% compliance confidence. PH-BBNs can render effective decision support for BMP risk assessment and adaptive watershed management under climate change.

Automated summary

This study developed a new modeling framework called Process-guided Hybrid BBNs (PH-BBNs) to bridge the gap between Process-based Watershed Models (PWMs) and uncertainty-based BMP planning. The findings suggest that PH-BBNs can effectively capture the critical pathways of water, sediment, and Total Phosphorus (TP) loss, and highlight the vulnerability of watersheds to climate change. The implementation of parallel terraces and filter strips is recommended for reducing TP loads with high compliance confidence.