Measure to Manage – Field boundary detection for adequate land management practices towards adapted digital learning resources

Kurzfassung

Accurate information on field boundaries is crucial for effective land management. It forms the basis for deriving property rights, managing resources, and planning agricultural activities. In Western Africa, where agriculture is a cornerstone of livelihoods, farmers often do not know the exact size of their cropped fields. This can lead to under- or overuse of inputs such as seeds and fertilizers, which can reduce yield and/or profitability, and have long-term ecological impacts. In Senegal, irrigated rice in the Senegal River Valley is a key contributor to the country’s food balance. Efficient field boundary detection helps to reduce the cost of profiling heterogeneous smallholder farms, building profit and loss sheets, and monitoring agricultural activity therein. Remote sensing can support the automatization of field boundary identification. By training a convolutional neural network (CNN) to delineate field boundaries from Planet data, field-level agricultural management can be monitored more cost-effectively. It curtails the need for physical plot circumnavigation or on-screen manual delineation, the cost of which can then be redirected to monitor other management practices for which proximal interactions are more critical (e.g., timing and dosing of fertilization). Within the project COINS (Co-developing innovations for sustainable land management in West African smallholder farming systems), the integration of Earth Observation (EO)-derived field boundaries in Manobi Africa´s agCelerant platform has been identified as one of 6 priority minimum value products (MVPs) from EO and aims to support smallholders’ implementation of sustainable intensification (SI) practices. Participatory approaches are mainstreamed to combine local knowledge and expertise with satellite-derived information and resolve persistent discordances linked, for example, to over- or under-segmentation of smallholder plots. Leveraging SI practices with phygital approaches (a combination of digital tools and proximal presence) is further promoted through elearning modules to be simultaneously embedded in a Massive Open Online Course (MOOC) accessible for students and scientists across disciplines, and in last-mile initiatives such as agCelerant Academy mobile.