Temporal variability in water quality of agricultural tailwaters: Implications for water quality monitoring
Introduction
The 2002 National Water Quality Inventory of the Environmental Protection Agency (EPA) identified agricultural non-point source (NPS) pollution as the leading cause of water quality impairment to rivers and lakes in the U.S. (U.S. - EPA, 2002). On irrigated lands, much of the NPS pollution is delivered to surface waters from tailwaters originating from gravity flow (flood or furrow) irrigation methods. Currently about 94,000 km2, 44% of the irrigated land area in the U.S., utilizes gravity flow irrigation (USDA-ASS, 2002). Due to its diffuse nature, agricultural return flows have remained largely unregulated. In 2003, California began the process of regulating agricultural water dischargers through adoption of the Irrigated Lands Conditional Waiver's Program (ILCWP). The ILCWP mandates that individual landowners or coalition groups develop monitoring programs to document that their contribution of NPS pollutants will not negatively impact surface waters. The current version of the ILCWP requires collection and analysis of one grab sample per month throughout the year (California Regional Water Quality Control Board, Central Valley Region, 2008). As regulation of agricultural NPS pollution becomes a more widespread reality, it is essential that regulators and growers have appropriate information for developing monitoring programs that: (i) document existing background conditions, (ii) identify exceedences of water quality constituents, (iii) verify the effectiveness of mitigation strategies, and (iv) are economically feasible to implement.
The aim of regulatory monitoring is to assess compliance with water quality objectives, usually concentrations or mass loading rates, for a given water body. However, most water quality monitoring programs have been designed on an arbitrary, rather than a statistically defensible basis (Strobl and Robillard, 2008). Proper evaluation of compliance is crucial, as growers may be unfairly punished for “false exceedances”, and the efficacy of the program is compromised if exceedances go undetected. The frequency of sampling necessary to accurately characterize water quality is dependent on the statistical distribution of the monitoring data (e.g., seasonal peaks, distribution, variance, and degree of autocorrelation) (Valiela and Whitfield, 1989). In many cases, cost constraints severely limit sampling frequency, despite the fact that accuracy and precision are a direct function of sampling frequency (Moustafa and Havens, 2001). A rational criterion for selecting sampling frequencies for regulatory monitoring is to choose a large enough sample number, based on statistical parameters (e.g., variance), to achieve a reasonably small and uniform confidence interval about the mean value (Loftis and Ward, 1980a).
Water quality monitoring involves sampling a “population” that is changing over time. In irrigation tailwater systems, sample statistics (e.g., sample mean) computed from water quality data are affected by: (i) random changes induced by irrigation timing and amount, fertilizer application, contributions from specific fields, etc., (ii) seasonal changes resulting from crop rotation, fertilizer application, plant nutrient demands, etc., and (iii) serial correlation of data (Loftis and Ward, 1980b). Given the diversity of cropping systems, nutrient management, irrigation practices, soils, and watershed size, it is difficult to recommend a universally acceptable monitoring program that transcends all water quality constituents of concern (e.g., nutrients, sediments, salts, and pesticides). There is currently a paucity of data quantifying the variability of water quality contaminants in agricultural tailwaters. Thus, it is generally not possible to evaluate the effectiveness of monitoring programs.
A large portion of the cost of monitoring is related directly to the collection and processing of water samples, so it is important to devise a monitoring scheme that minimizes sample number while preserving accuracy. Methods such as time-composite samples can be used to capture variability without increasing sample number (Moustafa and Havens, 2001). Continuous monitoring for certain constituents, such as salt (specific conductance) and sediment (turbidity) is feasible using microprocessor controlled sensors. However, the technology to quantify many pollutants (e.g., pesticides, nutrients) at a high frequency is either not available or prohibitively expensive. However, in some cases it may be possible to determine an easily measured proxy that displays a strong correlation to more difficult to measure constituents (e.g., turbidity versus total phosphorus, specific conductivity versus nitrate).
The goal of this study was to provide a statistical basis to quantify the variability of selected water quality constituents in five agricultural watersheds in the San Joaquin Valley, California, for the purpose of optimizing monitoring protocols both in terms of cost and accuracy. The number of samples necessary to calculate seasonal mean concentrations within given confidence bounds was evaluated for various sampling strategies (composite versus grab sampling). This is the first study of its type for California irrigation tailwaters, and therefore, it provides information critical for evaluation of the current ILCWP monitoring requirements, as well as a template for regulators and growers to design economically feasible and effective monitoring programs.
Section snippets
Study sites
Two sets of agricultural watersheds in California's San Joaquin Valley (SJV) were studied at two different temporal monitoring resolutions. Three small watersheds (<5000 ha) were monitored by the University of California, Davis (UC Davis) at a high frequency (6 h to daily), and two large watersheds (86 and 1245 km2) were monitored by the U.S. Geological Survey and UC Davis (USGS-UC Davis) at a low frequency (biweekly to monthly) over a decade. The high-resolution data were used to evaluate
Temporal variability
Temporal variation in water quality was considerably different among sites. Time-series plots of TN (Fig. 2) are shown to illustrate the magnitude of variability within a given site, and the heterogeneity of variability among sites. TN was highly variable at W-1 throughout both the 2006 and 2007 irrigation seasons, and showed the greatest range of TN values, <0.01–60 mg L−1 in 2006 and <0.01–40 mg L−1 in 2007. W-1 TN values fluctuated on short-time scales, often more than 20 mg L−1 in a 24-h period (
Conclusions
The dynamic nature of water quality variability in agricultural tailwaters in the San Joaquin Valley poses unique challenges for regulators to design and implement effective monitoring programs that are economically feasible. The current ILCWP takes a first step towards monitoring agricultural tailwaters. Results from this study show that the mandated monthly sampling frequency was grossly inadequate in addressing the short-term water quality variability inherent in tailwater systems with
Acknowledgements
Funding for this project has been provided in part through an agreement with the California State Water Resources Control Board. The contents of this document do not necessarily reflect the views and policies of the State Water Resources Control Board, nor does mention of trade names or commercial products constitute endorsement or recommendation for use. We gratefully acknowledge the assistance of Jiayou Deng, Tony Orozco, Jon Maynard, and Jeannie Evatt.
References (28)
- et al.
Using turbidity and acoustic backscatter intensity as surrogate measures of suspended sediment concentration in a small subtropical estuary
J. Environ. Manage.
(2008) - et al.
A comparison of river water quality sampling methodologies under highly variable load conditions
Chemosphere
(2007) - et al.
Characterization and prediction of highway runoff constituent event mean concentration
J. Environ. Manage.
(2007) - et al.
Network design for water quality monitoring of surface freshwaters: a review
J. Environ. Manage.
(2008) - et al.
Evaluation of sampling frequency for the water quality monitoring network of coastal Galicia (northwest Spain)
Water Environ. Res.
(1998) - California RWQCB (Regional Water Quality Control Board), Central Valley Region, 2008. Monitoring and Reporting Program...
- Clesceri, L.S., Greenberg, A.E., Eaton, A.D., 1998. Standard Methods for the Examination of Water and Wastewater. 20th...
- et al.
Spectrophotometric determination of nitrate with a single reagent
Anal. Lett.
(2003) - et al.
The fine structure of water-quality dynamics: the (high-frequency) wave of the future
Hydrol. Process.
(2004) - Kratzer, C.R., Dileanis, P.D., Zamora, C., Silva, S.R., Kendall, C., Bergamaschi, B.A., Dahlgren, R.A., 2004. Sources...
Cost-effective water quality assessment through the integration of monitoring data and modeling results
Water Resour. Res.
Sampling frequency selection for regulatory water quality monitoring
Water Resour. Bull.
Water quality monitoring—some practical sampling frequency considerations
Environ. Manage.
Capturing temporal variability for estimates of annual hydrochemical export from a first-order agricultural catchment in southern Ontario, Canada
Hydrol. Process.
Cited by (27)
HUM: A review of hydrochemical analysis using ultraviolet-visible absorption spectroscopy and machine learning
2023, Science of the Total EnvironmentWhat do we need for water and soil quality monitoring?
2022, UV-Visible Spectrophotometry of Waters and SoilsEffects of sampling strategies and estimation algorithms on total nitrogen load determination in a small agricultural headwater watershed
2019, Journal of HydrologyCitation Excerpt :In recent decades, numerous monitoring projects have been conducted to assess water quality degradation in riverine systems (Hudnell, 2010; Richards et al., 2008; Vörösmarty and Meybeck, 2004; Walling and Webb, 1996). Such monitoring programs have been conducted to: (i) assess current water quality conditions (Roygard et al., 2012); (ii) distinguish between point and non-point pollutant sources (Roygard et al., 2012); (iii) provide data for watershed modeling (Gassman et al., 2007; Ullrich and Volk, 2010); (iv) report field boundary conditions and monitoring time to evaluate long-term trends in river loads (Brauer et al., 2009; King and Harmel, 2003); and (v) evaluate the effectiveness of best management practices to guide policy and management decisions (Brauer et al., 2009; King and Harmel, 2003; Lennartz et al., 2010; Snelder et al., 2017). Ideally, the uncertainty of monitoring data (in terms of accuracy and precision) would be minimized by using appropriate collection and analysis methods, and estimates of the uncertainty would be reported alongside the data to guide data users (Birgand et al., 2010; Harmel et al., 2009; Moatar et al., 2013).