Scientists and policymakers are promoting Conservation Agriculture (CA) to reduce soil erosion and greenhouse gas emissions. While numerous studies have already highlighted the diversity of CA practices, no method has been proposed for identifying and evaluating it. This thesis proposes a multi-scale, interdisciplinary, participatory methodology to identify and assess this diversity. In Wallonia, five CA-types have been identified and differentiated by organic certification, the presence of temporary grassland, and the proportion of industrial crops in the rotation. Analysis shows significant differences in soil structural stability and carbon content: CA-types that include temporary grassland in their rotation, even if they occasionally plow, have higher stability and carbon content than those that have abandoned the plow altogether and grow industrial crops in their rotation. Farmers’ incentives to adopt these practices vary within the same CA-type, as do their prospects for changing CA practices. This study highlights the diversity of CA practices at the regional level, the diversity of impacts, and the diversity of incentives and prospects for change within CA-types, with implications for other areas and farming systems. More generally, it questions the boundaries of farming systems and the policy choices associated with them. ; Scientifiques et politiques promeuvent l’Agriculture de Conservation (AC) pour réduire l’érosion des sols et les émissions de gaz à effet de serre. Cependant, tandis que de nombreuses études ont déjà souligné la diversité des pratiques agricoles existantes en AC, aucune méthode n’a actuellement été proposée pour l’identifier et l’évaluer. Cette thèse propose une méthodologie multi-échelle, interdisciplinaire et participative, pour identifier et évaluer cette diversité. Sur le territoire wallon, cinq types d’AC ont été identifiés, se distinguant par la certification biologique, la présence de prairie temporaire, et la part de cultures industrielles ...