The cooling demand hidden in climate forecasts
Climate change is not only changing how much energy the world must produce. It is also changing when and where people need that energy. As summers become hotter, buildings require more electricity for cooling, while milder winters reduce demand for heating. A 2026 study in Nature Communications suggests that widely used climate and energy scenarios can miss a substantial part of this shift when they keep the historical climate fixed in their calculations.
Researchers led by Mengting Zhu found that projected cooling demand could be underestimated by as much as 23% in lower-warming scenarios and 79% in higher-warming scenarios by 2100. Their modelling also indicated that heating demand could be overestimated by up to 14% and 40%, respectively. Those are scenario-dependent maximum differences, not forecasts that every country will experience the same change.
The findings matter because decisions about electricity generation, transmission networks, building design and climate policy depend on the demand figures that planners expect decades ahead. A scenario that misses rising air-conditioning demand could make a future power system appear easier to supply than it will actually be.
Why the modelling baseline matters
Energy system models often separate the effects of human choices from the effects of climate. To do that, some established scenario frameworks calculate future building energy needs against historical heating and cooling conditions. This helps isolate changes caused by population, economic growth, technology and policy. But it also creates a blind spot when the climate itself is moving.
Heating degree days and cooling degree days are ways to express how far outdoor temperatures sit below or above a reference point associated with indoor comfort. They do not directly count electricity units consumed, but they help estimate the energy needed to keep buildings at usable temperatures. As temperatures rise, cooling degree days generally increase and heating degree days generally decrease.
The researchers updated those climate-sensitive inputs using temperature projections from the sixth phase of the Coupled Model Intercomparison Project, known as CMIP6. They then incorporated the revised heating and cooling degree days into the Global Change Analysis Model, an integrated assessment model used to explore energy, economic and emissions pathways.
By comparing results with and without the warming-sensitive building demand, the team could identify how much of the difference arose from accounting for a changing climate rather than holding the weather baseline constant. The approach examined several combinations of socioeconomic development and greenhouse-gas trajectories, so the result is not dependent on one imagined future.
The biggest numerical differences appear in cooling
The study reports that ignoring future warming can underestimate cooling demand by up to 23% in scenarios with relatively limited warming. In higher-warming scenarios, the difference can reach 79% by the end of the century. A nearly four-fifths gap in an estimate is consequential for any infrastructure plan built around that estimate, although it should not be confused with a prediction that total electricity demand everywhere rises by 79%.
Heating moves in the opposite direction. Without updated climate inputs, models can overestimate heating needs by up to 14% in lower-warming cases and 40% in higher-warming cases. The two shifts do not simply cancel one another out because heating and cooling use different equipment, fuels and seasonal operating patterns.
An additional unit of cooling demand may fall on an electricity system during hot afternoons, precisely when many buildings are trying to cool simultaneously. A reduction in winter heating demand might instead lower natural gas use, district heating or electricity consumption depending on local technologies. The consequences therefore depend on what powers the buildings, not only on how many units of thermal comfort are required.
The research finds important geographic variation. The United States, China, Southeast Asia and the Middle East are among the places where the modelling highlights meaningful changes, but their exposure, existing building stocks and energy systems differ. The same global average can conceal very different local infrastructure problems.
A surprising implication for carbon emissions
Changing building energy needs also changes the emissions pathway. Across the scenarios studied, including warming-sensitive demand produced estimated carbon dioxide emissions in 2100 that were 83 million to 1,600 million tonnes lower per year than estimates produced without those adjustments. That finding might initially sound counterintuitive: why would hotter weather, and more cooling, coincide with lower modelled carbon dioxide emissions?
The answer is that the result comes from a system-wide model rather than a direct equation between air conditioning and emissions. Less heating can reduce fuel combustion, while cooling demand, electrification and the changing mix of electricity generation interact in complex ways. The net carbon dioxide outcome depends on those combined assumptions and is not a claim that climate change is beneficial.
The study also points to a separate emissions concern. Cooling equipment can use refrigerants containing fluorinated gases, some of which are powerful greenhouse gases. If the models underestimate future cooling needs, they can also underestimate the associated rise in these F-gas emissions, especially in warmer developing regions.
This distinction is important for policy. Carbon dioxide from energy use and emissions from refrigerant leakage are different accounting categories. A scenario showing lower carbon dioxide emissions from heating and power generation can still miss a significant refrigerant-related climate burden. Better cooling efficiency and refrigerant management therefore matter alongside cleaner electricity.
Why this is relevant beyond climate science
Electricity systems are built for peaks as well as annual totals. A utility may generate enough energy over an entire year but still struggle to serve concentrated demand during a heatwave. Greater cooling use can raise peak loads, accelerate equipment wear and create reliability challenges in places where grids are already strained.
Building owners face a related planning problem. A structure designed for past temperature patterns may require retrofits, better insulation, shading, ventilation or more efficient cooling equipment as conditions change. The study does not prescribe one building technology, but it supports using forward-looking climate information rather than assuming historical weather remains representative.
Developing economies face a particularly difficult balance. Access to safe indoor temperatures can protect health and productivity, yet expanding mechanical cooling increases electricity needs and can worsen emissions if power supplies and refrigerants are not managed carefully. The answer is not to suppress necessary cooling; it is to plan for it accurately and provide it more efficiently.
For South Africa, the paper offers a planning lesson rather than a South Africa-specific numerical forecast. Hotter conditions, changing patterns of electricity use and uneven access to cooling make climate-sensitive demand modelling relevant to utilities, municipalities and property developers. The published global results should not be presented as precise estimates for Eskom or for any individual South African city.
Related Research Today reporting has examined the economics of geothermal heating in Europe and how environmental conditions can alter ecosystem benefits. Both research questions illustrate why assumptions about the future environment can change the conclusions of models used to guide decisions.
What the study can and cannot establish
This is a modelling study, not a field experiment measuring the electricity bills of individual households. Its numerical estimates reflect climate projections, model structure, assumptions about economic and technological development, and the relationship between temperature and energy demand. Different assumptions could produce different results.
The authors also distinguish global patterns from regional outcomes. A 79% underestimation in a high-warming scenario is an upper-end modelled discrepancy, not a universal effect. Some places will have much greater cooling exposure than others, while building efficiency, income, adaptation and technology adoption will influence the actual demand that emerges.
Cooling degree days are useful but incomplete proxies for comfort-related energy use. Humidity, building materials, insulation, household incomes, air-conditioning ownership, behavioural choices and electricity prices also affect consumption. Climate-sensitive inputs improve the model without eliminating every source of uncertainty.
The results also extend to 2100, a horizon over which technologies, building standards and energy policy could change substantially. They are best interpreted as evidence about the importance of including climate feedbacks in planning scenarios, not as a detailed timetable for any specific utility’s future investments.
One further caution is essential: a modelled reduction in net carbon dioxide emissions after accounting for lower heating needs does not mean warming should be welcomed. The paper focuses on energy-system accounting and does not subtract the broader human, ecological and economic harms caused by climate change.
The planning lesson: future weather cannot be treated as yesterday’s
The central insight is simple but consequential. An energy forecast can be internally consistent and still miss a major driver of future demand if it assumes that the climate remains unchanged. By updating temperature-sensitive building demand, the researchers exposed substantial differences in cooling, heating and emissions estimates.
Related research also considers how green technology trade relates to environmental performance. Better planning would connect climate projections with building energy models, electricity network forecasts, efficiency standards and refrigerant policies. It would also examine regional differences and the timing of peak loads rather than relying only on annual global totals.
The study does not offer a single number that every government or utility should plug into its budget. It offers a warning about the assumptions underneath those numbers. As the world warms, the energy required to keep buildings habitable is changing too. Models that fail to represent that change risk guiding investment with an increasingly outdated picture of demand.
Source Information
Study: Warming-driven shifts in global building energy use reshape climate mitigation planning
Authors: Mengting Zhu, Mengqi Zhao, Rongqi Zhu and colleagues
Journal: Nature Communications, volume 17, article 6181 (2026)
Method: CMIP6 climate projections incorporated into the Global Change Analysis Model across socioeconomic and climate scenarios
