Informal Settlements
EO Capability Benefits
Sustainable Development Goal Indicator 11.1.1 requires countries to quantify the “proportion of urban population living in slums, informal settlements or inadequate housing”. Especially in developing countries where accessibility and local resources are limited, traditional field mapping is time-consuming and costly. Satellite data offers a scalable and affordable alternative. The Extent and Type of Informal Settlement EO Capability provides national and local authorities as well as civil society with a scalable and replicable tool for monitoring informal settlements and assessing the efficacy of policy interventions or slum upgrading efforts.
EO Capability Description
While definitions of informal settlements can vary significantly between countries and contexts, the UN has established five key “deprivations” that must be present for an area to be classified as informal. These include: lack of access to improved water, improved sanitation, sufficient living area, durability of housing, and security of tenure. Some of these indicators—such as security of tenure or access to improved water and sanitation—are naturally not detectable from space. Moreover, local reference data may not always be available. When a widely accepted local slum definition is lacking, identifying informal settlements becomes even more challenging due to the diversity of conditions and characteristics these areas can exhibit. In such cases, it is necessary to work with local stakeholders to agree on the relevant slum characteristics, taking into account varying building typologies, levels of infrastructure, settlement locations (e.g., areas along railway tracks), and other metrics.
Once a set of characteristics has been agreed upon, optical Very High Resolution (VHR) imagery serves as the primary input data source. If a sufficiently large quantity of local training data is available, Convolutional Neural Network (CNN)-based models can be utilised. Otherwise, Object-Based Image Analysis (OBIA) relying on segmentation procedures and rule-based classification can be used, or a hybrid approach in combination with CNN-based models. Rather than a binary map of slum extent, probability-based heatmaps with a minimum probability threshold are more reflective of the uncertainties inherent in slum detection. To facilitate the generation of actionable insights for policy makers and urban planners, zonal statistics can then be used to aggregate these probabilities into larger units, for instance, at the sub-district or neighbourhood level.