Monitoring natural resources in conflict using an object-based multiscale image analysis approach

Natural resources are sometimes exploited beyond a sustainable level, spoiling natural habitats, displacing local communities, affecting people’s livelihoods and even fuelling armed conflicts. The lack of precise geographic information is a critical limit in the design of appropriate provisions for prevention and response to ongoing crises related to natural resources exploitation. Vast territories have to be observed within a narrow time frame as exploitation activities can easily shift from one area to another. The areas are very difficult to access because they are widely dispersed, too remote or too insecure. Conflict situations often prevent research teams from travelling freely and thus remote sensing provides the potential for complementing more traditional means of monitoring. The study is carried out using high and very high resolution optical data at test sites which are located in the Democratic Republic of the Congo. Various natural resources exist within the test sites such as cassiterite (tin), columbite-tantalite (coltan), wolframite (tungsten) and gold. Deforestation takes place in areas for mining operations and in forests when timber is harvested, transported and sold. This paper presents first steps towards a transferable, robust and fast analysis approach for detecting exploitation of natural resources such as mining. The method used for monitoring active sites combines object-based image and GIS analyses. The designed workflow is concentrating on three levels built on (1) a transferable feature extraction scheme using high spatial resolution (HR) images for identification of potential hot spots, (2) spatial aggregation of the results to regular grid cells and (3) a fine-scale spatial pattern analysis of the potential sites by the interpretation of very high spatial (VHR) imagery. The usability and reusability of rule sets using Cognition Network Language (CNL) enhance the efficiency of image analysis and contribute to efficient and constant monitoring at a service level to the user community. The proposed approach may contribute to focused reactions during crises and yield rapid identification of affected areas. * Corresponding author.