Buildings account for approximately 9% of global greenhouse gas emissions each year, yet emissions data for the building sector is often outdated or too low-resolution to support effective policymaking. Climate TRACE—a global nonprofit coalition that independently monitors emissions across 83% of global sources—seeks to fill this gap. As part of its effort to expand coverage to the building sector, Climate TRACE has partnered with Duke University on a project to estimate Energy Use Intensity (EUI) for both residential and non-residential buildings worldwide.
We used EUI data from the World Bank’s CURB Tool, representing 483 locations across the globe, as the dependent variable. Feature data for these locations was collected from satellite imagery, as well as climatic, geographic, and socioeconomic sources. Using these inputs, we trained machine learning models to predict EUI. After evaluating several algorithms, we found that the Random Forest model delivered the best performance. Among the features, Income Index, Average Temperature, and Latitude emerged as the most influential in predicting energy use intensity. These results can support future researchers in generating high-resolution EUI predictions to improve emissions estimates while also helping policymakers better understand spatial patterns of energy consumption and design more targeted emission reduction strategies.
Mentor: Kyle Bradbury
