Herbicide Sampling in Watersheds: When Enough is Enough

Abstract

Herbicide losses from agriculture represent potential environmental and human health hazards, and are the target of various conservation practices. Measuring herbicides at the watershed scale can be rigorous and expensive, and sampling frequency should be based upon the monitoring purpose and expected associated uncertainties. Atrazine, simazine, alachlor, metolachlor, and glyphosate were monitored in tile-fed drainage ditches for conservation effects assessment and source water protection. A total of 20,428 water samples were collected during the 2004 to 2009 cropping seasons at eight monitoring sites located at the outlets of sub basins ranging in size from 298 to 19,341 ha. However, it may be possible to analyze fewer samples without incurring unacceptable errors. To identify other possible methods to reduce analytical cost for determination of drinking water source suitability, we further analyzed the potential errors associated with compositing every 2 daily samples into a single sample representing 2 days by electronically simulating composites, while sampling more frequently during storm flow events. Compositing samples in this matter was found to introduce infrequent and minimal error in water quality compliance for atrazine and simazine only, and represent an average reduction in analytical cost of 21%. Since herbicide transport is closely associated with the first few rainfall events following application, targeting sampling when levels are temporally most variable may be another efficient way to assess conservation efforts and ensure adequate water filtration methods are employed at the right time. The abilities of truncated the sampling seasons (105, 122, and 226 days) to predict water quality compliance on an annual basis were evaluated. Results indicate that the ideal sampling season of atrazine and simazine is April 1 – November 15, while sampling of other measured herbicides can be truncated by August 15 with infrequent and minimal errors.