Monitoring Herbicide Concentrations and Loads during a Flood Event: A Comparison of Grab Sampling with Passive Sampling Andrew Joseph Novic,* ,† Dominique S. O’Brien, ‡ Sarit L. Kaserzon, † Darryl W. Hawker, § Stephen E. Lewis, ‡ and Jochen F. Mueller † † Queensland Alliance for Environmental Health Sciences, The University of Queensland, 39 Kessels Road, Coopers Plains, Queensland 4108, Australia ‡ Catchment to Reef Research Group, TropWATER, ATSIP, DB145, James Cook University, Townsville, Queensland 4811, Australia § Griffith School of Environment, Griffith University, 170 Kessels Road, Nathan, Queensland 4111, Australia * S Supporting Information ABSTRACT: The suitability of passive samplers (Chemcatcher) as an alternative to grab sampling in estimating time-weighted average (TWA) concentrations and total loads of herbicides was assessed. Grab sampling complemented deployments of passive samplers in a tropical waterway in Queensland, Australia, before, during and after a flood event. Good agreement was observed between the two sampling modes in estimating TWA concentrations that was independent of herbicide concentrations ranging over 2 orders of magnitude. In a flood-specific deployment, passive sampler TWA concentrations underestimated mean grab sampler (n = 258) derived concentrations of atrazine, diuron, ametryn, and metolachlor by an average factor of 1.29. No clear trends were evident in the ratios of load estimates from passive samplers relative to grab samples that ranged between 0.3 and 1.8 for these analytes because of the limitations of using TWA concentrations to derive flow-weighted loads. Stratification of deployments by flow however generally resulted in noticeable improvements in passive sampler load estimates. By considering the magnitude of the uncertainty (interquartile range and the root-mean-squared error) of load estimates a modeling exercise showed that passive samplers were a viable alternative to grab sampling since between 3 and 17 grab samples were needed before grab sampling results had less uncertainty. ■ INTRODUCTION Comprehensive monitoring networks are often established to characterize and evaluate water quality because of the risks that offsite transport of micropollutants (including herbicides) pose to receiving ecosystems. 1 In such networks, the estimation of average concentration and total loads of micropollutants over specific periods of time are tools that help define and evaluate progress toward water quality targets. In micropollutant monitoring, uncertainty and error in the estimation of concentration or loads can arise from discharge measurement, sampling, storage/preservation of samples and analytical results. Of these, sample collection has often been shown to be the most important for a range of matrices. Not only can it be the greatest source of uncertainty, but the variation of this uncertainty can also be the greatest. 2,3 To date, grab (or spot) sampling remains the predominant surface water-sampling mode for evaluating concentration and loads of micropollutants despite its well-documented limita- tions. For example, a common criticism of grab sampling for measuring concentration is that it only provides a snapshot of concentration in time. 4 In the absence of continuous monitoring, varying concentrations and flow rates of rivers and streams over time-with flood events being scattered within low flow periods for example-represent a form of measure- ment and stochastic uncertainty that may preclude the obtainment of reliable, representative data from grab samples. Representative samples are a requirement in micropollutant monitoring. 5,6 The relationship between flow rate and chemical concen- tration is complex and depends upon the characteristics of individual catchments (e.g., scale) as well as the nature of the rainfall event(s) (e.g., intensity) and micropollutants (e.g., physicochemical properties). 7 Periods of maximum flow do not necessarily coincide with times of maximum chemical concentration that can also vary temporally. Thus, the chemograph (a temporal plot of concentration) of each target chemical responds differently to a flow event than a hydrograph (a temporal plot of flow rate) and as a result, the representativeness of sampled data is also different for each Received: June 7, 2016 Revised: February 6, 2017 Accepted: February 14, 2017 Published: February 14, 2017 Article pubs.acs.org/est © 2017 American Chemical Society 3880 DOI: 10.1021/acs.est.6b02858 Environ. Sci. Technol. 2017, 51, 3880-3891