Watershed modeling of BMP scenarios to improve agricultural water quality a case study in the San Joaquin River watershed, California

Quantifying effectiveness of agricultural BMPs at watershed scale is a challenging issue, requiring robust algorithms to simulate not only the agricultural production system but also pollutant transport and fate. This research takes this challenge to simulate and potentially improve the performances of BMPs in reducing organophosphates (OPs) runoff at field and watershed scales. A literature review with meta-analysis quantifies BMP effectiveness at field scale. Then, the SWAT model is calibrated and validated following a sensitivity analysis combining Latin Hypercube sampling and One-factor-At-a-Time simulation. Finally, the calibrated model is applied in the San Joaquin River Watershed and its sub-watershed Orestimba Creek Watershed to simulate BMPs including buffer strips, sediment ponds, vegetated ditches, use reduction, and their combinations. The meta-analysis revealed that buffer width, slope and vegetation were important factors in determining buffer's effectiveness. The model Y = K · (1- e -b·w) successfully captured the relationship between buffer width (w) and effectiveness (Y), where the estimated removal capacity (K) were 90.9 and 93.2 for sediment and pesticides. A 20 m buffer under favorable slope conditions (≈ 9%) would remove over 92% pesticides. The SWAT model was successfully calibrated with Nash-Sutcliffe coefficients over 0.92 and 0.82 for monthly simulation of diazinon and chlorpyrifos, respectively. Pesticide transport and fate is greatly impacted by surface runoff and their physico-chemical properties. BMP simulation suggested that combining vegetated ditches and buffer strips in addition to use reduction would decrease by over 94% the dissolved diazinon and chlorpyrifos. Buffer strips and vegetated ditches removed over 89% and 30% dissolved diazinon and chlorpyrifos, respectively, while sediment ponds removed 3-10%. Simulation of almond pest management practices found that OP concentrations in surface water were reduced by applying reduced-risk pesticides during summer and no spray during winter. This study has demonstrated that the SWAT model reasonably predict BMP effectiveness at watershed scale. However, the model can be further improved by enhancing the irrigation algorithm and by including more adjustable parameters to represent BMP mitigation processes. The findings can be widely applied to facilitate BMP implementation and evaluation through: (1) simulating BMP performance under various environmental conditions; (2) estimating annual pollutant removal, and (3) evaluating BMP design options.