Power systems with the lowest annual fossil fuel shares experience the sharpest spikes in fossil generation during temperature extremes, according to a study published in Nature on October 7, 2026. Researchers from Peking University, Tsinghua University, and Stanford examined 22 U.S. balancing authorities and found that daily fossil generation averages 43% higher on extreme hot days and 19% higher on extreme cold days compared with mild days. The amplification is most pronounced in grids that appear cleanest by annual measures, where daily carbon dioxide and air pollutant emissions on hot days can more than double.
Across the 22 balancing authorities studied, mean daily electricity generation rises 28% on extreme hot days and 13% on extreme cold days relative to mild days, with fossil fuels supplying a disproportionate share of the increase. The fossil share of generation climbs from 62% on mild days to 69% on extreme hot days and 65% on extreme cold days, driven primarily by natural gas during heat waves. Wind and solar respond inconsistently—sometimes increasing, sometimes declining—and total non-fossil generation varies relatively little in most systems. The relative increase in fossil generation on extreme hot days grows systematically as a balancing authority's annual fossil share declines, with a correlation coefficient of −0.77. In systems where non-fossil generation exceeds 60% of annual output, daily CO₂ and air pollutant emissions on hot days and the associated premature deaths can more than double compared with mild days, with emission intensities and mortality rates rising by more than 50%.
The report finds that "extreme temperatures may therefore do more than raise demand: because thermal generation remains an important source of flexibility for meeting short-lived peaks, these events may deepen the very fossil dependence that the transition is meant to end." The authors identify a structural mechanism: as variable renewables claim a larger share of the annual mix, thermal units shift from steady operation toward flexible balancing, and demand rises while renewable generation falls on nearly a third of days in the year. According to the researchers, "a power system is not decarbonized because its annual fossil share is low; it is decarbonized when it can withstand climatic stress reliably and cleanly."
The balancing role of thermal units creates a capacity-utilization tension that compounds the problem. Meeting short-lived peaks requires capacity that sits idle much of the year, because reliability depends on sufficient dispatchable capability to cover episodic spikes. In wind-solar-large and non-fossil-dominant systems, maximum daily fossil generation reaches roughly twice the annual mean, compared with about 1.5 times in fossil-dominant grids. Even where fossil generation supplies only about 40% of annual electricity, fossil units may still be called on to provide close to 80% of average daily demand during peak periods. The peak-to-average ratio rises systematically as annual fossil share declines, implying fossil utilization rates below 50% in some non-fossil-dominant systems and raising the prospect of substantial standby assets and economic losses as more frequent and intense extremes raise peak demand.
The study concludes that decarbonization should be planned and judged against peak conditions rather than annual averages. Energy storage, demand-side flexibility, and other forms of dispatchable low-carbon generation must be deployed at a pace that keeps up with renewable expansion, while carbon capture can complement any remaining fossil capacity, so that the flexibility now provided by thermal units is progressively replaced rather than merely retained. The implication isn't that fossil generation should be preserved as the default source of flexibility, but that cutting annual fossil generation doesn't by itself remove a system's dependence on fossil-based flexibility—and as long as that dependence persists, so do the associated emissions and health burdens.