Climate models may be underestimating future warming by a quarter
A new evaluation of climate models indicates that for a given level of carbon dioxide, the planet could warm 25% more than previously thought. This would mean that existing policy-based projections, such as a 2.6°C rise by 2100, might actually be closer to 3.25°C. The study suggests that models previously dismissed as too sensitive may be more accurate than assumed.
The study’s method hinges on filtering out natural climate variability, such as La Niña’s cooling effect, which masked warming between 1981 and 2014. That period was previously used to judge model accuracy, unfairly penalizing models that projected stronger responses to CO2. By removing this bias, the team found those “hot” models align better with reality. Additionally, satellite records of incoming and outgoing radiation—available only since 2001—now offer a more direct test of a model’s energy balance. Models matching these radiation trends consistently predict higher equilibrium warming, suggesting the IPCC’s central estimate of 1.8°C per CO2 doubling may be too low.
If correct, this finding could reshape climate risk assessments for governments, insurers, and infrastructure planners. A 25% upward revision means adaptation budgets and emissions targets based on current projections may fall short, potentially leaving coastal cities and agricultural systems more exposed than planned. It may also intensify public pressure for faster decarbonization, though the uncertainty inherent in model evaluation means policymakers should treat the higher range as a plausible scenario rather than a certainty.