Comparing and evaluating pesticide leaching models: Results of simulations with PELMO

Abstract The PELMO model was used independently by five modellers to reproduce the results of a lysimeter study performed at Tor Mancina in Italy and a field study performed at Vredepeel in the Netherlands. For the comparisons of the Tor Mancina data set the main features of the measured fluxes of water and bromide were well reproduced by the simulations. The deviations between simulated and experimental cumulative amounts of water leached were generally less than 50%. The measured leaching of metolachlor was small (typical concentrations considerably below 0.1 μg/l). These trace amounts were not reproduced by any of the simulations, not even by those calibrated for bromide leaching in the re-packed lysimeters. For Vredepeel, the agreement between the measured and simulated water tables were generally poor, even on a qualitative level. This was mainly due to PELMOs inability to deal with shallow, fluctuating groundwater tables. Concentrations of both the tracer and the pesticides were generally satisfactorily reproduced in the initial phases of the experiment but not at later stages. In most cases, the penetration depth of the centre of mass was over-estimated by the model and the dispersion of the pesticide under-estimated. The correct determination of the parameters to simulate the degradation (and adsorption) of pesticide in the field seemed to be of much greater importance for accurately modelling the transport of such chemicals in soils than improvements in the water balance. The degradation data from long-term laboratory studies clearly did not reflect field conditions. Additional sampling dates to determine more concentration profiles and to measure DT50 values from the field would have helped reducing the differences in picking different input data by the modellers and would have improved the accuracy of the model predictions. Validation tests, user guidance and good modelling practice are recommended as essential tools to improve the confidence of the scientific community in modelling results.