Modelling of pesticide leaching is paramount to managing the environmental risks associated with the chemical protection of crops, but it involves large uncertainties in relation to climate, agricultural practices, soil and pesticide properties. We used Latin Hypercube Sampling to estimate the contribution of these input factors with the STICS-MACRO model in the context of a 400 km² catchment in France, and two herbicides applied to maize: bentazone and S-metolachlor. For both herbicides, the most influential input factors on modelling of pesticide leaching were the inter-annual variability of climate, the pesticide adsorption coefficient and the soil boundary hydraulic conductivity, followed by the pesticide degradation half-life and the rainfall spatial variability. This work helps to identify the factors requiring greater accuracy to ensure better pesticide risk assessment and to improve environmental management and decision-making processes by quantifying the probability and reliability of prediction of pesticide concentrations in groundwater with STICS-MACRO.

农药淋洗模型对于管理与作物化学保护有关的环境风险至关重要，但涉及气候，农业实践，土壤和农药特性方面存在很大的不确定性。我们在400 km ²的背景下使用STICS-MACRO模型使用Latin Hypercube Sampling来估计这些输入因子的贡献。法国的一个集水区，玉米中使用了两种除草剂：bentazone和S-异丙甲草胺。对于这两种除草剂，影响农药淋洗模型的最主要输入因素是气候的年际变化，农药吸收系数和土壤边界水导率，其次是农药降解半衰期和降雨空间变异性。这项工作通过量化使用STICS-MACRO预测地下水中农药浓度的可能性和可靠性，有助于确定需要更高准确性的因素，以确保更好的农药风险评估并改善环境管理和决策过程。