In the French West Indies, chlordecone (CLD), an organochlorine pesticide, was applied in banana plantations from 1972 to 1993. Our study was aimed at identifying the factors that determine soil CLD contamination and that thus can describe the spatial distribution of this contamination. For each of the 1045 plots analyzed for CLD, we collected spatial information on the physical environment and agricultural activities. Then, we searched for determinants of soil contamination and used a multiple regression model to predict the spatial distribution of CLD. First, the results showed that the soil type had a strong impact. Andosols, with a high organic carbon content (up to 100 g kg−1) and a high sorption capacity (Koc 20 000 L kg−1), had the highest CLD concentrations and stocks. Nitisols had 10-fold lower sorption capacities. Second, the results showed a "farm effect" on soil contamination, due to the more or less intensive application of inputs by farmers and to the shorter or longer application times, depending on the banana land-use frequency. However, despite these effects and given the weak accuracy of the usually available information on agricultural activities, it remained difficult to predict the spatial distribution of CLD contamination by a multiple regression model. In conclusion, this study highlighted the multifactorial nature of pollution. The main perspectives are related to better assessments of agricultural activities using remote sensing strategies and designing agricultural practices observatories focusing on pesticide use.