Monitoring Charcoal Production Impacts in Somalia Using Multisensor Satellite Data

Charcoal production, through indiscriminate logging, poses significant threats to the bushland ecosystems in Somalia, exacerbating an already fragile socio-political and environmental context. Particularly in the Somali semi-arid climate, the reduction in vegetation cover can have negative impacts on soil moisture retention, biodiversity, and increased runoff, thereby undermining the resilience of the whole system essential for local pastoral livelihoods. In this study, we investigated the application of freely available multi-sensor remote sensing products (Sentinel-1 and Sentinel-2) combined with machine learning techniques (Random Forest - RF and Support Vector Classification - SVC) to monitor and spatially detect the presence of charcoal production sites (i.e., kilns) over large areas. The approach, based on applying the RF and SVC models to the difference between multisensor composites of consecutive years, was tested between 2019 and 2021 over a remote area of approximately 11,500 km 2 northwest of Mogadishu. An extensive reference dataset of visually identified kilns from very high-resolution imagery (ca. 75,000 points, FAO-SWALIM) supported training and validation. The classification of kiln presence or absence achieved stable and robust performances (overall accuracy: 0.81-0.88; F1 score: 0.81-0.87) and the resulting maps captured the distribution of charcoal kilns at 10 m resolution, providing a spatial picture of how this phenomenon is driving bushland degradation in Somalia. Our results demonstrate that open-access data and tools can replace the costly very high-resolution imagery for continuous monitoring, providing a scalable and replicable methodology transferable to other dryland regions and informing land management strategies aimed at curbing unsustainable charcoal production.