Predicting Economic Indicators through Nighttime Satellite Imagery

Research Poster Engineering 2025 Graduate Exhibition

Presentation by Janice Tran

Copresented by Muhammad Faizan Raza and Tanuja Voruganti

Exhibition Number 6

Abstract

Nighttime satellite imagery offers a scalable and cost-effective approach to monitoring economic activity, particularly in regions with inconsistent reporting. This study examines the relationship between nighttime light emissions and key economic indicators—GDP, population, and CO emissions—by integrating global satellite data with economic datasets from institutions such as the World Bank and the U.S. Department of Education. Through geospatial analysis, data preprocessing, and machine learning techniques, we develop predictive models to quantify economic trends. Our ensemble model demonstrates high predictive accuracy (R² = 0.9733) for GDP, with minimal errors for the U.S. (0.41%) and India (0.54%). Beyond GDP estimation, we apply this framework to predict CO emissions for U.S. universities, identifying discrepancies between institutional sustainability commitments and actual emissions. Results indicate that light intensity reliably captures economic fluctuations, reflecting downturns associated with global events like the oil crash and COVID-19 lockdowns. The findings support nighttime luminosity as a robust energy proxy, providing a scalable alternative for assessing economic health and emissions without reliance on traditional self-reported data. This research highlights the potential of nighttime light data as an innovative tool for economic and environmental monitoring. By leveraging machine learning and satellite imagery, we introduce a framework that enhances data-driven policymaking, offering insights for researchers and institutions focused on sustainability and economic development.

Importance

This study highlights the potential of nighttime satellite imagery as a valuable tool for monitoring economic activity and environmental impact. By analyzing light emissions, we can estimate key indicators like GDP and CO2 emissions in a scalable and cost-effective way, even in regions with limited data availability. Our findings also demonstrate how universities, as hubs of research and innovation, play a role in energy consumption, offering a new method to assess sustainability efforts. By integrating machine learning with satellite data, this research provides policymakers, researchers, and institutions with a reliable approach to tracking economic health and environmental trends, supporting informed decision-making for sustainable growth and development.

DEI Statement

This research leverages nighttime satellite imagery to bridge data gaps in economic monitoring, particularly in regions with limited reporting. Traditional economic data often overlook underserved communities, reinforcing disparities in policymaking and resource allocation. By using satellite-based methods, this study provides a scalable, cost-effective approach to estimating GDP and CO2 emissions, helping to address economic justice and sustainability concerns. Additionally, the study examines university emissions, offering insights into institutional sustainability efforts and their broader societal impact. By integrating machine learning with geospatial data, this work contributes to equitable access to reliable economic indicators, supporting data-driven policies that promote social mobility, environmental responsibility, and economic inclusivity for historically underrepresented populations.