Duke Experts Provide Clearest Picture Yet of Global Building Emissions

The Energy Data Analytics Lab is helping strengthen Climate TRACE, which provides the world’s largest inventory of greenhouse gas emissions.

Greenhouse gas emissions have been rising for decades. To reverse this trend swiftly and effectively, decision-makers need to know exactly where emissions are coming from.

Duke University is helping tackle this challenge as part of Climate TRACE, a non-profit coalition that offers an independent, transparent inventory of emissions data and sources worldwide. Governments, companies and others seeking to decrease emissions can use this free and open resource to make high-impact decarbonization choices.

In November 2024, Climate TRACE released the largest and most comprehensive emissions inventory to date. Duke’s contribution to the massive dataset is the highest-resolution information available today on direct emissions from residential and non-residential buildings (including retail, warehouses, hospitals, schools and more).

The building sector contributes an estimated 7 percent of total global emissions.

“Our data for buildings help Climate TRACE provide a detailed picture of emissions sources at a high spatial resolution and illuminate high-impact mitigation opportunities at all levels from national to municipal,” said Kyle Bradbury, a Pratt School of Engineering faculty member who directs the Energy Data Analytics Lab at the Nicholas Institute for Energy, Environment & Sustainability. Bradbury leads the building sector team, which includes Duke alumnus and former Pratt faculty member Jordan Malof, now at the University of Missouri.

Bradbury said the building sector data can inform individual cities’ climate action plans and strategies, as well as decarbonization efforts by advocacy groups. With further research, the data could eventually be used regionally to help monitor emissions reduction progress and to identify communities that could benefit from energy efficiency retrofits.

Launched in 2020 by former U.S. Vice President Al Gore and WattTime Executive Director Gavin McCormick, Climate TRACE is comprised of a dozen organizations (including Duke) and more than 100 additional contributors. The group leverages advanced data science techniques and artificial intelligence to trace greenhouse gas emissions back to their sources.

Each member of the Climate TRACE coalition brings its own unique expertise to track emissions from a particular sector. The Duke-led team estimating building sector emissions uses data derived from applying machine learning techniques to satellite imagery. The Energy Data Analytics Lab has pioneered similar approaches to address a variety of energy and climate questions. The team is also investigating new ways of integrating AI tools into this process.

The team’s work has resulted in a neighborhood-level inventory of direct building emissions, which primarily come from onsite fossil fuel combustion for space heating, water heating and cooking. (Bradbury notes that emissions from lighting, consumer electronics and most air conditioning are typically electric and accounted for elsewhere in the broader Climate TRACE inventory.)

The dataset covers every 1-square-kilometer area of Earth’s land with buildings—a resolution 100 times sharper than the previous best estimates for the sector. For reference, the Energy Data Analytics Lab’s model results in 450 datapoints for Manhattan, compared to just four or five from the earlier estimates, Bradbury noted in a recent interview.

The lab’s approach relies on publicly accessible, high-resolution estimates of the size of residential and nonresidential buildings derived from satellite imagery. The team combines these data with available regional estimates of energy use to “super-resolve” emissions data from the Emissions Database for Global Atmospheric Research (EDGAR). The higher resolution is critical to break down the EDGAR data to make it more useful in national, subnational and local emissions inventories, Bradbury explained.

The Duke-led team is further refining its estimates of energy use intensity through machine learning techniques to incorporate weather data, economic factors and regional preferences and norms, Bradbury said. The researchers are also working to reflect seasonal changes in energy consumption at an even higher temporal resolution.

“Understanding climate challenges—and advancing the most timely and high-impact solutions—requires access to top-quality, actionable data,” said Brian Murray, director of the Nicholas Institute. “Climate TRACE is an extraordinary resource for decision-makers aiming to accelerate climate progress, and the scholars at Duke’s Energy Data Analytics Lab are helping make it even more robust and useful.”