Simulation of pesticide concentration in a small agricultural catchment under different climate change scenarios.
Abstract
The present work looked at the potential to simulate trends in pesticide concentrations in a small French agricultural catchment. A comprehensive modelling approach taking into account water flow and pesticide transfers in the soil and subsoil aquifer of Brévilles catchment was adopted. Results from an 1D soil leaching model (MACRO) were fed into a 2D groundwater model (MARTHE). The Brévilles catchment was divided in a series of homogeneous zones to represent the catchment area. The outputs of interest from the MACRO simulations included: i) daily predictions for percolation at the bottom of the leaching columns; and, ii) daily concentrations of atrazine in leachate. The input files of the 2D MARTHE model were adapted to receive the output files of the MACRO simulations. The coupled transport model of Brévilles basin was calibrated over the period 1988-2005 with a daily time step while a simulation over the period 1972-1987 was carried out for model initialisation. To predict the evolution of atrazine concentration in Brévilles spring, simulated meteorological data corresponding to climate change were generated using 12 Regional climate models from the PRUDENCE project (Christensen et al., 2002). Ten replications of each scenario were generated.
On the fitting period (2001-2006), the piezometric levels were well reproduced, but the discharge at the spring was overestimated. All the peaks at the spring were not reproduced by the combined model, the simulated discharge curve was smoother than the real one. The general trend was however simulated. The simulated maximal average discharge over the period 2006-2019 ranged from 18.7 to 27.5 l/s. The coefficient of variation (CV) was c.a. 3.5% for all climate change scenarios except for one which had a CV of 6.2%. The predicted concentrations of atrazine at the spring were overestimated in 2005 and underestimated in 2007. The general trend was however adequately described. The minimal concentration predicted in 2019 was 0.016 µg/l whereas the maximum was 0.025 µg/l.
This work showed that over the 10 years period studied, the variability induced by the use of different climate change models was found to be equivalent to that originating from the use of replicated datasets for each scenario.