Abstract

In 2009, the U.S. Environmental Protection Agency (EPA) declared six greenhouse gases a threat to public health and welfare. This Endangerment Finding became the legal basis for federal emissions controls. The EPA based this decision mainly on climate model projections from the IPCC Fourth Assessment Report—projections that sixteen years of real-world data have now shown to be wrong. In July 2025, the Department of Energy (DOE) Climate Working Group—John Christy, Judith Curry, Steven Koonin, Ross McKitrick, and Roy Spencer—published a detailed review showing the actual evidence does not support the catastrophic predictions behind the original finding. This paper looks at the DOE report alongside sixteen years of observational data and finds that climate models consistently overpredict warming, U.S. extreme weather has not gotten worse, CO2 is delivering real benefits to agriculture that were left out of the policy analysis, sea level rise has not sped up beyond normal historical rates, and warming actually saves more lives from cold than it costs from heat. The paper also documents a string of high-profile climate predictions—ice-free Arctic summers, 50 million climate refugees, the death of the Great Barrier Reef—that simply did not happen. The conclusion is straightforward: the science behind the 2009 Endangerment Finding has not held up, and the finding should be reconsidered.

1. Introduction

In December 2009, the EPA issued its Endangerment Finding under Section 202(a) of the Clean Air Act, declaring that atmospheric concentrations of six greenhouse gases endanger public health and welfare [EPA, 2009]. The agency based this almost entirely on computer model projections from the IPCC Fourth Assessment Report (AR4), not on direct observational evidence [IPCC, 2007]. As Spencer (2021) put it, the finding was “based almost entirely on theoretical climate model projections that had not been validated against real-world observations.”

Sixteen years later, we now have the observational record to test those projections. The DOE Climate Working Group published its critical review in July 2025, concluding that the evidence has shifted decisively against the assumptions behind the Endangerment Finding [DOE Climate Working Group, 2025]. Their analysis used satellite temperature records, surface station data, extreme weather databases, agricultural productivity numbers, and sea level measurements—the real-world evidence that the original finding lacked.

In response, Kopp et al. (2025) assembled 85 scientists to produce a 450-page rebuttal. But the Kopp response mostly just restated model-based projections rather than addressing the core problem: the models don’t match what we actually observe. This paper looks at the real-world evidence across five critical areas and asks a simple question: does the scientific case for the Endangerment Finding hold up when tested against what has actually happened?

2. Climate Models Have Failed Their Most Important Test

2.1 The Prediction That Matters

The whole case for the Endangerment Finding rests on one central claim: that increasing atmospheric CO2 will produce dangerous warming. This claim comes from global climate models (GCMs) that project equilibrium climate sensitivity (ECS)—the warming you get from doubling CO2—at levels high enough to cause serious harm. If the models are overpredicting sensitivity, the entire regulatory case falls apart.

2.2 Satellites vs. Models

The satellite temperature record, maintained by Spencer and Christy at the University of Alabama in Huntsville (UAH) since 1979, gives us the most complete and consistent picture of atmospheric temperature trends across the globe. Christy and McNider (2017) showed that the CMIP5 model ensemble overpredicts tropical mid-tropospheric warming by a factor of roughly two to three compared with both satellite and weather balloon data. This is not a minor point—the tropical troposphere is exactly where greenhouse theory says the strongest warming should show up, the so-called “tropical hot spot.” The fact that it is largely missing from the observations is a fundamental problem for the models.

Spencer (2013) showed that the actual warming trend in the satellite era is about 0.14°C per decade, while the model average projects 0.27°C per decade—nearly double what we actually see. McKitrick and Christy (2020) carried this analysis forward into the CMIP6 model generation and found the same warm bias, with models still projecting warming rates well above what is being observed in the tropical troposphere.

2.3 Energy Budget Estimates Contradict Models

When you use actual observations to calculate ECS—what is called an energy budget approach—you consistently get values well below what the climate models produce. Lewis and Curry (2018) estimated ECS at 1.50°C (1.05–2.45°C, 5–95% range), putting the most likely warming from doubled CO2 at roughly half what the models assumed when they informed the Endangerment Finding. Lewis (2023) refined this estimate further downward using updated aerosol data.

Even the IPCC AR6, despite raising its lower bound to 2.5°C, acknowledged that “whichever line of evidence is used, a best estimate of ECS below 2°C cannot be ruled out” when observational data are given proper weight [IPCC, 2021]. The AR6 range of 2.5–4.0°C is really a compromise between conflicting lines of evidence—with models and paleoclimate pulling the number up while the actual measurements pull it down.

2.4 What This Means for Policy

Curry (2025) put it bluntly: it is “difficult to argue that global climate models are fit for any policy-relevant purpose.” When the instruments measuring reality disagree with computer simulations predicting catastrophe, a sensible regulatory framework should follow the measurements. The Endangerment Finding followed the simulations.

3. Extreme Weather: The Dog That Didn’t Bark

3.1 The Big Claim

No claim has done more to build public support for climate regulation than the idea that greenhouse gas emissions are making extreme weather worse. The 2009 Endangerment Finding used this connection to justify regulatory action. Sixteen years later, the U.S. extreme weather record does not back it up.

3.2 U.S. Tornado Trends

The DOE report showed that strong tornado frequency (EF3+) in the United States has no statistically significant trend since reliable records began in the 1950s [DOE Climate Working Group, 2025]. The apparent increase in total tornado counts is simply an artefact of better detection—Doppler radar, more storm chasers, and smartphones—not a real meteorological trend [Brooks et al., 2014]. When you account for detection improvements, the data actually suggest a possible decline in the most violent tornadoes.

3.3 U.S. Hurricane Landfalls

The continental United States went twelve years (2005–2017) without a major hurricane landfall—the longest gap in the modern record. While individual intense storms have hit since then, the long-term frequency of U.S. landfalling hurricanes shows no upward trend [Klotzbach et al., 2022]. The IPCC AR6 itself admitted it has only “low confidence” in linking tropical cyclone frequency changes to human-caused warming [IPCC, 2021].

3.4 Floods and Droughts

Recent years have brought devastating floods to American communities. Hurricane Helene in September 2024 killed more than 250 people and destroyed the village of Chimney Rock, North Carolina, while cutting off Asheville’s water supply for 53 days. The Texas Hill Country flash floods of July 2025 killed at least 135 people, including 37 children, when the Guadalupe River rose 26 feet in 45 minutes—the deadliest inland flood in the U.S. since 1976. These are real tragedies that demand serious attention to flood preparedness and infrastructure.

However, individual catastrophic events—no matter how terrible—are not the same thing as a long-term trend. The USGS stream gauge network, spanning more than 8,000 stations across the contiguous United States, shows no systematic national increase in peak flood magnitudes. Lins and Slack (1999, updated 2005) found that while average streamflows have risen (the country is getting wetter), annual maximum flows—the measure that captures flood peaks—show no significant upward trend. Archfield et al. (2016) analysed 2,683 gauges over the period 1940–2013 and found only “fragmented patterns” with few statistically significant trends in any direction. The most recent USGS study (SIR 2023-5064), covering 1921–2020, found mixed regional trends but no uniform national increase.

Even the IPCC’s Sixth Assessment Report rates confidence in human influence on high river flows at the global scale as “low,” and explicitly notes that “heavier rainfall does not always lead to greater flooding” [IPCC, 2021]. Flood losses as a share of U.S. GDP have declined by roughly 75–90% since 1940, reflecting improved flood control infrastructure, better forecasting, and land-use management [Pielke, 2025]. As Pielke testified to Congress in September 2025: “It is misleading, and just plain incorrect, to claim that disasters associated with hurricanes, tornadoes, floods or droughts have increased on climate timescales either in the United States or globally.”

A fair reading of the evidence is this: some of the worst flood disasters in living memory have occurred in the past three years, yet the long-term stream gauge record does not show that floods as a class are getting more frequent or more severe across the United States. The Palmer Drought Severity Index for the contiguous United States shows no long-term trend toward worse droughts either.

3.5 Heatwaves

The 1930s Dust Bowl era is still the benchmark for extreme heat in the United States. Heatwave frequency and intensity in the U.S. Historical Climatology Network (USHCN) have not exceeded 1930s levels, which directly contradicts claims of “unprecedented” heat. While 2023 and 2024 had high global mean temperatures, these were heavily influenced by the 2023–2024 El Nino and the Hunga Tonga submarine volcanic eruption, which pumped record amounts of water vapour into the stratosphere [Millan et al., 2022].

3.6 The Bottom Line

The data are clear: U.S. extreme weather has not gotten worse in ways you can pin on rising CO2 over the past century. Claims that it has rely on attribution modelling—using the same climate models that overpredict warming to attribute individual events to human activity. This is circular reasoning: models that can’t accurately predict global temperature trends are used to blame specific weather events on CO2, and those attributions are then used as evidence that the models are right.

4. CO2: The Molecule They Want You to Fear

4.1 Plant Food, Not Pollution

Carbon dioxide is not a pollutant. It is an essential part of photosynthesis and the basis of the entire terrestrial food chain. The EPA classifying CO2 as an “air pollutant” under the Clean Air Act requires the legal fiction that a molecule every plant on Earth needs to survive is somehow a danger to public welfare.

Satellite data since the 1980s show a 14% increase in global green leaf area—an area equal to two times the continental United States [Zhu et al., 2016]. CO2 fertilisation was identified as the main driver, responsible for about 70% of the observed greening. Even the IPCC AR6 Working Group I confirmed that elevated CO2 “directly enhance plant growth, globally contributing to greening the planet” [IPCC WGI, 2021].

4.2 Agricultural Productivity

Global crop yields have gone up substantially during the period of rising CO2 levels. While other factors play a role (better crop varieties, fertiliser, irrigation), Free-Air CO2 Enrichment (FACE) experiments consistently show yield increases of 10–15% for C3 crops (wheat, rice, soybeans) and 5–8% for C4 crops (maize) under elevated CO2 [Ainsworth and Long, 2005]. These benefits were completely left out of the 2009 Endangerment Finding’s risk assessment.

4.3 Natural Carbon Sink

The land biosphere has absorbed about 30% of human CO2 emissions, with CO2 fertilisation boosting this natural carbon sink [IPCC WGI, 2021]. This is a self-correcting feedback that slows CO2 buildup in the atmosphere—a mechanism that climate models have historically underestimated, causing them to overpredict future CO2 levels for any given emissions scenario.

4.4 The Missing Cost-Benefit Analysis

The Social Cost of Carbon (SCC) calculations used to justify climate regulation assign zero value to CO2 fertilisation benefits. This is not a scientific decision—it is a bookkeeping choice that tilts the whole analysis toward finding net harm. A proper assessment that included the farming benefits, stronger forest growth, and the greening of dry regions would substantially reduce—and possibly reverse—the estimated net economic damage from CO2 emissions.

5. Sea Level Rise: Modest, Manageable, and Not Accelerating

5.1 The Long View

Global mean sea level has risen about 20 centimetres since 1900, a rate of roughly 1.5–2.0 mm per year [IPCC WGI, 2021]. This rise started well before significant human CO2 emissions and is consistent with the long-term recovery from the Little Ice Age, which ended in the mid-1800s. As Spencer (2010) has pointed out, “sea level has been rising naturally for thousands of years” since the last ice age.

5.2 Satellite Era Data

Satellite measurements since 1993 show global mean sea level rise of about 3.1 mm per year. Claims of “acceleration” depend heavily on which start and end dates you pick and what adjustments you apply to the satellite data. Nerem et al. (2018) reported an acceleration of 0.084 mm/year², but this number changes depending on whether you include or exclude the cooling effect of the 1991 Mount Pinatubo eruption at the start of the record and what Glacial Isostatic Adjustment (GIA) corrections you use. Houston and Dean (2011) found no statistically significant acceleration in tide gauge records covering the entire twentieth century.

5.3 Regional Factors

A lot of the coastal flooding blamed on “climate-driven sea level rise” in the media is actually caused by the land sinking—from groundwater pumping, sediment compaction, and geological processes. The U.S. East Coast, particularly around Chesapeake Bay, is sinking at 1–3 mm per year, and that has nothing to do with climate change. Mixing up these local geological factors with global climate change makes the problem look worse than it is.

5.4 Manageable Adaptation

At current rates, we are talking about millimetres per year—a pace that allows engineering solutions far cheaper than trying to eliminate fossil fuels worldwide. The Netherlands has managed sea levels for centuries. Calling modest, gradual sea level change an existential crisis that requires getting rid of fossil fuels is a policy choice, not a scientific conclusion.

6. Temperature and Human Health: The Warming Benefit

6.1 Cold Kills More Than Heat

The most comprehensive global study on temperature and death, Gasparrini et al. (2015), looked at 74 million deaths across 384 locations in 13 countries and found that cold kills roughly 17 times more people than heat. Cold weather caused 7.3% of all deaths studied, compared with 0.4% for heat. That ratio completely undermines the Endangerment Finding’s claim that warming threatens public health.

6.2 Warming Reduces Net Deaths

Modest warming cuts the bigger burden of cold-related deaths while slightly increasing the smaller burden of heat-related deaths. Zhao et al. (2021) projected that under moderate warming scenarios, fewer people would die from cold than the additional people dying from heat through at least mid-century—a net benefit. The 2009 Endangerment Finding ignored this completely.

6.3 Adaptation Works

Heat-related death rates have dropped dramatically in developed nations even as temperatures have risen. In the United States, heat deaths fell by roughly 80% between the 1960s and 2010s, thanks to air conditioning, better medical care, and heat warning systems [Bobb et al., 2014]. This shows that people adapt—something the Endangerment Finding’s static risk assessment failed to account for.

6.4 The Developing World Argument

Some argue that developing nations can’t adapt the way wealthy countries can. That is true—but the answer is economic development, and economic development needs affordable energy. Making fossil fuels more expensive slows development and actually reduces the ability to adapt. A climate policy that makes poor countries poorer to address a theoretical future risk while ignoring the real present benefit of affordable energy has it backwards.

7. The Track Record: Failed Predictions and Unfulfilled Warnings

7.1 The Pattern

Scientific credibility depends on getting predictions right. A theory that keeps overpredicting harm should have its assumptions questioned, not its conclusions amplified. The climate alarm camp has produced a long list of specific, dated predictions that simply did not come true. These are not minor misses—they come from the same models and reasoning behind the Endangerment Finding.

7.2 Ice-Free Arctic Summers

In 2009, Al Gore cited Naval Postgraduate School research claiming a “75 per cent chance” the Arctic would have completely ice-free summers within five to seven years [Gore, 2009]. Cambridge Professor Peter Wadhams predicted ice-free summers by 2015–2016, then quietly shifted it to 2020 [Wadhams, 2012]. As of 2026, Arctic sea ice is still there every summer. Ice extent has declined since 1979, but the predictions of complete disappearance were wrong by over a decade and counting. The 2012 minimum has not been exceeded since.

7.3 Himalayan Glaciers Gone by 2035

The IPCC AR4 Working Group II report claimed Himalayan glaciers could vanish by 2035 [IPCC WGII, 2007]. This turned out to be based on a World Wildlife Fund report that cited a 1999 New Scientist magazine article, which itself was based on an unpublished phone interview—not peer-reviewed science. The IPCC got the glacier area wrong by a factor of fifteen. In January 2010, the IPCC issued an unprecedented retraction, admitting “poor application of well-established IPCC procedures” [IPCC, 2010b]. So much for the rigorous review process.

7.4 Hansen’s 1988 Scenario A

James Hansen’s famous 1988 congressional testimony presented three warming scenarios. Scenario A, assuming continued growth in greenhouse gas emissions, projected about 0.5°C warming per decade. Actual CO2 emissions grew faster than Scenario A assumed (1.97% versus 1.5% per year), yet observed warming came in at roughly one-third to one-quarter of what Scenario A predicted [Christy, 2019]. More emissions, less warming than predicted—which points directly to climate sensitivity being much lower than Hansen’s model assumed.

7.5 The Missing Tropical Hot Spot

Climate models all predict that greenhouse warming should produce amplified warming in the tropical upper troposphere—the “tropical hot spot”—at two to three times the surface rate. This is not an optional extra; it is a direct result of the water vapour feedback that makes models so sensitive. Christy and McKitrick have shown through multiple analyses of satellite and weather balloon data that this amplification is either missing or greatly reduced in the observations [McKitrick and Christy, 2020]. All 73 CMIP5 models show more tropical tropospheric warming than what the three independent measurement systems (UAH, RSS, radiosondes) actually record. When a theory’s key prediction fails, the theory needs rethinking.

7.6 Fifty Million Climate Refugees by 2010

In 2005, the United Nations Environment Programme (UNEP) predicted 50 million climate refugees by 2010. The deadline came and went. Populations in the supposed “danger zones”—the Bahamas, St. Lucia, Seychelles, Solomon Islands—had actually grown. UNEP quietly removed the prediction from its website and denied it was ever a “UNEP prediction” [Watts, 2011].

7.7 The Great Barrier Reef

After the 2016 mass bleaching event killed 30% of coral, media and scientists declared the reef was in an irreversible death spiral, with some predicting 70–90% coral loss from another 0.5°C of warming. Then the Australian Institute of Marine Science’s 2022 survey found hard coral cover had bounced back to record-high levels across two-thirds of the reef [AIMS, 2022]. Coral bleaching is a stress response, not death, and tropical corals have recovery abilities that doom-and-gloom projections consistently ignore.

7.8 Polar Bear Numbers

Polar bears became the poster species for climate alarm, with predictions of dramatic population collapse from Arctic ice loss. The IUCN estimated 20,000–25,000 bears in 2005. Current numbers range from 22,000 to 31,000, with most studied populations stable or increasing [Crockford, 2019]. It turns out polar bears adapted to less summer ice by changing their hunting habits—a flexibility the decline models did not account for.

7.9 Snow as “A Thing of the Past”

In March 2000, David Viner of the Climatic Research Unit at the University of East Anglia told The Independent that “children just aren’t going to know what snow is” [Viner, 2000]. Two weeks later, Luton Airport was shut down by heavy snowfall. The Northern Hemisphere has had multiple record-breaking snow events since, and satellite data show no declining trend in winter snowfall extent.

7.10 What This Tells Us

Every one of these failed predictions came from the same modelling approach and institutional consensus that produced the Endangerment Finding. They all overpredict harm. They all err in the same direction. That is not random noise around a correct answer—it is systematic bias.

8. What the IPCC Process Actually Is

8.1 Not an Independent Scientific Body

The IPCC does not do original research. It is an intergovernmental panel—a political organisation whose members are government representatives, not independent scientists. Its assessment reports come with “Summary for Policymakers” documents that are negotiated line-by-line by government delegates—a process that puts political agreement ahead of scientific accuracy.

8.2 The Consensus Problem

The scientific “consensus” used to shut down sceptical research is itself shaped by institutional incentives. Funding bodies favour research that identifies climate risks, creating a publication bias. Scientists who question the catastrophe narrative face real consequences—less funding, trouble getting published, and public attacks—creating a chilling effect on honest scientific debate. Spencer, Christy, Curry, and Lindzen have all spoken about these pressures from personal experience.

8.3 The Kopp Rebuttal in Context

The Kopp et al. (2025) response to the DOE report was thrown together in weeks and presented as the verdict of “85 scientists.” But a rushed rebuttal is not the same as the kind of careful, years-long assessment you would need to properly address the DOE report’s findings. The Kopp rebuttal mostly just restated model projections and waved away the observational problems rather than explaining them. The number of names on a document is not a scientific argument.

9. Discussion

9.1 The Basic Question

The question for policymakers is simple: does the real-world evidence support the claim that greenhouse gas emissions are dangerous enough to public health and welfare to justify the massive economic costs of emissions regulation?

Based on sixteen years of data since the 2009 finding, the answer is no.

9.2 What the Data Actually Show

• Climate sensitivity is likely 1.0–2.5°C per doubling of CO2—well below what the models assumed when they informed the Endangerment Finding
• Extreme weather in the United States has not gotten worse in ways linked to CO2 over the past century
• CO2 fertilisation is greening the planet, boosting crop yields, and strengthening the natural carbon sink—benefits completely excluded from the regulatory analysis
• Sea level rise is modest, gradual, and fits long-term natural trends rather than model-predicted acceleration
• Net deaths from temperature changes are likely going down, because reduced cold deaths far outweigh increased heat deaths

9.3 The Precautionary Principle

Supporters of the Endangerment Finding point to the precautionary principle: even if risks are uncertain, we should act to prevent possible catastrophe. But the precautionary principle works both ways. Aggressive decarbonisation imposes certain, immediate costs—higher energy prices, less competitive industry, slower economic growth—against uncertain and possibly small future benefits. A truly cautious approach would weigh the definite costs of acting against the uncertain costs of not acting, rather than just assuming the worst-case climate scenario while ignoring the price tag of the proposed solution.

9.4 U.S. Policy Makes No Difference Anyway

The DOE Climate Working Group pointed out that “U.S. policy actions are expected to have undetectably small direct impacts on the global climate” [DOE Climate Working Group, 2025]. Even if the catastrophic projections were right, the U.S. cutting emissions while China and India keep expanding fossil fuel use would make no measurable difference to global temperatures. This is not a political argument; it is basic maths.

10. Conclusion

The 2009 EPA Endangerment Finding was built on climate model projections that sixteen years of real-world data have shown to be wrong. The models overpredict warming. The predicted worsening of extreme weather has not happened. The benefits of CO2 for farming and ecosystems were left out of the analysis. Sea level rise remains modest and manageable. Warming saves more lives from cold than it costs from heat. And a long list of specific climate predictions—ice-free Arctic, vanishing Himalayan glaciers, 50 million refugees, the death of the Great Barrier Reef—have simply not come true, showing a clear pattern of overprediction from the institutions that built the case for the original finding.

The DOE Climate Working Group’s review, by five scientists with decades of hands-on experience measuring the atmosphere, presents a solid evidence-based case for reconsideration that the Kopp et al. rebuttal does not successfully counter. The core weakness of the mainstream position is its dependence on model projections when the actual observations tell a different story. As Richard Feynman said: “It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it disagrees with experiment, it’s wrong.”

Reconsidering and repealing the 2009 Endangerment Finding is not an attack on science. It is a return to the scientific method—the principle that policy should be based on what the data actually show, not on what computer models predict they should show. The data support this reconsideration. The regulatory framework should follow.