When field pollution is heterogeneous due to localized pesticide application, as is the case of chlordecone (CLD), the mean level of pollution is difficult to assess. Our objective was to design a decision support tool to optimize soil sampling. We analyzed the CLD heterogeneity of soil content at 0-30- and 30-60-cm depth. This was done within and between nine plots (0.4 to 1.8 ha) on andosol and ferralsol. We determined that 20 pooled subsamples per plot were a satisfactory compromise with respect to both cost and accuracy. Globally, CLD content was greater for andosols and the upper soil horizon (0-30 cm). Soil organic carbon cannot account for CLD intra-field variability. Cropping systems and tillage practices influence the CLD content and distribution; that is CLD pollution was higher under intensive banana cropping systems and, while upper soil horizon was more polluted than the lower one with shallow tillage (<40 cm), deeper tillage led to a homogenization and a dilution of the pollution in the soil profile. The decision tool we proposed compiles and organizes these results to better assess CLD soil pollution in terms of sampling depth, distance, and unit at field scale. It accounts for sampling objectives, farming practices (cropping system, tillage), type of soil, and topographical characteristics (slope) to design a relevant sampling plan. This decision support tool is also adaptable to other types of heterogeneous agricultural pollution at field level. (Résumé d'auteur)