One current challenge in sustainable agriculture is to redesign cropping systems to reduce the use and impacts of pesticides, and by doing so protect the environment, in particular groundwater, and human health. As a large range of systems could be explored and a wide number of pesticides used, field experiments cannot be carried out to study the sustainability of each of them. Thus, the objectives of this work were (1) to measure water flows and pesticide leaching in six contrasted low input cropping systems based on sunflower-wheat rotation, oilseed rape-wheat-barley rotation, and maize monoculture, experimented for three years in three different soil and climatic conditions, and (2) to assess and to compare the ability of three pesticide fate models (MACRO, PEARL, PRZM) to simulate the observed water flows and pesticide concentrations. The systems were designed using various crop rotations, including cover crops and intercrops. The models were parameterized with generic parameter estimation routines as done for regulatory risk assessment, and a method was developed to parameterize intercrops, not represented in the models: the use of average crop factors, maximum LAI, crop height and rooting depth of the crops constituting the intercrop allowed acceptable simulations of cumulative water flows, but not their dynamic. Twelve pesticides of 70 applied were quantified in lysimeter samples (e.g. bentazone, glyphosate, S-metolachlor), and their concentrations exceeded 0.1 μg L^-1 in several occasions. The performance of the models to reproduce pesticide concentrations was generally poor illustrating the great challenge and the progress needed to simulate accurately pesticide transfers into the soil. The best fits to measured data were attained using "worst-case" pesticide sorption and degradation parameters. Overall, MACRO performed better than PEARL and PRZM. The method developed to parameterize intercrops could be used for risk assessment of groundwater contamination by pesticides in low input cropping systems, but the use of the three models without any calibration is likely to underestimate pesticide leaching in several situations.