How will our warming climate stabilize? Scientists look to the distant past

Thanks to unbridled greenhouse gas emissions, our planet is stitching together a climate version of Dr. Frankenstein's monster. We still have ice from the warmer parts of the Pleistocene even as our temperature approaches the warmer Pliocene levels of 3 million years ago. Meanwhile, our CO₂ level is between the Pliocene and the Miocene of 10 million years ago, and we risk an Eocene hothouse not seen in 40 million years. At some point, this unnatural fusion of incongruous climate parts must resolve into a new equilibrium—but at what point? And what does that equilibrium look like? Much of that is up to us, based on how fast we reach net-zero greenhouse gas emissions. But it’s also up to our planet—how “sensitive” it is to greenhouse gases and how quickly it reacts to changes. Discovering our planet's sensitivity to greenhouse gases has been a "holy grail" for scientists since the 1970s, but it has stubbornly resisted attempts to constrain it. The best we can do is a wide range: 1.5° to 4.5° C of warming if CO₂ levels double. That’s a huge temperature range, and we’re likely to double preindustrial CO₂ levels this century even as we want to avoid warming above 2° C. Narrowing this range will be key to understanding what our Frankenstein-like climate will look like when it settles into a new equilibrium.

We have left the Holocene behind

Part of the uncertainty is because our instrument records only capture a short bit of the cool climate we've left behind. By 2100, we’re on track for global temperatures between ~2.7° C and 3.6° C warmer than the preindustrial era. This is warmer than the entire Holocene, the geological period since the last glaciation in which human society with agriculture, cities, and industry flourished. “We’re talking about a warming that is on par with the Pliocene,” Dr. Jessica Tierney of University of Arizona told me, “but at CO₂ levels that… would take us back farther in time to… probably [the] Miocene warm period.” In the Pliocene, our ancestors were Australopiths not yet fully committed to life out of trees. Most of the ice currently on Greenland and West Antarctica was not there, and sea levels were between five and 25 meters higher than in 1900. In the Miocene, our ancestors were still apes, sea levels were perhaps 48 meters higher, and parts of Antarctica were lushly forested. “Ancient climates are our only context for what a warm world looks like,” said Tierney. “We ask our models to simulate it for us, but if we want to know what actually happens in a high-CO₂ world, we have to look to the past for the examples.”

Dipping a toe in ancient seas

To do that, you first have to measure those ancient temperatures and CO₂ levels. Since there were no thermometers or infrared spectroscopes back then, scientists use indirect measurements, or “proxies,” items that change their chemical, isotopic, or physical makeup in proportion to changes in temperature, CO₂ levels, or even rainfall. There’s a cornucopia of such proxies to choose from, ranging from the banal (various element and isotope ratios) to the bizarre (packrat urine, leaf wax, moth scales, and leaf pores). Each proxy is calibrated to enable conversion into the relevant climate value, like temperature, CO₂, rainfall regime, and so on. That calibration is often far from straightforward, so proxies were generally considered too uncertain and noisy to constrain climate sensitivity numbers much. But that has changed. To give one example, temperature estimates for the Eocene used to vary wildly. It was clearly one of the hotter periods of the planet’s history, with conditions thought to match our extreme and unlikely worst-case, high-emissions future. Yet estimates for exactly how much hotter it was than the preindustrial era varied hugely, from 9° to 23° C warmer. Inconsistencies between proxies, methods, and timeframes all added to the noise. Several projects set out to fix those issues. In one example, Dr. Gordon Inglis of the University of Southampton, with help from colleagues, carefully curated and analyzed a large collection of proxies with reasonable global coverage. “We have looked at multiple methods (for the first time) and applied this to the same data set (for the first time),” Inglis told me via email last year. “This enables an apples-for-apples comparison between methods.” The team's estimate of early Eocene temperatures suggests the era was 10° to 16° C warmer than the preindustrial era. That estimate is more robust and precise than before, enabling scientists to estimate a climate sensitivity for our time of 3.1° C of warming per CO₂ doubling, validating the value in the latest IPCC report. But the team's rigorous approach highlighted some problems. A marine temperature proxy based on microbe fat called “TEX86,” for example, tends to give warmer temperatures than other proxies, and some land-based proxies (leaf fossils, pollen, and chemicals called “branched GDGTs” from soil and peat bacteria) max out when their environmental temperature is around 25° to 30° C. “If we exclude the terrestrial data … we get much higher temperature estimates,” Inglis explained. Marine proxy data is often measured from tiny plankton corpses that are only .01 millimeter (.004 inch) wide. Oxygen isotopes in the plankton remains tell us the temperature of the seawater the plankton lived in, but those isotopes can change in their sedimentary tombs over the millions of years before they are sampled. This problem afflicted early work that suggested past warm climates had tropics that were about as cool as today. It turned out that the isotopes of those tiny skeletons had been reset after death by cold groundwater, so now scientists use only pristine corpses that were quickly buried in clay, sealing out the water. At other times, inconsistencies between different labs have swamped the climate signal, so scientists made standard reference chemicals to ensure that labs have a consistent baseline. This reduced uncertainty in ancient CO₂ measurements based on Boron isotopes by an order of magnitude. Meanwhile, increasingly sensitive instruments have made it possible to get better results from smaller samples, opening the door to new, more robust techniques. These efforts have earned ancient climates (“paleoclimates”) an equal footing with other lines of evidence as scientists seek to narrow climate uncertainty for our future. “These paleo time periods are potentially a very powerful tool to allow us to tune our models,” Professor Dan Lunt of the University of Bristol told me.

Looking at clouds from both sides now

Climate models used to do a poor job of simulating a full picture of ancient climates—if they matched the proxy temperatures in the tropics, they made the poles too cold. In past warm climates, warmth was amplified in polar regions, just as the Arctic today is warming twice as fast as lower latitudes. In the Miocene, for example, with CO₂ levels that we’re likely to see late this century, models underestimated Arctic temperatures by a whopping 10° C. This failure to get the ancient “polar amplification” right was known as the “Equable Climate Problem.” It was partly due to those too-cold tropical proxies I mentioned, but it was also due to the way that models simulated clouds. About a fifth of the world’s oceans today are covered by thick, low-altitude clouds that reflect sunlight back into space before it can warm the planet. Conversely, thin, high-altitude clouds let sunlight in and stop heat from escaping, so atmospheric scientists had been unsure if cloud changes would amplify warming or work against it. After flying planes crammed with instruments through clouds, particularly over the Southern Ocean, they eventually showed that the cloud changes that come with warming do, in fact, amplify global warming. When they then upgraded climate models with this more accurate cloud behavior, the models got polar warming about right. “We don't really have this Equable Climate Problem anymore, at least for some models,” Lunt told me. But this fix introduced a new problem. About a third of the new models warmed too much—their “sensitivity” to CO₂ was too high. Ancient climates have been key to diagnosing and correcting this. When Tierney’s group gave one of those hypersensitive models the job of simulating the Last Glacial Maximum (LGM) some 20,000 years ago, when Earth was coldest and ice sheets largest, it got things very wrong. The model doubled the cooling indicated by proxies, and it kept getting colder as sea ice built up in high latitudes, amplifying cooling. Tierney’s colleague, Dr. Jiang Zhu of the University of Michigan, looked under the hood of the climate model and found that its excess sensitivity was tied to cloud behavior.  “He could actually just tweak a few things in [the model] and knock down that sensitivity a bit, and then we could simulate a good Last Glacial Maximum,” said Tierney. “So that gave us the idea that we can actually start to use paleoclimates as ways to improve models.” They constrained a minimum climate sensitivity to around 2.4° C for an Ice Age climate, but warmer climates are more sensitive. “As you move into the warmer climates, that climate sensitivity actually goes up,” explained Tierney. To see just how much, climatologists turned to the Eocene. But you can’t simply take an ancient climate like the Eocene and compare it to ours. The continents were in different places, oceans were connected differently, vegetation cover was far from similar, the Sun was slightly dimmer, and the planet was free of ice. These differences must be calculated and subtracted to isolate the effect of higher CO₂ for comparison to our possible futures. When Lunt and colleagues ran climate models with Eocene geography, vegetation, solar brightness, and orbit, but with preindustrial CO₂, they still ended up with a warmer planet—3° to  5° C warmer, revealing the non-CO₂ contribution to Eocene warmth. That left the planet about 10° C above today due to the Eocene’s CO₂ levels. These were roughly 1,500 ppm, which translates to a climate sensitivity of 3° to 5° C in Eocene conditions (it’s just coincidence that those numbers match the non-CO₂ part of Eocene warmth). But hypersensitive climate models overcook the Eocene. When Zhu tested one of them in Eocene conditions, the model produced tropical land temperatures over 55° C—too hot for life. Since there’s fossil evidence of tropical rainforests then, we know the model got it wrong. These results are even worse when you consider that Zhu used the low end of Eocene CO₂ estimates, which are lower (and thus cooler) than the likely Eocene carbon levels. Consequently, climate models with sensitivities above or below what ancient climates allow were given less weight in the 2021 IPCC report. Some climate models, however, pass the paleoclimate test with flying colors: “The CESM1.2 [Community Earth System Model version 1.2], seems to do a very, very good job at the Pliocene, the LGM, and the Eocene, and [it] has a climate sensitivity of about 4° [C] under modern conditions,” said Lunt. Together, improved proxies and cloud cover calculations have halved the uncertainty of climate sensitivity for our time to between 2.5° C and 4° C, with a best estimate of 3° C, in the 2021 IPCC report. But that uncertainty still spans 1.5° C, as big as the warming that we ideally need to limit ourselves to. If the value is closer to 4° C, then we’re closer to overshooting that limit.

Permanent impermanence

Unlike today, the Pliocene, Miocene, and most of the Eocene weren’t Frankenstein climates—they were the result of millions of years adjusting to an equilibrium, with volcanic CO₂ emissions balanced by CO₂ burial through rock weathering. Vegetation, ice, and temperatures were all in harmony with CO₂ levels and the periodic influence of orbital wobbles. Our maladjusted climate has more in common with past CO₂ shocks driven by freakishly large volcanic CO₂ emissions. A relatively mild example is the Paleocene-Eocene Thermal Maximum (PETM), 56 million years ago, in which large amounts of CO₂ and methane flooded into the atmosphere over a few millennia, causing an already warm climate to warm by another 5° C. It was a time eerily reminiscent of today, with ocean acidification, oxygen-starved seas, species extinctions, and floods. “The PETM event is characterized by intense flood deposits," said Tierney. "We just see this really rapid, extreme precipitation.” Today’s warming climate is already producing unprecedented extreme weather: downpours, heat waves, droughts, and wildfires, which are set to get worse. This is what we see in past climates, too. “I think there's this realization that, for a lot of the paleoclimates we're studying, some of the changes that we interpret as a mean change are actually being driven by a change in the number or intensity of extreme events,” Tierney told me. To get a better handle on past weather extremes like that, Tierney has begun using past climates to reconstruct “paleoweather.” Some examples include looking at how “atmospheric rivers” filled huge lakes in the western US during the last ice age, inferring hurricane records from deposits left by storm surges, seeing how monsoons changed, and figuring out which areas got wetter and which got drier. “It’s how people experience climate change in their backyard,” said Tierney. “I mean, we've seen them all this summer in the United States, every single one! Drought, heat wave, hurricane, flood—record-setting in all cases. For me, it's a great way to connect the paleoclimate record to future climate change.”

Sea level rise is more insidious. If we eventually equalize at something like the Pliocene, that implies sea levels around 5-25 meters (16-82 feet) above today’s, which would drown much of the Eastern US seaboard and gulf coast. “How long it would take… is one of the most uncertain numbers in climate science,” said Lunt. It all depends on whether the ice melts gracefully or collapses. Should it melt gracefully, if we don’t exceed 2° C, we’re looking at roughly half a meter of average sea level rise by 2100, according to the IPCC. Unfortunately, levels will rise for millennia until they reach a Pliocene-like 8-13 meters (26-43 feet) higher in 10,000 years. But in the Pliocene and the Miocene, Antarctic ice was “highly dynamic,” regularly gaining and losing as much as 80 percent of its mass in response to millennial-scale orbital wobbles. “Dynamic” to a geologist, however, can mean changes that unfold over tens of millennia—unfortunately, the sediments offshore Antarctica and elsewhere that record those ice retreats can’t resolve changes at the century scale needed to constrain their speed. Instead, professor Robert DeConto of the University of Massachusetts Amherst simulated Antarctic ice melt with nearly 200 different combinations of melt rate and future warming. He discarded any simulations that did not match three constraints: recent ice loss, sea levels 125,000 years ago, and Pliocene sea levels. His team found that if we stay under 2° C by 2100, melting continues at today’s graceful pace. At 3° C, however, about 10 percent of the simulations went into a runaway collapse abruptly after 2060, accelerating sea level rise to 10 times faster than today. Once triggered, this “rapid and unstoppable sea-level rise” will, they say, carry on regardless of net-zero achievements or artificial CO₂ reduction. Runaway ice sheet collapse is referred to by the IPCC as a “low-likelihood, high-impact outcome” and remains a very contentious topic, Dr. Zeke Hausfather, a climate scientist at Berkeley Earth and the Breakthrough Institute, told me. “We're still trying to guess how fast ice sheets really disintegrate,” said Tierney. Beyond raising sea levels, ice loss also amplifies warming. This is because bright, white ice reflects sunlight, whereas the water and vegetation that replace it absorb sunlight and warm the planet.  This slow feedback is part of how the Miocene and Pliocene came to equilibrium. For our time, while sea level rise itself will cause disruption and abandonment of infrastructure over generations, the small amount of additional warming caused by ice loss, even with a runaway ice sheet collapse, will take millennia. “You're not going to have Antarctica turned into a verdant, green, ice-free area in the course of 100 years,” Hausfather said, “simply because [in] most of the places you're seeing large-scale ice loss, there's a lot of ice.” 

Healing a monster with a long tail

Humanity’s hope lies in the very fact that we do have a Frankenstein climate. The ice sheets have not yet passed that runaway tipping point. Forests have not yet replaced much Arctic tundra or sprouted in Antarctica. The Sahara is not yet green, and the Amazon is not yet all burned. Our climate fate is not yet sealed. And as soon as we reach net-zero CO₂ emissions, temperatures will essentially stabilize for a century or two.  “If you then go out for 1,000 years or 10,000 years, you are going to probably have larger, positive warming effects,” said Hausfather. “It depends a bit on some of the uncertainties in the carbon cycle. If you had 2° warming before you hit net-zero, even after 1,000 years you're probably still going to be 1.6°, 1.7° above preindustrial levels.”

At those timescales and temperatures, the long tail of slow feedback effects like ice sheet melt and vegetation changes add up. That means it will be hard to avoid a Pliocene-like future without removing CO₂ from the air—so-called “negative emissions.” “Most of the scenarios in the IPCC report that get us to net-zero emissions this century have net negative emissions thereafter,” explained Hausfather. “We're certainly [going to need] a large amount of negative emissions to deal with a long tail of residual positive emissions in the economy that are hard to decarbonize.” Because of remaining uncertainty in climate sensitivity, we can’t be sure how close we are to exceeding 1.5° or 2° C, so all efforts to bend our future back toward the Holocene will reduce the damage inflicted on generations of our descendants. “I don't think anyone is claiming that we have confidence that a 3° C world would necessarily result in flat temperatures for the next century or two, compared to, say, a 2° C world,” said Hausfather. “The higher the temperatures we get, the higher the risk that we trigger longer-term Earth System feedbacks.” Emergency first aid in the form of methane cuts, with drastic surgery in the form of rapid CO₂ cuts, followed by long-term rehab in the form of negative emissions, may all be required to restore our Frankenstein climate to something like Holocene health, albeit with permanent scars. “We can avoid making it worse, in the near term, if we emit less,” said Tierney. “I'm still hopeful for 1.5° [warming],” Lunt told me, adding, “I think I'm a phenomenally naive optimist!” In Mary Shelley’s novel, Victor Frankenstein dies pursuing his monstrous creation. We may yet heal ours.