CONTEXT: Chronic environmental pollutions are one of the most striking examples of negative externalities of intensive agriculture. To address this issue, an agroecological transition is necessary, requiring new design methods: While most design methods are suited to the field or farm scale, some innovations require an approach at the scale of the territory. OBJECTIVE: In this article, we propose a new method to design an agricultural territory aiming to overcome an environmental pollution issue, i.e. to identify one or more combinations of agricultural practices, implemented by the farmers of the territory that reduce water pollution. METHODS: We implemented the new method following three steps: first, we used a sociotechnical diagnosis to identify cognitive biases that prevent the actors from innovating in a disruptive way. Second, we used the K C method (based on C K theory) to allow actors to innovate in a disruptive way. Third, we used a serious game to continue the design of the innovations and to test and assess them in a safe place. We implemented this new method in the Galion River watershed in Martinique (French West Indies) whose river is polluted by herbicides. RESULTS AND CONCLUSION: Using the method, actors have overcome existing path dependencies. Based on this, the method made it possible to design different innovations across scales in a participative way and with a systemic territorial point of view. Here, actors designed 15 innovations that enhance either the territory or the pooling of its resources. Among these, there was the sharing of the service of an itinerant breeder or the implementation of an ecotouristic circuit. Actors combined innovations across scales into three scenarios that created a new agricultural territory with fewer agricultural herbicides in the river. SIGNIFICANCE: We suggest adding two participatory steps to improve the level of genericity of the method: a first step to discuss the initial question and the general organisat