Long-term pollution by chlordecone of tropical volcanic soils in the French West Indies: A simple leaching model accounts for current residue

https://doi.org/10.1016/j.envpol.2008.12.015Get rights and content

Abstract

Chlordecone was applied between 1972 and 1993 in banana fields of the French West Indies. This resulted in long-term pollution of soils and contamination of waters, aquatic biota, and crops. To assess pollution level and duration according to soil type, WISORCH, a leaching model based on first-order desorption kinetics, was developed and run. Its input parameters are soil organic carbon content (SOC) and SOC/water partitioning coefficient (Koc). It accounts for current chlordecone soil contents and drainage water concentrations. The model was valid for andosol, which indicates that neither physico-chemical nor microbial degradation occurred. Dilution by previous deep tillages makes soil scrapping unrealistic. Lixiviation appeared the main way to reduce pollution. Besides the SOC and rainfall increases, Koc increased from nitisol to ferralsol and then andosol while lixiviation efficiency decreased. Consequently, pollution is bound to last for several decades for nitisol, centuries for ferralsol, and half a millennium for andosol.

Introduction

Chlordecone (CLD), an organochlorine insecticide, was used in the French West Indies to control the banana weevil Cosmopolites sordidus during two periods: (i) 1972–1978 under the trade-mark Kepone™ (5% CLD) manufactured by Life Science Products Co. (Hopewell, VA), and (ii) 1982–1993 under the trade-mark Curlone™ (5% CLD), manufactured by Calliope S.A. (Port-la-Nouvelle, France). Kepone received a temporary use license from the French Ministry of Agriculture in 1972. Its use ceased in 1978 after its prohibition in the USA. A French paper (Snegaroff, 1977) warned against the hazard of soil and water contamination by CLD. Several years after the beginning of Kepone spreading, the topsoil of some banana fields exhibited CLD content higher than 9 mg kg−1. Despite a report intended for the French Environment Ministry (Kermarrec, 1980), which highlighted the contamination of fresh water and terrestrial fauna, the French authorities delivered in late 1981 a license for the use of CLD (Curlone). CLD spreading was restricted to banana fields. Its use was definitively banned in early 1993. Later surveys conducted by the French Department of the Environment (DIREN, 2001) and the French Department of Health (DSDS, 2001) revealed its wide presence in soils, rivers, spring water, but also in drinking water and food crop produce. Authorities immediately asked scientists to investigate pollution duration and trends.

Because chlordecone had been prohibited for a long time in the USA, the scientific literature on chlordecone environmental behavior was sparse and more or less confusing. Given its high peripheral chlorination (Fig. 1), CLD hydrophobicity is high. It has a low solubility in water, except at pH > 9 where chlordecone-hydrate appears (Dawson et al., 1979). It is only largely soluble in hydrogenated organic solvents like benzene or hexane, as attested by the high partitioning ratio between octanol and water (log Kow = 4.5). Its partitioning coefficient (Koc) between the sorbed part on soil organic matter (estimated by the soil organic carbon, SOC) and the dissolved part in water is high, and its value varies widely: 17 500 L kg−1 according to Kenaga (1980), log Koc of 3.3 (Bonvallot and Dor, 2004) or between 3.38 and 3.41 (Howard et al., 1981, cited in ATSDR, 1995), i.e. Koc between 2000 and 2500 L kg−1. Those differences could derive from using soils with different physico-chemical properties. Spatial variability of soil pesticide sorptivity has been described for atrazine by Coquet and Barriuso (2002), but the variation did not cover such a wide range of relative values.

The vapor tension of CLD is less than 3 × 10−2 Pa at 25 °C, and the sublimation temperature is over 350 °C (INCHEM, 1984). Solar irradiation of CLD in the presence of ethylene–diamine results in 78% degradation after 10 days, but no study of the degradation products or their toxicity has been undertaken (Dawson, 1978).

There is no evidence of microbial CLD degradation. Using Pseudomonas aeruginosa strain KO3 and a bacterial pool isolated from Hopewell plant sludge under aerobic conditions, Orndorff and Colwell (1980) obtained an apparent de-chlorination into mono-hydro-chlordecone and di-hydro-chlordecone. But those authors did not describe the initial mono-hydro- and di-hydro-chlordecone contents of the CLD, so that those values could derive from uncompleted chlorination during the production process. Later, George and Claxton (1988) studied CLD degradation by three Pseudomonas spp. But again details were lacking. Moreover, the HPLC chromatogram of non-inoculated CLD solution, with only yeast and glucose addition, showed peaks corresponding to hydro- and di-hydro-chlordecone. The authors underlined that its highly chlorinated cage-like structure makes CLD a poor carbon source for bacteria: after two weeks, the apparent degradation reached only 15–25% with a high relative uncertainty of 10% on chlordecone HPLC concentration measurements.

Only one reference was dedicated to CLD absorption by plants. Topp et al. (1986) studied the uptake of several 14C labeled organic molecules by young barley (Hordeum sativum L.) and cress (Nasturtium officinale). Measured CLD uptakes were negligible for both species. We therefore speculated that soil leaching was the main factor of decontamination.

To account for inter-annual chlordecone desorption and current level of soil pollution, we chose to build a simple leaching model of CLD, WISORCH, based on first-order desorption kinetics. Because of slow decontamination and long-term soil pollution, existing models of pesticide dissipation were not appropriate: they account for short-term sorption, leaching, and degradation so that applying them over years should result in excessive error. In WISORCH, we assumed that neither degradation, nor plant capture was efficient to remove CLD from the various volcanic tropical soil types of the French West Indies. For validation, because inter-annual CLD leaching experiments were impossible, we used a “space for time” approach: we reconstructed the various schedules of CLD spreading on fields and compared the simulated soil CLD contents to the measured ones.

Section snippets

Soils

Table 1 summarizes the distribution pattern of the volcanic soil types and properties according to Colmet-Daage and Lagache (1965). All primary minerals of andesitic rocks are weathered, so that soils have a high content of secondary minerals: halloysite for nitisol, halloysite and Fe-oxihydroxides for ferralsol, and allophane for andosol, the three main soil types contaminated in Guadeloupe. All have a centimeter-size granular structure in the topsoil. Intermediate horizons have polyhedral to

Soil pollution levels

No parcel was contaminated without CLD spreading, even if it was downstream from contaminated parcel.

The [CLD]m measured contents and WISORCH parameters and variables are listed in Table 2 for the 45 parcels. Untilled andosol showed the highest [CLD]m in the 0–30 cm layer. This is in accordance with the persistence of CLD mainly in the first 0–10 cm layer, which reached more than 50 mg kg−1 of CLD. The 50–70 cm layer never reached more than 1 mg kg−1, and it generally contained less than 0.3 mg kg−1.

Discussion

Some of our results were expected. Notably, the soils were polluted only where chlordecone had been spread, except at emergence of contaminated water-tables (data not shown). Indeed, the very low volatility of chlordecone makes atmosphere unable to carry any measurable contamination. Moreover, the CLD loads in runoff flow accounted for less than 3% of those in drainage. Consequently, the inter-parcel contamination remained rare. The soil pollution was spatially delimited. We observed only

Conclusion

Chlordecone, an organochlorine insecticide in the POPs prohibition list of the Stockholm Convention (UNEP, 2005), was regularly applied in the banana crops of the French West Indies at the end of the last century. Our results show that the simple WISORCH model, which applies a first-order kinetics of lixiviation based on SOC, soil carbon/water partitioning coefficient Koc, and drainage to the soil chlordecone stock, accounts for the current chlordecone residues of volcanic soils in Guadeloupe

Acknowledgments

This study was supported by funding from the French Ministry of Ecology and Sustainable Development (MEDD) and the local services of the French Ministry of Agriculture and Fishing. Thanks to J. André, F. Burner, A. Mulciba, C. Palmier, and F. Razan from INRA, and G. Onapin from CIRAD for their technical help.

References (29)

  • Y. Coquet et al.

    Spatial variability of pesticide adsorption within the topsoil of a small agricultural catchment

    Agronomie

    (2002)
  • Dawson, G.W., 1978. Kepone Mitigation Feasibility Report, Appendix A: the Feasibility of Mitigating Kepone...
  • G.W. Dawson et al.

    Kepone – a case study of a persistent material

    The American Institute of Chemical Engineers (AIChE) Symposium Series

    (1979)
  • DIREN, 2001. Le suivi de la contamination des rivières de la Martinique par les produits phytosanitaires. Bilan à...
  • Cited by (217)

    • Chlordecone determination in serum by LC-MS/MS and the importance of low limit of detection

      2023, Journal of Chromatography B: Analytical Technologies in the Biomedical and Life Sciences
    View all citing articles on Scopus
    View full text