Open data for algorithms: mapping poverty in Belize using open satellite derived features and machine learning

Several methods have been proposed for using satellite imagery to model poverty. These include poverty mapping using convolutional neural networks applied either directly or via transfer learning on high-resolution satellite images, or combinations of methods that merge satellite imagery with standard approaches. However, many of these methods require proprietary imagery which, due to cost and infrequent acquisition, restricts practical deployment.

This study investigates whether satellite-derived poverty maps can be improved by incorporating features from Sentinel-2 and MODIS, two open-source and freely available imagery sources. The authors estimate a poverty map for Belize that integrates spatial and time-series features extracted from these sensors, both with and without survey-derived variables.

Results show an 8% improvement in model performance when open-source satellite features are included. The authors conclude that Open Data for Development should adopt open-data pipelines wherever possible.