Distributed modelling of the fate of pesticides in groundwater, approaches for calibration and validation

Groundwater and raw water pollution risks of pesticides were assessed by analyses of groundwater samples and by modelling groundwater flow and chemical transport with a coupled model ‘train’ consisting of: a regional Modflow model for simulation of groundwater flow in the saturated zone (IBRAHYM, TNO 2007)) a 1-D model for simulation of water and pesticide transport in the unsaturated zone (PEARL, Leistra, 2001) a pathline-based chemical transport model for the saturated zone (RESPOND, Vink, 2008) The use of pesticides over the years was reconstructed with a series of detailed land-use maps specifying agricultural use over the period 1985-2009 in groundwater protection zones in Limburg Province, in the southern part of the Netherlands. Thus, reconstructed historic pesticide loads were used as input to unsaturated and saturated zone models. Relevant chemical properties such as CEC, REDOX conditions and organic matter concentrations in both the soil and the groundwater were mapped into a 3-D interpolated raster so that the chemical environment of every pathline could be characterized. Various checks and validations have been applied as to validate modelling results throughout the process: comparison of PEARL results with experimental data from shallow monitoring wells simulation of diffuse ‘guide’ parameters nitrate, sulphate and total hardness in pumping wells as to validate the flow patterns and travel times that were calculated with the groundwater model (IBRAHYM) comparison of observed and calculated pesticide concentrations in deep monitoring wells and pumping wells Calibration of the chemical transport simulations of both the diffuse guide parameters as the pesticides was carried out with a genetic algorithm that enables multiple objective calibration and permits to assess discrepancies between groundwater model and chemical transport model results. The pathline approach enabled relatively modest calculation times and thus also calibration, in spite of the relatively large level of detail of this regional modelling session.