Evaluating the potential of constructed wetlands to mitigate pesticide runoff from agricultural land to surface waters.

Abstract

Pesticides make important contributions to modern agriculture. However, diffuse transfers of pesticide from agricultural land to surface water can lead to significant compliance failures in drinking water supplies under the EU Drinking Water Directive. Article 7 of the Water Framework Directive promotes a prevention-led approach in drinking water supply catchments which is based on a range of catchment management solutions to mitigate diffuse pollution. This includes non-structural interventions, such as modifying crop rotations and changing the timing and amount of pesticide applied and structural measures, such designating buffer and no-spray zones and the construction of attenuation ponds and wetlands. Although the performance of constructed wetlands for diffuse-source nutrients is now well understood, there is currently little in the literature on the behaviour of pesticides in these systems – particularly under transient conditions typical of storm events. One of the simplest and cheapest constructed wetland designs has a free-water-surface. Understanding the performance of these systems is particularly important because they are most likely to be implemented. The primary aim of this project was, therefore, to improve understanding of the processes affecting pesticide fate and transport in free-surface constructed wetlands in order to assess their utility as attenuation features. The study focused on six pesticides: metaldehyde, quinmerac, carbetamide, metazachlor, propyzamide and pendimethalin. Metaldehyde has been reported as the biggest pesticide challenge currently facing the UK water industry. Two free-surface constructed wetland systems (the South and North Wetlands) situated at Hope Farm in Knapwell (Cambridgeshire, UK) were monitored over two winter field seasons. Discharge was measured at control structures (v-notch weirs and Venturi flumes) installed at the inflow and outflow of each feature (using a combination of pressure transducers and ultrasonic sensors) and pesticide concentrations were determined in water samples collected typically every eight hours using automotic water samplers. Concentrations were measured using direct injection liquid chromatography coupled with tandem quadrupole mass spectrometry (LC-MS-MS). A multi-component method was developed to allow high rates of sample throughput with minimal preparation. The LOQs obtained ranged from 0.2 to 1.0 µg L⁻¹, which is acceptable for detecting concentrations in natural water samples in this project. After autumn wet-up, discharge response to rainfall was flashy in both wetland catchments. Pesticide concentrations typically increased rapidly in the first significant post-application storm event and then decreased during hydrograph recession. Concentrations measured in the inflow and outflow of both wetland systems were often very similar, suggesting little attenuation for the pesticides monitored – particularly during storm events when both concentrations and loads increased but when retention times decreased. The main explanation for poor performance was the very short hydraulic residence time of these systems (determined using pulse-injection dye tracing exercises employing rhodamine WT). The solute residence time estimated during hydrographs was typically ~32 minutes, giving little time for pesticide sorption to sediment or vegetation, degradation or plant uptake. In the North Wetland, which is bunded at the outflow, discharge is intermittent in the autumn, immediately following post-harvest pesticide applications. This means that static periods exist with no inlet and outlet flows, during which some pesticide losses were observed (approximately 12% for metaldehyde and 20% for metazachlor). This suggests that these systems may be of some value in reducing fluxes and concentration for limited periods. During the subsequent winter period, however, when water levels in the wetland over-topped the outflow pipe, residence times were short and fluxes in the outflow were similar to those in the inflow (i.e. negligible removal was observed). A set of degradation and sorption experiments were conducted in the laboratory to investigate potential mechanisms of pesticide attenuation. The sorption experiments were based on the OECD test 106 guidelines. Results showed a linear isotherm for sorption and desorption for all individual pesticides (R² > 0.97). They suggested that the apparent affinity of the pesticides evaluated for organic matter in the Hope Farm wetlands was significantly lower than that reported previously in soils for metaldehyde and carbetamide, was similar for quinmerac and propyzamide and slightly higher for metazachlor. Sorption experiments conducted using a mixture of pesticides in wetland sediment showed that the presence of other compounds can influence the sorption capacity of the sediment. The degradation studies looked at water-sediment systems, following the OECD 308 test guidelines. A short “pseudo lag phase” of approximately seven days was apparent in many of the experiments, in which the rate of concentration change was lower than the rate observed thereafter. This is a commonly reported phenomenon in laboratory degradation studies which can be caused by slow adaptation of degrading microorganisms. Interestingly, this phenomenon does not appear to have been apparent in the North Wetland under static conditions. This may be due to the development of a competent microbial community in this ,system as a consequence of repeated exposure. Rate constants were explicitly hypothesised to be inversely proportional to water depth because the majority of the degrading microbes are assumed to inhabit the sediment in fixed biofilms (i.e. freely suspended cells and cells associated with suspended solids are relatively unimportant as degraders). The water volume to sediment surface area (i.e. the water depth) is, therefore, expected to control the concentration change in the water column. In the outdoor mesocosms, fitted rate constants were higher in treatments with 20cm water depth than when water depth was 40cm, suggesting that the depth control hypothesis may have some validity. When the depth was 10cm, factors such as sediment-water exchanges (owing to the particularly low water volume to sediment mass ratio) may have confounded the depth effect. Overall, the field and laboratory data reported here suggest that small free-surface wetland features may be relatively ineffective at reducing pesticide concentrations and loads, unless the catchment size is small relative to the wetland dimensions. This means that a large number of such features would need to be constructed to make an appreciable difference at the catchment scale.