Modelling of pesticide exposure in ground and surface waters used for public water supply

Abstract

Diffuse transfers of pesticides from agricultural land to ground and surface waters can lead to significant drinking water quality issues. This thesis describes the development and application of a parameter-efficient, numerical model to predict pesticide concentrations in raw water sources within an integrated hydrological framework. As such, it fills an unoccupied niche that exists in pesticide fate modelling for a computationally undemanding model that contains enough process complexity to be applicable in a wide range of catchments and hydrogeological settings in the UK and beyond. The model represents the key processes involved in pesticide fate (linear sorption and first-order degradation) and transport (surface runoff, lateral throughflow, drain flow, percolation to the unsaturated zone, calculated using a soil water balance) in the soil at a daily time step. Soil properties are derived from the national soil database for England and Wales and are used to define the boundary conditions at the interface between the subsoil and the unsaturated zone. This is the basis of the integrated hydrological framework which enables the application of the model to both surface water catchments and groundwater resources. The unsaturated zone model accounts for solute transport through two flow domains (accounting for fracture flow and intergranular matrix flow) in three hydrogeological settings (considering the presence and permeability of superficial deposits). The model was first applied to a small headwater sub-catchment in the upper Cherwell. Performance was good for drainflow predictions (Nash Sutcliffe Efficiency > 0.61) and performed better than the MACRO model and as well as the modified MACRO model. Surface water model performance was evaluated for eight pesticides in five different catchments. Performance was generally good for flow prediction (Nash Sutcliffe Efficiency > 0.59 and percentage bias below 10 %, in the validation period for all but two catchments). The 90th percentile measured concentration was captured by the model in 62 % of catchment-pesticide combinations. In theremaining cases predictions were within, at most, a factor of four of measured 90th percentile concentrations. The rank order of the frequency of pesticides detected over 0.1 μg L-1 was also predicted reasonably well (Spearman’s rank coefficient > 0.75; p < 0.05 in three catchments). Pesticide transport in the unsaturated zone model was explored at the point scale in three aquifers (chalk, limestone and sandstone). The results demonstrate that representing the unsaturated zone processes can have a major effect on the timing and magnitude of pesticide transfers to the water table. In comparison with the other catchment scale pesticide fate models that predict pesticide exposure at a daily time-step, the model developed stands out requiring only a small number of parameters for calibration and quick simulation times. The benefit of this is that the model can be used to predict pesticide exposure in multiple surface and groundwater resources relatively quickly which makes it a useful tool for water company risk assessment. The broad-scale approach to pesticide fate and transport modelling presented here can help to identify and prioritise pesticide monitoring strategies, to compare catchments in order to target catchment management and to highlight potential problems that could arise under different future scenarios.