Changes in agricultural land use over recent decades have exacerbated the pressures and impacts of agriculture on the environment. Researchers model these pressures and impacts using data on soils, climate and agricultural land use characteristics, which have to be represented at the spatial and temporal scales appropriate to the ecological or biophysical processes being studied. Representation of agricultural land use across large areas (several thousand of km2 or more) is hampered by lack of data on the diversity of farming practices and their spatial distribution, usually reduced to land cover. We propose in this article a methodological approach for combining generic and local information sources to characterise, across a large area, the diversity of crop managements and their spatial distribution, at a resolution appropriate to a specific environmental issue: assessment of nitrate contamination of groundwater. The approach involves four steps: (i) devise a system of spatial units that is appropriate to the research context and the available information sources, (ii) use multivariate analysis to characterise the diversity of crop managements obtained from a national database, (iii) conduct a survey, holding semi-directive interviews with experts and local actors to identify specific features that may locally modify this diversity, (iv) synthesise the statistical findings and survey results to establish a set of representative crop management "types". Illustrated in connection with the problem of diffuse nitrogen pollution of groundwater in the Seine-Normandie basin in France, it enabled to describe 919 crop managements out of the 18 main crops in the basin and especially highlighted the high diversity of practices on wheat with more than 200 crop managements. The knowledge produced on the diversity of crop managements and their spatial distribution was used for environmental modelling.