Nonpoint source pollution from agriculture is one of the main causes of water quality degradation. To cope with this issue, diverse policy measures have been implemented to promote farming practices that favour water quality. Among these measures, many instruments are based on the voluntary commitments of farmers. Farmers'' participation is therefore important for water quality outcomes. To analyse this issue, we built an agent-based model of a drinking water catchment area using the social-ecological systems framework and the theory of planned behaviour. This model aims to better understand the impact of behavioural dynamics on the effectiveness and efficiency of policies targeting water pollution. Our model allows us to explore how the evolution of farming practices is influenced by, first, the behavioural characteristics of farmers and, second, the characteristics of the policy measures implemented. Our main results are threefold: 1) The characteristics of farmers influence policy effectiveness and efficiency; 2) taking into account farmers' behavioural characteristics in the definition of policies may improve the efficiency of policies; and 3) in situations where behavioural characteristics are unknown, policies combining both financial incentives and training measures are costlier but may be a better option than policies relying on a single measure.