Persistent organic pollutants like organochlorine pesticides continue to contaminate large areas worldwide raising questions concerning their management. We designed and tested a method to link soil and water pollution in the watershed of the Galion River in Martinique. We first estimated the risk of soil contamination by chlordecone by referring to past use of land for banana cultivation and took 27 soil samples. We then sampled surface waters at 39 points and groundwater at 16 points. We tested three hypotheses linked to the source of chlordecone pollution at the watershed scale: (i) soils close to the river, (ii) soils close to the sampling point, (iii) throughout the sub-watershed generated at the sampling point. Graphical and statistical analysis showed that contamination of the river increased when it passed through an area with contaminated plots and decreased when it passed through area not contaminated by chlordecone. Modeling showed that the entire surface area of the watershed contributed to river pollution, suggesting that the river was mainly being contaminated by the aquifers and groundwater flows. Our method proved to be a reliable way to identify areas polluted by chlordecone at the watershed scale and should help stakeholders focus their management actions on both hot spots and the whole watershed. (Résumé d'auteur)

Persistent organic pollutants like organochlorine pesticides continue to contaminate large areas worldwide raising questions concerning their management. We designed and tested a method to link soil and water pollution in the watershed of the Galion River in Martinique. We first estimated the risk of soil contamination by chlordecone by referring to past use of land for banana cultivation and took 27 soil samples. We then sampled surface waters at 39 points and groundwater at 16 points. We tested three hypotheses linked to the source of chlordecone pollution at the watershed scale: (i) soils close to the river, (ii) soils close to the sampling point, (iii) throughout the sub-watershed generated at the sampling point. Graphical and statistical analysis showed that contamination of the river increased when it passed through an area with contaminated plots and decreased when it passed through area not contaminated by chlordecone. Modeling showed that the entire surface area of the watershed contributed to river pollution, suggesting that the river was mainly being contaminated by the aquifers and groundwater flows. Our method proved to be a reliable way to identify areas polluted by chlordecone at the watershed scale and should help stakeholders focus their management actions on both hot spots and the whole watershed.