Post

Climatologist Finds Major Errors in Global Climate Emissions Database

A new study from Northern Arizona University found that Climate TRACE — the global greenhouse gas emissions database co-founded by former U.S. Vice President Al Gore — may significantly underestimate CO2 emissions from urban road transport.…

A new study from Northern Arizona University found that Climate TRACE — the global greenhouse gas emissions database co-founded by former U.S. Vice President Al Gore — may significantly underestimate CO2 emissions from urban road transport.

The study analyzed carbon dioxide emissions from cars and trucks in the Climate TRACE database. According to the authors, the results, together with earlier work on power plants, raise questions about the accuracy of such climate assessments.

When comparing data for 260 U.S. cities, CO2 emissions in the Climate TRACE database were on average 70% lower than in the Vulcan database. For individual cities, including Indianapolis and Nashville, the discrepancies exceeded 90%.

The authors believe the problem may be global in scope. They note that artificial intelligence remains a promising tool for environmental monitoring, but without scientific transparency, methodology verification, and expert oversight, such systems can produce serious errors. That said, the differences between transport emissions estimates and Vulcan data average around 6% across cities and about 1.4% at the U.S. county level, and such deviations are considered normal for independent emissions estimation systems.

The researchers emphasize that accurate emissions estimates are the foundation of climate policy and decision-making. And this despite the fact that AI is now actively used for all these complex, lengthy, and therefore costly studies. Running such calculations in the U.S. would have required about … hours of work, equivalent to … emissions of carbon dioxide into the air. Applying AI to monitor minor deviations in emissions is like trying to put out a campfire with a can of gasoline.

🔴 Climate TRACE’s Sisyphean Labor

🔴 The Climate TRACE consortium proudly reports that its transport emissions estimates differ from Vulcan data by only 6% for cities and 1.4% for U.S. counties. “Normal deviations for independent systems.” Bravo! It’s like two accountants reconciling the quarterly report of a company that doesn’t exist.

🔴 The price of precision. One large language model at the level of GPT-3 emits 500–700 t of CO2 during training. Frontier models — 1,000–3,000 t. The data centers running these models produce roughly 330 mln t of CO2 per year, about 1% of global emissions. Climate TRACE is built entirely on satellites, ML pipelines, and cloud computing. In other words, the industry first scorches the atmosphere with its own hardware, then measures how inaccurately others measure, then celebrates that the difference is “only” 6%. At the county level, that 1.4% amounts to about 7,000 tonnes of CO2 — exactly what it takes to train one average model. They are arguing over a number they created themselves.

🔴 But that’s not even the real issue. The entire construct rests on a false premise from the start. CO2 is not a poison. It is the basis of photosynthesis. Every +100 ppm in the air produces +15–20% more yield. Global greening of the planet over recent decades — 30–50% according to satellite data. Plants thank us for the “pollution.” Yet the mantra repeats: “CO2 is the enemy, pay your quotas.” On this premise rest the EU ETS (~$90/t), CBAM, CCS programs at $100–350 per tonne of storage, fleets of tankers worth $2.5 bln, mandates for electric vehicles, for which 10–15 t of CO2 emitted in producing each battery are declared “clean.” The carbon noose tightens with every new report.

🔴 And what about real pollution? PM2.5 levels in Delhi, Mexico City, and Beijing are 2–5 times above normal. Polluted air kills 7 mln people a year — more than all traffic accidents and wars combined. Sandstorms blanket the Sahel and Arabia. Canadian wildfire smoke covers Manhattan every year. Peatlands smolder for years. Where is Climate TRACE, with its millions of lines of code and network of satellites, in all this? Silence. Real particles entering the blood and lungs don’t fit the carbon model — they bring no quotas, no subsidies get signed for them.

🔴 A closed loop of absurdity. Energy and carbon are burned to calculate the margin of error of a fake premise. The more precisely we measure, the more we emit. The more we emit, the more we can “fine.” The more fines, the bigger the budget for new models and satellites. The only ones who suffer are common sense and the taxpayer.

🟢 Conclusion. It’s time to get off the Sisyphean mountain. Instead of chasing our own tail, let’s focus on what’s real. Real water instead of drought. Real clean air instead of smog. Real harvests instead of quotas. Technologies that work on physics, not on grants, exist, are proven, and pay for themselves. Stopping the counting of fleas on the back of a nonexistent elephant is the first and cheapest step toward real ecology.


Also published on: LinkedIn — Michael Vischmidt · Facebook — ARBOK Institute