Mapping Gridded Wealth Index Using Open Geospatial Data in Zambia

Low- and middle-income countries are usually poor in statistical systems which particularly necessitates satellite imagery and geospatial data to inform their well-being at the subnational level. While satellite images are overwhelmingly utilized in poverty mapping in these countries, the open geospatial data remains limited integration. This study aims to explore the potential of freely available geospatial data in poverty mapping in sub-Sahara Africa. The household cluster Wealth Index (WI) including 535 samples, derived from Zambia 2018 Demographic and Health Surveys, provides the ground truth observations of poverty. We developed a feature framework from the perspective of potential household poverty that might be influenced by regional poverty. The specific features proxied the economy, population, infrastructure, travel time to public services, and agriculture were then extracted using open geospatial data. The results indicated that the proxy features derived from open geospatial data were able to obtain an \({R}^{2}\) value of 0.78 on the test set in WI prediction in Zambia. Meanwhile, the variable importance metrics highlight that the top 5 high contribution variables are derived from nighttime light images, Open Street Map, Global Human Settlement Layer. The final 4km grid wealth map of Zambia was generated, demonstrating that most of the areas in Zambia fall between “poor” and “middle” living standards. This map presents the wealth disparities within the country, providing a detailed snapshot of the living conditions across Zambia.