Persistent socio-economic and environmental inequalities pose major challenges to sustainable development in the global South. However, comprehensive and spatially clear data on environmental conditions and socio-economic well-being remain scarce, preventing a thorough analysis of intersecting inequalities. This study assesses economic well-being and its relationship to environmental factors in South Africa by proposing a method for analysing environmental and socio-economic inequalities using remote sensing data and transfer learning, using publicly available satellite imagery and statistics. We take the established correlation between nighttime light intensity and economic activity and propose a framework to analyze it in parallel with environmental indicators derived from daytime satellite imagery. Our approach centers on training convolutional neural network (CNN) models to extract economic and environmental features from high-resolution daytime satellite data. CNNS are trained to predict nighttime light intensity, act as proxies for economic activity, while learning to recognize environmental features. Patterns indicating economic activity and environmental conditions can be identified from daytime images alone. By linking the extracted features to known socio-economic indicators obtained from census data and surveys, a spatially clear map of South Africa's economic well-being and environmental quality was created.