Management-oriented sensitivity analysis for pesticide transport in watershed-scale water quality modeling using SWAT
Introduction
In recent years, non-point source pollution from agriculture is increasingly responsible for the degradation of surface water quality. This, in turn, increases the need of integral water-quality management with enhanced hydrologic models. Off-site movement of agrochemicals, such as organophosphate (OP) pesticides, to streams and aquifers, in agricultural watersheds, may potentially cause adverse effects on human health and ecosystem. OP pesticides are widely used in orchards and other crops. According to the pesticide market estimates by U.S. Environmental Protection Agency (USEPA, 2004), OP use as a percent of total insecticide use has increased from 58% in 1980 to 70% in 2001. In the San Joaquin Valley of California, one of the most productive agricultural regions in the world, about 450 tons of active ingredients of OP pesticides were used per year from 1990 to 2007, and OP pesticide residues have been routinely detected in surface water bodies of the San Joaquin River watershed. According to the sampling results during 1992–1995, pesticide levels in 37% of the streams in the San Joaquin Valley exceeded the criteria for the protection of freshwater aquatic life (Dubrovsky et al., 1998).
GIS-based distributed or semi-distributed modeling is widely applied to simulate chemical transport and predict pollution reductions with management practices in agricultural watersheds. The in-field and in-stream transport processes of OP pesticides are determined to a great extent by the dominant hydrologic processes of a river watershed. Therefore, a reliable hydrologic simulation has to be established for the dynamic pesticide exposure assessment. Modeling of pesticide fate and transport might be more complex and associated with more sources of uncertainty than hydrologic simulation (Dubus et al., 2003, Holvoet et al., 2005). Even if the rates and timing of a particular pesticide application are fully recorded for some agriculturally dominated areas such as the San Joaquin Valley, there are other data inputs associated with greater uncertainties, such as soil properties (e.g., curve number and erosion factors) and chemical properties of pesticides (e.g., half-lives and partition coefficient). Therefore, it is very important to present clearly the propagation of input variances into model outputs for environmental persistence of OP pesticides.
In most studies of water-quality modeling at watershed scale, sensitivity analysis is usually performed for one catchment as a whole, without the consideration for spatial arrangement of sub-catchments in the stream network. Therefore, the spatial effects on the model performance and management implications are not fully evaluated. As indicated by Arabi et al. (2006), for example, in-stream transport processes and associated conservation practices must be discussed at watershed scale because their effects cannot be detected at fields. In order to provide useful information for agricultural management strategies, simulation of pesticide transport should consider hydrometeorology and water-quality processes at various spatial scales. The spatial dependence of environmental fate of pesticide species were traditionally evaluated based on measurement data. For example, Capel et al. (2001) examined monitoring data of 39 pesticides as a function of scale across 14 orders of magnitude. There are a few but increasing number of studies modeling fate and transport of pesticides at area-varying watersheds (Brown et al., 2002, USEPA, 2006, Luo et al., 2008).
The Soil and Water Assessment Tool (SWAT) was chosen in this study to predict pesticide loads of OP species in surface water. In addition to hydrologic simulation, SWAT also allows dynamic predictions of pesticide outputs at various spatial scale (Gassman et al., 2007). In our previous study, SWAT had been calibrated for the hydrologic conditions in the San Joaquin River basin (14 983 km2), and applied to evaluate residue distribution of two OP pesticide diazinon and chlorpyrifos (Luo et al., 2008). Based on the calibrated model, effects of pesticide management practices on water quality were evaluated (Zhang et al., 2008). The modeling efforts were extended in this study by predicting pesticide transport and potential efficiency of structural best management practices (BMPs) at spatial scales of [1] small watershed (the Orestimba Creek watershed, a tributary watershed of the San Joaquin River, 563 km2), and [2] agricultural fields (agricultural drainage area in the Orestimba Creek watershed, 146 km2 in total). The objectives of this study were threefold: (1) to evaluate the modeling capability of SWAT in predicting the fate and transport of pesticides in agricultural watersheds with different spatial extents, by comparing with measured pesticide loads in surface water; (2) to identify the most influential model parameters for simulating pesticide distribution based on a management-oriented sensitivity analysis; and (3) to represent the functionality of selected management practices in SWAT, and to assess the water-quality impacts at both field and watershed scales. Results in this study were anticipated to provide useful information for agricultural BMP planning in reducing pesticide residues, and for future model development and evaluation in agrochemical transport and mitigation.
Section snippets
Site description
The Orestimba Creek watershed was selected as a representative sub-region in the San Joaquin River basin for further investigation (Fig. 1). Large amounts of organophosphate insecticides are sprayed to almonds and other stone-fruit orchards in the watershed (Cryer et al., 2001). Compared to other regions in the San Joaquin Valley, a greater variety of pesticides were detected in this watershed (Dubrovsky et al., 1998). Chlorpyrifos and diazinon loadings per unit area from the Orestimba Creek
Baseline modeling
The results of statistical evaluation of the model performance for stream flow, sediment, and pesticides predicted at the two USGS gauges during 1992–2007 are summarized in Table 4, reported for rainfall season, irrigation season, and the entire simulation period. The model efficiencies (NSE) by comparing the SWAT-predicted monthly stream flow and USGS measurements were 0.82 and 0.78 at gauges #11274500 and #11274538, respectively, for the entire simulation period (Fig. 2). This indicated good
Conclusion
In this paper, pesticide fate and transport in an agriculturally dominated watershed were evaluated by SWAT modeling. The model simulation was applied in the field conditions of the Orestimba Creek watershed, California, with two widely used organophosphate pesticides chlorpyrifos and diazinon during 1990–2007. The calibrated SWAT generated reliable simulation results for the stream flow, sediment, and pesticides in the studied watershed. By comparing with the results of SWAT model previously
References (35)
- et al.
Semidiscrete pesticide transport modeling and application
Journal of Hydrology
(2004) - et al.
Sources of uncertainty in pesticide fate modelling
The Science of the Total Environment
(2003) - et al.
Sensitivity analysis for hydrology and pesticide supply towards the river in SWAT
Physics and Chemistry of the Earth
(2005) - et al.
A multimedia environmental model of chemical distribution: fate, transport, and uncertainty analysis
Chemosphere
(2007) - et al.
Dynamic modeling of organophosphate pesticide load in surface water in the northern San Joaquin Valley watershed of California
Environment Pollution
(2008) - et al.
Representation of agricultural conservation practices with SWAT
Hydrologic Processes
(2007) - et al.
Role of watershed subdivision on modeling the effectiveness of best management practices with SWAT
Journal of the American Water Resources Association
(2006) - et al.
Modeling long-term water quality impact of strctural BMPs
Transaction of the American Society of Agricultural and Biological Engineers (ASABE)
(2006) - et al.
Evaluation of structural best management practices 20 years after installation
Journal of Soil and Water Conservation
(2004) - et al.
Prediction of pesticide concentrations found in rivers in the UK
Pest Management Science
(2002)
The behaviour of 39 pesticides in surface waters as a function of scale
Hydrologic Processes
California Interagency Watershed Map of 1999
California Irrigation Management Information System. The Office of Water Use Efficiency
California Land and Water Use: Survey Data Access
Rules for dormant season insecticides. California Environmental Protection Agency
San Joaquin River Watershed Surface Water Ambient Monitoring Program (SWAMP)
Pesticide Use Reporting (PUR). California Environmental Protection Agency
Cited by (60)
Performance of a pesticide fate model for predicting multi-year surface runoff contamination in a Mediterranean vineyard
2024, Science of the Total EnvironmentAssessment of agricultural pesticide inert ingredient transport following modeling approach: Case study of two formulation agents in Sacramento River watershed
2023, Journal of Environmental ManagementCitation Excerpt :By adding pesticide parameters in the sensitivity analysis, the most sensitive parameters were revealed and the estimated direct losses of pesticide could be corrected (Holvoet et al., 2005). Luo and Zhang (2009) have reported that the sensitivity analysis results in their study provided more meaningful information for the evaluation of water quality and management practices. Meanwhile, previous studies have indicated the uncertainty sources (e.g., model parameterization, rates and timing of pesticide application) regarding pesticide modeling by the SWAT model (Boithias et al., 2014; Chen et al., 2017a; Gassmann et al., 2014b).
Environmental prioritization of pesticide in the Upper Citarum River Basin, Indonesia, using predicted and measured concentrations
2020, Science of the Total EnvironmentEmission estimation and fate modelling of three typical pesticides in Dongjiang River basin, China
2020, Environmental PollutionCitation Excerpt :The result of output flux was used for sensitivity analysis, as shown in Table S8 (b) and Fig. S4. The most sensitive parameters for pesticide fate were obtained by referring to publications and from Oregon State University Extension Pesticide Properties Database (Luo and Zhang, 2009; Ficklin et al., 2013; Boithias et al., 2014; Fohrer et al., 2014; Chen et al., 2017). Measured concentrations for the three pesticides in water and suspended particulate matters (SPMs) of the Dongjiang River basin were obtained through field campaign in December 2008 and July 2009 to verify the model, which included sampling in the dry and wet seasons.
A review of pesticide fate and transport simulation at watershed level using SWAT: Current status and research concerns
2019, Science of the Total Environment