Climate TRACE, the world's most widely used greenhouse gas emissions database co-founded by former US Vice President Al Gore, may seriously underestimate CO2 emissions from urban road traffic, a new study shows, raising concerns in the scientific and policy community about the reliability of the data. The research, led by a team at Northern Arizona University, shows that within U.S. cities, vehicle CO2 emissions given by Climate TRACE are on average about 70% lower than another mature database, with individual cities even underestimating them by more than 90%.
The research was published in the journal Environmental Research Letters and was led by Kevin Gurney, a professor in the School of Informatics, Computing and Network Systems at Northern Arizona University. The research team focused on the urban road vehicle emissions part of Climate TRACE, systematically compared its estimation methods and results, and cross-validated them with the "Vulcan" highway emissions database built by Gurney's team for a long time. The Vulcan database is based on official traffic statistics and fuel use data and uses standardized methods to accurately characterize CO2 emissions from burning fossil fuels on urban roads.
According to Bilal Aslam, co-author of the paper and a postdoctoral researcher at the school, the Vulcan highway emissions data itself has an uncertainty of about 14%, but this is far smaller than the huge difference between the two databases this time. In a head-to-head comparison of vehicle CO2 emissions in 260 U.S. cities, Climate TRACE’s urban road emissions estimates were on average 70% lower than Vulcan’s. Another co-author, researcher Pawlok Dass, added that in cities such as Indianapolis and Nashville, Climate TRACE’s emissions data were more than 90 percent lower than Vulcan’s.
The research team believes that this systematic underestimation is likely not limited to U.S. cities and may also appear in urban emission estimates in other countries and regions. In addition, Gurney's team also found similar underestimation of CO2 emissions in a previous analysis of Climate TRACE power plant emissions data. The superposition of multiple results has led researchers to raise broader questions about this global climate monitoring system that relies on the rapid development of artificial intelligence technology.
Climate TRACE is a project that uses satellite remote sensing, big data and artificial intelligence technology to conduct high-resolution tracking and estimation of global greenhouse gas emissions. In recent years, more and more governments and cities have used its data as an important basis for formulating climate policies and evaluating emissions reduction progress. In this study, Gurney's team focused on examining the project's algorithm path and data sources for urban vehicle emissions, and pointed out that if emissions from key sectors are systematically underestimated, it will directly affect the accuracy of the city's formulation and assessment of emission reduction targets.
Gurney said urban vehicle CO2 emissions accounted for such a large proportion of a city's overall carbon footprint that any emerging "high-tech" emissions data needed to be rigorously examined. He stressed that while new methods based on artificial intelligence are "promising", using such data directly for policy development in the absence of transparency, independent verification and adequate peer review could send misleading signals to policymakers and the public. In his view, current results suggest that Climate TRACE data may significantly underestimate more than half of fossil fuel CO2 emissions in U.S. cities.
The study authors pointed out that artificial intelligence has the potential to become an important tool for environmental monitoring in the future, but only if it is operated within a rigorous scientific framework. This means that algorithmic assumptions, data sources and uncertainties need to be transparent and cross-checked with traditional, longer-tested emissions inventories. Only in this way can AI-driven monitoring systems truly provide credible support for climate policy, rather than creating new blind spots.
To this end, the paper puts forward a number of improvement suggestions for Climate TRACE, including: further strengthening the coupling and comparison with official energy and transportation statistics; establishing more detailed calibration parameters for different departments and regions; and introducing an independent research team to conduct regular audits and methodological evaluations. The research team believes that this is not only related to the accuracy of a single database, but also related to how governments arrange their emission reduction budgets and prioritize which emission "hot spots" to control.
"We will never be able to quantify emissions with 100% accuracy, but we have a responsibility to ensure that the data provided to policymakers and the public are statistically unbiased and meet the most rigorous scientific standards available." Gurney said when talking about the significance of the research. He warned that if there are systematic biases in emissions data, it could mislead decision-making and undermine public trust in climate governance capabilities.
Gurney's research career has lasted for more than 20 years, and he has long been committed to the refined quantification of greenhouse gas emissions. The Vulcan and Hestia projects he leads have important influence in the American academic and policy circles. With funding from multiple federal agencies, these two programs have constructed a greenhouse gas emissions grid map covering the entire United States, detailing emissions down to individual power plants, city blocks and even specific roads, providing tools for identifying high-emission "hot spots" and formulating differentiated emission reduction plans. Comparison of relevant studies with atmospheric observation data shows that these emission estimates are in good agreement with actual monitoring results.
Currently, countries are increasingly demanding high-resolution emissions data as they implement their emission reduction commitments, which has also driven the rapid rise of projects like Climate TRACE that rely on AI and remote sensing technology. However, the data bias revealed in this study shows that while pursuing "new" technology and "fast" monitoring, we cannot ignore the value of the traditional statistical system and the basic requirements of scientific processes. The research team calls for closer collaboration and mutual verification mechanisms between innovative methods and mature inventories to ensure that climate policies are based on the most reliable facts possible when building a global emissions monitoring system in the future.