Capacity Building & Knowledge Transfer In Seaweed Mapping In Indonesia

Deskripsi

The formulation of effective strategies to develop the seaweed industry in Indonesia is crucial in the effort to combat poverty in coastal communities. The Commodities Research Group’s Strategic Integrated Project (SIP) has recently demonstrated how high-resolution and high-frequency satellite image datasets could contribute to this objective by its ability to enhance current production monitoring systems and provide vital information to value chain actors, stakeholders, and policymakers. This research project will expand on SIP’s fourth research objective by refining and enhancing the satellite image processing and analysis methods to develop a well-defined seaweed production mapping and monitoring system using deep learning. The temporal spectral and spatial information extracted will enable a detailed study of the production dynamics of an area of interest to facilitate the characterisation of the growing areas based on their production characteristics. This will help the industry better understand the production dynamics of an area for improved planning and policy formulation to enhance the coordination between value chain actors, stakeholders, and policymakers. The seaweed production mapping and monitoring system will also create opportunities to monitor and study significant environmental and socio- economic shocks. The final research output is an enhanced, rapid, and automated but relatively low-cost image processing and analysis method that will provide reliable seaweed multitemporal maps and potential production status information to inform the formulation of policies, strategies, and actions within the growing industry. The deep learning image segmentation approach using UNet implemented in this study using PlanetScope data in 3-m spatial resolution produced a high accuracy (> 0.9) with low to moderate boundary information accuracy (0.2 – 0.6). The most stable deep learning model produced can be used to monitor the multitemporal progression of seaweed farming to show the dynamics of seaweed production farming over time. The training data, model, and script to produce the analysis will be provided as an open-access source to facilitate capacity building and adoption of the production monitoring system by the stakeholders in the industry.