Models of pesticide emissions and impacts adapted to LCA have been developed according to temperate pedoclimatic and farming conditions. Because, tropical farming practices and environmental conditions differ widely from those in temperate regions, the aim of our study is to assess pesticide impacts in a cradle-to-farm-gate LCA study on tomato production in Martinique and to investigate the use of the models under these tropical conditions. To that purpose, we tested the most up-to-date models, the adapted PestLCI 2.0, USEtox 2.1 and dynamiCROP, to 9 tomato fields from Martinique with variable climates, soils and pesticides. Our first results highlight that improvements of the models are required. Indeed, in PestLCI, no drift curve for tropical conditions is available, and water processes such as leaching and runoff are not well taking into account. An update and extension of climate and pesticide databases is also necessary. In the cradle-to-farm-gate LCA results for non-cancer human toxicity impacts, pesticides account for up to a maximum of 42%, mainly associated to pesticide residues in crop, while emission-related exposure to pesticides are less relevant. Pesticide can account for up to one third of cradle-to-farm-gate freshwater ecotoxicity impacts. In conclusion, in the cradle-to-farm-gate LCA of tomato from Martinique, the contribution from pesticide residues in crops to non-cancer human toxicity may be large. First improvements of PestLCI have been identified (e.g. drift curve), but to better assess the applicability of the models to tropical conditions, comparisons between two tomato systems, one in temperate and one in tropical conditions, is required.