The standard way to measure cooling demand, counting how many degrees the temperature rises above a baseline, then summing those degrees over a season, has a fundamental blind spot: it treats a 30-degree humid day the same as a 30-degree dry day. A new study in Nature Communications shows that blind spot matters, and that it has been getting worse.
The paper, by Jake W. Casselman and Christina Karamperidou at the University of Hawaii at Manoa, introduces the “efficiency-weighted cooling degree day” (eCDD). Unlike the conventional cooling degree day (CDD), eCDD accounts for the fact that air conditioner efficiency depends on both temperature and humidity. When it is hot and humid, the compressor has to work harder to reject heat to a hotter environment, and the additional latent heat from moisture in the air adds an extra burden.
The metric
Conventional CDD is calculated by taking the positive difference between daily mean temperature and a base temperature, typically 18 degrees Celsius (65 degrees Fahrenheit), and summing those differences across a season. It assumes that every degree of warming demands the same amount of energy for cooling, regardless of how humid it is or how hard the air conditioner has to work.
eCDD replaces that assumption with a physically grounded alternative. It links both ambient temperature and humidity to the thermodynamic work required for cooling, using the coefficient of performance (COP) of refrigeration, a measure of how efficiently a heat pump or air conditioner can move heat. COP degrades as the temperature difference between indoors and outdoors increases, and humidity adds a latent load that does not appear in temperature-only metrics.
The result is that eCDD can differ substantially from CDD, and the difference is growing.
What the numbers show
Applying eCDD to historical climate data across North America, the researchers found that cooling efficiency, the amount of cooling delivered per degree day, has declined by 2% to 4% per decade since 1971. The decline is driven by trends in humidity that conventional CDD does not capture.
Under future climate scenarios, the study projects that eCDD will increase by 10% to 80% across different regions by the end of the century, depending on emissions pathway and local humidity regime shifts. The wide range reflects a key finding: temperature and humidity trends are spatially opposing across North America. Some regions are getting hotter and more humid, compounding the efficiency problem. Others are getting hotter but drier, partially offsetting the efficiency loss.
In fact, the study identifies a projected eastward extension of dry heat across the continent, which could enhance cooling efficiency during hot extremes in historically more humid regions. But this is a net negative for overall cooling demand because temperatures are rising everywhere.
Why this matters for energy planning
The practical implication is that temperature-only metrics misrepresent cooling demand, and the misrepresentation is systematic. Regions with humid heatwaves stress the grid far more than temperature alone predicts. For utilities and grid operators planning capacity, reserve margins, and transmission investments, the difference matters.
The same applies to emissions accounting. Regions with fossil-fuel-heavy electricity grids may see disproportionately higher emissions from cooling, because the actual energy demand per degree day is higher under humid conditions than CDD suggests. The emissions impact of cooling is being underestimated in any planning framework that uses conventional CDD.
The researchers also note that building codes and appliance standards could benefit from eCDD-based design criteria, and that climate adaptation investments, in air conditioning deployment, grid upgrades, and heatwave preparedness, would produce more accurate demand projections if they accounted for the humidity-efficiency coupling.
The underlying gridded analysis outputs are available on Zenodo.
Sources
[1] Casselman, J.W. & Karamperidou, C. “Efficiency-weighted cooling degree days reveal opposing temperature and humidity effects on energy demand.” Nature Communications (2026). DOI: 10.1038/s41467-026-75388-9