This year marks the 100th anniversary of the birth of my mentor, climatologist Mikhail Budyko (1920–2001). Fifty years ago, when the science of climate change prediction was in its infancy, this scientist from the Soviet Union made a series of climate predictions that have proven to be surprisingly on target in the years since.
These predictions were not as well known as some of his other work. The reason for this had to do, as Budyko recalled in a 1990 interview, with most of his colleagues at the time rejecting the idea of unavoidable, long-term global warming as something absolutely impossible. It took almost 20 years for the scientific establishment to accept this paradigm.
Here I summarize the methods behind Budyko’s predictions and demonstrate why his forecast was an important step in the development of contemporary climate science and why it should be the basis of a new “business-as-usual” scenario of global warming, characterizing changes on Earth that may result if we do not make any additional efforts to mitigate or reverse climate change.
A Prescient Look at Today’s Climate
Budyko is better known for his work in several other areas of study. In 1969, he authored the first global energy balance model describing “snowball Earth” conditions, in which nearly the entire planet freezes over [Budyko, 1969]. (This actually happened on Earth during a period that ended about 635 million years ago, when complex life was just starting to develop.) He also developed the “Budyko curve” for modeling plant evapotranspiration as a function of how wet or dry the environment is. And his “Budyko blanket” idea for artificial solar radiation management, published in 1974, imagined the use of sulfate aerosol particles injected into Earth’s stratosphere to control global warming.
In 1972, before the mean global temperature began the sharp, steady rise seen in the half century since, Budyko published a lesser known climate forecast extending 100 years into the future. He predicted that Earth’s mean global temperature would increase about 2.25°C by 2070 and that the Arctic would no longer be covered by ice year-round by 2050 [Budyko, 1972]. (Budyko briefly discussed the part of his forecast dealing with Arctic ice in a 1972 Eos article cited more than 100 times since.) Despite his confidence in his work, he cautioned that his estimates were made under assumptions of a significantly simplified climate system and should be viewed accordingly [Budyko, 1972], so it might have surprised him to see how closely actual events aligned with his predictions.
Comparing 2019 to 1970, Budyko predicted an increase in the global mean temperature of 1°C and the disappearance of about 50% of Arctic multiyear ice. Observations have borne out these trends, demonstrating that mean global temperature increased by 0.98°C over this period and that the extent of multiyear Arctic sea ice in September 2019 was about 46% smaller than in 1970 (Figure 1).
The accuracy of these predictions is especially fascinating in light of the prevailing uncertainty in modern, complex global circulation models [Zelinka et al., 2020]. Even though these newer models incorporate more complexity, until about 2009, most models of Arctic sea ice dynamics consistently underestimated the actual rate at which the Arctic lost ice over the past several decades.
Predictions for the New Millennium
Budyko’s temperature forecast was based on several works, including his own global energy balance model [Budyko, 1969], that were presented at a 1970 symposium about the Study of Man’s Impact on Climate (SMIC) [Matthews et al., 1971]. He cited a global carbon cycle model [Machta, 1972] that consisted of two atmospheric and two oceanic layers as well as “fast”—and “slow”—cycling pools of terrestrial carbon and a pool of carbon from marine biota. The most crucial parameter in this model was the rate of carbon exchange between the ocean and atmosphere. Machta derived this parameter from 1955–1969 observations of the fate of “bomb” radiocarbon—isotopes linked to nuclear weapons tests—and predicted that in 2000, the concentration of carbon dioxide in the atmosphere would reach 375 ± 10 parts per million (ppm). The actual concentration, measured at Mauna Loa on Hawaii and averaged for 2000, was 369 ppm.
Budyko also cited Manabe’s global circulation model that used a simplified land topography, a “swamp” ocean without horizontal or vertical heat transport, and no sea ice dynamics [Manabe, 1971]. Despite these simplifications, the model accounted for several major atmospheric feedbacks that affect radiative forcing, including those due to atmospheric moisture and snow cover. At a fixed relative humidity, this model demonstrated a climate sensitivity of about 2°C for every doubling of the carbon dioxide concentration and amplified warming at high latitudes by a factor of 2. Recent research has shown that the actual transient sensitivity of the climate system to radiative forcing (the excess heat energy that Earth absorbs from the Sun compared with how much it radiates back into space) is about 1.8°C per doubling of carbon dioxide [Nijsse et al., 2020] and that Arctic air temperatures are rising at more than double the global mean rate.
The same climate sensitivity and prediction for the atmospheric concentration of carbon dioxide in 2000 allowed the authors of the SMIC report to conclude that by 2000, Earth’s surface would warm by about 0.5°C [Matthews et al., 1971]. This prediction was based on only radiative forcing caused by changing carbon dioxide levels. It did not include forcing from methane, nitrous oxide, and ozone, yet it turned out to be correct (Figure 1a). One possible explanation for this accuracy is that the prediction also did not include the cooling effect (negative radiative forcing) produced by aerosol particles in the atmosphere, which could have compensated for the heating effect of these other gases.
A 100-Year Forecast
In 1972, Budyko went further than these earlier efforts had and published his global warming forecast out to 2070 [Budyko, 1972]. To an extent, his forecast was an extension of the SMIC prediction, but it envisioned a different rate of warming. He estimated this new rate with simple calculations that assumed a linear relationship between the annual growth rate of global energy production and Earth’s surface temperature. Budyko suggested an annual increase in the rate of global primary energy consumption by a factor of 1.5, from about 4% in the early 1970s to 6% after 2000, resulting in a calculated rate of global warming of 0.25°C per decade after 2000, rather than the rate of 0.5°C over 30 years (i.e., roughly 0.17°C per decade) predicted in the SMIC report [Budyko, 1972]. Budyko did not explain his reasoning for assuming a 1.5-fold increase in energy consumption, but he wrote that the energy consumption rate would unavoidably increase, possibly reaching 10% per year sometime during 21st century [Budyko, 1972].
From 1970 to 1999, the mean global temperature indeed increased at 0.17°C per decade, rising to 0.25°C per decade after 2000 [GISTEMP Team, 2020] (Figure 1a). During the past several decades, however, annual primary energy consumption has been growing only at about 2.9% per year [BP, 2019]. Apparently, Budyko overestimated the relative growth rate of primary energy consumption but correctly guessed the proportion of increase in the absolute rate at which energy was consumed, as well as the corresponding increase in temperature. From 1970 to 1999, the absolute growth of energy consumption was 15 petawatt-hours (PWh) per decade, and from 2000 to 2019, it was 25 PWh per decade [BP, 2019]. During these periods, the ratio of the temperature change to the increase in primary energy consumption was nearly the same, 0.011°C/PWh for 1970–1999 and 0.010°C/PWh for 2000–2019.
Why the Linear Relationship Works
This nearly linear response of global temperature to increases in energy consumption can be explained by several linear relations within the energy-carbon-climate system. First, because about 87% of global energy demand is still satisfied by fossil fuels [BP, 2019], the linear link between energy consumption and carbon emissions still stands, despite increases in alternative energy usage during the past 50 years.
Second, the fraction of emitted carbon that remains in the atmosphere held nearly constant at about 45% throughout the 20th century. Third, the logarithmic relation between radiative forcing and atmospheric carbon dioxide concentrations is reasonably approximated by a linear function for carbon dioxide levels between 320 and 580 ppm—the difference between linear and logarithmic approximations of radiative forcing is 10% at most.
Fourth, until recently, the main contributor to increases in global radiative forcing was rising atmospheric carbon dioxide and related climate feedbacks. And fifth, Budyko’s energy balance model shows a linear response of mean global temperature to small (up to 1%) deviations in radiation balance [Budyko, 1969]. Budyko’s suggestions about linearity between energy consumption and global warming thus turned out to be quite accurate.
Budyko believed, however, that his 1972 forecast might underestimate warming trends. He was concerned about the absence of a positive feedback between rising temperatures and decreasing polar ice cover in Manabe’s  original estimate of climate sensitivity. Also, he expected to see a greater contribution to warming from heat generated directly from fossil fuel combustion (distinct from the greenhouse effect it produces). This heat was minimal in 1970 and remains quite low today, accounting for less than 0.1 watt per square meter [BP, 2019], compared with contemporary mean global radiative forcing of 2.3 watts per square meter.
About 20 years after Budyko published his 1972 forecast, two similar scenarios, called A (high emission) and B (medium emission), were considered in the first report of the Intergovernmental Panel on Climate Change (IPCC). These scenarios, now called Representative Concentration Pathway (RCP) scenarios RCP 8.5 and RCP 6.0, respectively, still exist in more recent IPCC reports. Scenario A, referred to as “business as usual,” suggested a coal-intensive energy supply, continuing deforestation, unregulated agricultural emissions of methane and nitrous oxide, and only partial implementation of the Montreal Protocol, which regulated emissions of stratospheric ozone-depleting materials. It predicted an increase in the mean global temperature between 0.2°C and 0.5°C per decade (with an average of 0.3°C per decade). Scenario B was characterized by a shift toward more use of natural gas, large increases in energy efficiency, stringent carbon monoxide control, a reverse trend in deforestation, and full implementation of the Montreal Protocol. Therefore, scenario B predicted a lower increase in mean global surface temperature at 0.2°C per decade (Figure 1a).
A Model of Arctic Ice
Budyko based his forecast for perennial Arctic sea ice (ice that does not melt every year) on a semiempirical energy balance model of sea ice freezing and melting cycles (Figure 1b) [Budyko, 1966]. He assumed that the cumulative annual drift of floating ice from the Arctic through the Fram and Bering Straits, one of the most important parameters, would be constant over time. And he estimated that the equivalent volume of liquid water in this drifting ice would be about 2,000 cubic kilometers per year [Budyko, 1966]. Models of ice dynamics and satellite images demonstrate that during 2010–2016, annual export of ice through the Fram Strait east of Greenland varied from 1,970 to 2,400 cubic kilometers [Min et al., 2019] (export through the Bering Strait was more than an order of magnitude smaller and can be neglected in this estimate). Therefore, the net discharge of ice from the Arctic was about 2,200 cubic kilometers, a volume only about 10% higher than Budyko’s 1972 estimate.
Budyko recognized that temperatures were rising faster in the Arctic than at lower latitudes and that this polar amplification would speed up melting of Arctic ice. He calculated that with a 4°C Arctic temperature anomaly relative to 1970, a 4-meter-thick ice layer in the central Arctic should disappear in about 4 years [Budyko, 1966]. And with polar amplification by a factor of 2, implying Arctic warming at twice the global mean, he found that a 4°C anomaly in the Arctic should be reached by 2050–2060, when the mean global temperature would be 2oC higher than in 1970 (Figure 1a). Today’s climate models forced by RCP 8.5 and RCP 6.0 scenarios predict that the first ice-free summer will likely occur between 2042 and 2054 [Peng et al., 2020].
A Straightforward Vision of Business as Usual
Today, anthropogenic carbon emissions remain high. To avoid warming of 1.5° by 2060 (relative to the preindustrial level), global emissions must be reduced by 7% per year starting immediately [Höhne et al., 2020]. For reference, the COVID-19 pandemic and resulting lockdowns are projected to produce a temporary 4% to 7% decline in annual carbon emissions in 2020 [Le Quéré et al., 2020]. Previous crises, such as the global financial crisis of 2008–2009 and the early 1970s oil crisis, have also reduced carbon emissions temporarily, but emissions have always bounced back after these events. And the same is likely to hold true following the current pandemic, as low oil prices and economic recovery efforts spur more consumption.
Regardless of such short-term dips and spikes in emissions, even if we stopped burning all fossil fuels today, Earth would continue to warm by a few tenths of a degree per century for a century or more because of stored thermal energy in the ocean and because of the reduced cooling effect of aerosols released by fossil fuel combustion. The actual temperature increase, however, will likely be larger than this residual warming from past emissions because of other factors. For example, early IPCC scenarios did not consider emerging methane sources, such as methane seepage from warming Arctic Ocean bottom sediments and methane released by abrupt permafrost thaw.
Underlying the IPCC’s business-as-usual scenario is the absence of any political or economic actions to control emissions. During the past 30 years, however, the world has become more interconnected and aware of global warming’s gloomy consequences, and it has taken several practical steps to reduce carbon emissions. Therefore, it was recently proposed to treat the modern business-as-usual RCP 8.5 as the worst-case scenario [Hausfather and Peters, 2020]. Yet adding natural methane emissions from the Arctic, say, or the potentially rapid increase of carbon emissions supported by cheap oil and other fossil fuels during the post-COVID-19 era could produce an even worse scenario.
Looking back 20 years, the mean global temperature continued to rise by 0.25°C per decade, coinciding with Budyko’s projection. Moving forward, temperature trends might worsen if the world follows the worst-case trajectory, or they might improve somewhat if pledged policies and future technologies allow for intensive decarbonization of the economy. The simplest projection, however, is that observed trends will continue for at least the near future, just as Budyko predicted in 1972.