In mountain landscapes, agricultural abandonment is taking place in the most vulnerable areas, while intensification increases in the most productive lands. These contrasting processes, which have different impacts on biodiversity and ecosystem services (BES), are related to changes in the farming system component of these landscapes.Farming systems are identified based on farmer's decisions on, for example, type of crop and level of fertilizers, which represent the descriptors of farming systems and can be grouped into several dimensions (e.g. land use and intensity). Since obtaining this data at farm-level is often difficult, an alternative is to study the spatial combinations of farming systems at parish-level, i.e., Farming System Mixes (FSM), relying on agricultural census data.Other biophysical (e.g. climate, soil) and socioeconomic (e.g. labour, farmer's age) variables, independent of farmers' decisions, represent the exogenous drivers of these decisions. The separation between descriptors and drivers is important to improve knowledge about what drives farmers' decisions regarding farming system choice, as these choices are often the focus of policies aiming the support of BES.In this study, we explored the underlying drivers of FSM and assessed the role of socioeconomic drivers, main target for policy makers, in a context of strong biophysical gradients. Biophysical drivers emerge as those that primarily discriminate between the FSM located in different topographic positions (valleys, mountains and plateau). In the situations where there is a greater range of productive choices available for farmers, such as in valleys, socioeconomic drivers assume a preponderant role on farming system choice.