4.8. Conclusion and outlook
In this study, we benefited from a unique dataset collected at the scale of the national territory, covering a wide diversity of production situations and offering a large sample of farms with contrasting management strategies. Using partitioning methods, we first identified several types of production situation associated with a first level of variability in pesticide use. Using once again partitioning methods, we secondly identified, in each production situation, various profiles of managements strategies associated with low vs. high pesticide use, i.e. explaining a second level of variability in pesticide use. We showed that MSs with low pesticide use were different from a production situation to another, confirming the importance to consider the production context when addressing the question of drivers to reduce pesticide reliance. In a given production situation, we found that several profiles of management strategies displayed low levels of pesticide use, suggesting that there is not only one but several ways to be low reliant on pesticides. Although we highlighted the diversity of strategical options as a function of the situation, some management measures, such as the crop diversification, the soil tillage strategy and the pesticide dose reduction appeared as rather transverse discriminative factors, i.e., they were associated with low pesticide use across diverse situations. Mechanical weeding, wheat cultivar and sowing date were conversely more punctual discriminative factors as they contributed to discriminate strategies with high vs. low TFI in only one or two of our categories of production situation. The analysis of TFI variability could still be re- fined by collecting additional variables to better characterise factors from production situation and management strategy, in particular by getting information on farmer's decision-making.