Author: Huajin Chen

Publisher:

Published: 2018

Total Pages:

ISBN-13: 9780438290266

Over the last century, use of pesticides and fertilizers has enabled a tremendous increase in agriculture productivity worldwide. At the same time, documentation of contamination and adverse effects of agrochemical residues in the environment and aquatic ecosystem grows. A mechanistic and quantitative understanding of pesticide fate and transport in agricultural fields and watersheds is essential for informed decisions on regional water quality management. This study focuses on modeling agrochemical transport in California’s surface water and the pesticide removal efficacy of Vegetated Filter Strips (VFS’s). First, two newly-released, widely-used, field-scale numerical models, the Root Zone Water Quality Model (RZWQM) and the Pesticide Root Zone Model (PRZM), were compared and evaluated in terms of their abilities to predict sediment and pesticide runoff from two alfalfa fields in Davis, California. A composite metric (based on R2, NSE, d, and PBIAS) was developed and employed to ensure robust, comprehensive assessment of model performance. The RZWQM and PRZM models predicted surface water runoff reasonably well (absolute percent bias 31%) after adjusting important hydrologic parameters but with a consistent underestimation bias (-89% ≤ percent bias ≤ -36%) for sediment yield. Chlorpyrifos losses were simulated with reasonable accuracy especially for field A (absolute percent bias ≤ 22%), whereas diuron losses were underestimated to a great extent (-98% ≤ percent bias ≤ -65%) by both models. Based on this study, both RZWQM and PRZM performed well in predicting runoff that carried moderately adsorbed pesticides on an event basis, although the more physically based RZWQM is recommended when field-measured soil hydraulic properties are available. Next, to quantify VFS efficacy in reducing pesticide runoff, a meta-regression model was developed based on hydrologic responses of VFS’s, incoming pollutant characteristics, and interactions within and between these two factor groups (Q2 = 0.83). In cross-validation analysis, the developed model (Q2 = 0.81) outperformed the existing pesticide retention module of VFSMOD (Q2 = 0.72) by explicitly accounting for interaction effects and the categorical effect of pesticide adsorption properties. Based on the 181 data points studied, infiltration had a leading and positive influence on pesticide retention, followed by sedimentation and interactions between the two. Interactions between infiltration and pesticide adsorption properties were prominent, and the clay content of incoming sediment was negatively associated with pesticide retention. Next, we applied the Soil and Water Assessment Tool (SWAT) to watershed-scale simulation of pesticide fate and transport in the agriculturally intensive San Joaquin River watershed. This watershed is a major contributor to elevated pesticide levels in the downstream Sacramento-San Joaquin Delta where a biological weed control plan is under development. The Sequential Uncertainty Fitting version 2 algorithm was employed to perform calibration and uncertainty analyses. A combination of performance measures and standardized performance evaluation criteria was applied to evaluate model performance, and prediction uncertainty was quantified using the 95% Prediction Uncertainty band (95PPU). Results showed that streamflow simulation was at least “satisfactory” at most stations, with more than 50% of the observed data bracketed by the 95PPU. Sediment simulation was rated as at least “satisfactory” based on two performance measures, and diuron simulation was judged as “good” by all performance measures. The 95PPU of sediment and diuron bracketed about 40% and 30% of the observed data, respectively. Results also showed that the majority ( 70%) of agricultural diuron was transported during winter months, when direct exposure of biocontrol agents to diuron runoff is limited. Finally, the SWAT model together with the Sequential Uncertainty Fitting version 2 algorithm were applied to model riverine nitrate-nitrogen loads in the San Joaquin River watershed, where part of the area is poorly drained. We found that the new tile drainage routine in SWAT significantly improved nitrate load simulations. Nitrate simulations were rated as at least “satisfactory” for all the monitoring stations based on R2 and PBIAS. The P-factor was more than 0.5 and the average R-factor was about 1.2, indicating that a balance was reached between uncertainty measures and the strength of calibration. The major sources of uncertainty include the lack of information regarding the timing and magnitude of fertilizer and manure applications. The SWAT simulation indicated that 40% of riverine nitrate loads in the San Joaquin River watershed came from tile drainage nitrate. The San Joaquin River plays an important role in supplying nitrogen to the Delta by exporting 3,150 tons of nitrate-nitrogen annually. In summary, results from this dissertation indicate that: (1) preferential flow through macropore is an important transport pathway in undisturbed soil and mechanistic hydrologic models should be used to account for such field conditions; (2) interactions among hydrological and pesticide adsorption processes have significant effects on VFS pesticide retention; and (3) dominant hydrologic processes and associated parameter sensitivity vary spatially, and therefore it is important to perform rigorous multi-site calibration when simulating pollutant transport across large watersheds. The study not only facilitates a deeper understanding of hydrologic processes occurring in croplands, VFS’s, and agricultural watersheds, but also serves as a foundation for the continued research on calibration, evaluation, and uncertainty analysis of spatially distributed, physically based hydrologic models.