A modelling framework to simulate river flow and pesticide loss via preferential flow at the catchment scale

• The catchment system can be simulated as the sum of multiple field-scale processes.
• Pesticide concentrations in stream flow were driven by field-scale processes.
• In-stream processes had little effect on simulations.
• Uncertainties in rain gauge recording affected the simulation of low-flow periods.
• SPIDER simulates important lateral flow losses that can occur when drains are not flowing.

A modelling framework with field-scale models including the preferential flow model MACRO was developed to simulate transport of six contrasting herbicides in a 650 km2 catchment in eastern England. The catchment scale model SPIDER was also used for comparison. The catchment system was successfully simulated as the sum of multiple field-scale processes with little impact of in-stream processes on simulations. Preferential flow was predicted to be the main driver of pesticide transport in the catchment. A satisfactory simulation of the flow was achieved (Nash-Sutcliffe model efficiencies of 0.56 and 0.34 for MACRO and SPIDER, respectively) but differences between pesticide simulations were observed due to uncertainties in pesticide properties and application details. Uncertainty analyses were carried out to assess input parameters reported as sensitive including pesticide sorption, degradation and application dates; their impact on simulations was chemical-specific. The simulation of pesticide concentrations in the river during low flow periods was very sensitive to uncertainty from rain gauge measurements and the estimation of evapotranspiration.