Enlarge / Map of Antarctica today showing rates of retreat (2010-2016) of the "grounding line" where glaciers lose contact with bedrock underwater, along with ocean temperatures. The lone red arrow in East Antarctica is the Totten Glacier, which alone holds ice equivalent to ~3m (10ft) of sea level rise.

"What's past is prologue"- Shakespeare's The Tempest.

The year 2100 stands like a line of checkered flags at the climate change finish line, as if all our goals expire then. But like the warning etched on a car mirror: it's closer than it appears. Kids born today will be grandparents when most climate projections end.

And yet, the climate won't stop changing in 2100. Even if we succeed in limiting warming this century to 2ºC, we'll have CO2 at around 500 parts per million. That's a level not seen on this planet since the Middle Miocene, 16 million years ago, when our ancestors were apes. Temperatures then were about 5 to 8ºC warmer not 2º, and sea levels were some 40 meters (130 feet) or more higher, not the 1.5 feet (half a meter) anticipated at the end of this century by the 2013 IPCC report.

Why is there a yawning gap between end-century projections and what happened in Earth's past? Are past climates telling us we're missing something?


One big reason for the gap is simple: time.

Earth takes time to respond to changes in greenhouse gases. Some changes happen within years, while others take generations to reach a new equilibrium. Ice sheets melting, permafrost thawing, deep ocean warming, peat formation, and reorganizations of vegetation take centuries to millennia.

These slow responses are typically not included in climate models. That's partly because of the computing time they would take to calculate, partly because we're naturally focused on what we can expect over the next few decades, and partly because those processes are uncertain. And even though climate models have been successful at predicting climate change observed so far, uncertainties remain for even some fast responses, like clouds or the amplification of warming at the poles.

Earth's past, on the other hand, shows us how its climate actually changed, integrating the full spectrum of our planet's fast and slow responses. During past climate changes when Earth had ice sheets (like today) it typically warmed by around 5ºC to 6ºC for each doubling of CO2 levels, with the process taking about a millennium. That's roughly double the "Equilibrium Climate Sensitivity" (ECS) values used in climate model projections for 2100, which are calculated mainly from historical observations."What is past is prologue", inscribed on Future (1935, Robert Aitken) National Archives Building in Washington, DC.

Enlarge / "What is past is prologue", inscribed on Future (1935, Robert Aitken) National Archives Building in Washington, DC.

"We do expect the Earth System Sensitivity (change CO2 and have all the systems react—including ice sheets, vegetation, methane, aerosols etc.) to be larger than ECS. Work we did on the Pliocene suggested about 50 percent bigger, but it could be larger than that," Gavin Schmidt, director of the NASA Goddard Institute for Space Studies in New York, told me.

Or, as Dana Royer of Wesleyan University put it, "In short, climate models tend to under-predict the magnitude of climate change relative to geologic evidence."

Part of that greater magnitude is simply down to Earth's slow responses, which produce a net warming. Even if greenhouse gas emissions were to cease completely tomorrow, sea levels are committed to keep rising for centuries from thermal expansion and melting glaciers; ice sheets in Antarctica and Greenland are also committed to keep melting from the heat already built into the climate over recent decades. And because COlasts a long time in the atmosphere, in the absence of geoengineering to remove it, the world will overshoot any of our end-century temperature targets and stay elevated for centuries.

But those don't explain the entire gap, which suggests we're missing some other amplifying feedbacks. As the 2017 US National Climate Assessment put it: "model-data mismatch for past warm climates suggests that climate models are omitting at least one, and probably more, processes crucial to future warming, especially in polar regions."

Can the Miocene tell our future?

The Mid-Miocene Climate Optimum (MMCO) was an ancient global warming episode when CO2 levels surged from less than 400ppm to around 500ppm. (Ancient CO2 is measured in a variety of indirect ways like isotopes of boron or carbon in fossils and ancient soils, or from the pores on fossil leaves.) The cause of that surge was a rare volcanic phenomenon called a "Large Igneous Province" that erupted vast quantities of basalt in the Western USA 16.6 million years ago. Yvette Eley and Michael Hren of the University of Connecticut have been investigating how that changed the climate.

The tool? Fat molecules left in sediments by plants and microbes that lived at the time. Eley and Hren exhumed the chemical remains of microbes from Miocene muds in Maryland and then converted ratios of different fat molecules into soil temperature, using calibrations based on more than a decade of study of microbe fats in modern soils all over the planet. "Certainly, the timing of those flood basalts and the timing of when we see the shifts are pretty, pretty tight," said Eley. "Our biomarkers definitely track what CO2 was doing. Whatever is happening in the terrestrial system in terms of what's driving this event, it's definitely following pCO2."

As ancient climate changes go, the MMCO was mild compared to the end-Permian, end-Triassic, and others linked to mass extinctions. Miocene CO2 emissions were slow enough to avoid significant ocean acidification, for example, unlike today and during extreme past climate changes.

They also calculated sea temperatures in a similar way using chemically distinct remains of marine microbes: "We have a relative change across the MMCO of about 4-5 degrees [Celsius] in sea surface temperature, and sea surface temperatures that are about 6 degrees warmer than modern," said Eley.

Listing image by UMass Amherst / Edward Gasson

Page 2

They gauged Miocene atmospheric moisture by analyzing chemical traces of the waxy coating on plant leaves, calibrated to modern values from a wide variety of environments. "If we use our leaf-wax biomarkers as a proxy for atmospheric moisture, the data we get suggests that it was getting wetter across the MMCO," said Eley. "It's interesting to place our site in the context of other reconstructions. The Western US became more arid, South America gets wetter, parts of Europe get wetter, parts of Europe get dryer."

Places as far afield as East Coast USA, the Pacific Northwest, Western China, Patagonia, Central Asia, and the Atacama in South America, all became much wetter, causing a global uptick in erosion. The result was a general expansion and densification of forests. Remarkably, there's no sign of deserts either in North Africa or Asia, where today we have the Sahara and Gobi deserts.

That widespread wetting and greening is at odds with projected changes for our future, where areas that are currently wet are projected to get wetter, but dry areas are expected to get dryer. The difference may reflect the abrupt and unfinished nature of our climate change compared to the much slower Miocene change.

Even though landscapes right before the MMCO were already extensively forested (unlike today's, which reflect interglacial habitat deforestation by humans for millennia), the Miocene warming still produced clear changes in vegetation around the world, preserved in fossils, especially fossil pollen.

Across much of Europe, subtropical vegetation replaced cool-adapted plants, and dense swamp-forests resembling modern Louisiana clogged coasts and estuaries in Denmark and Germany (the European shoreline was 120 miles inland from today's coast). Those swamps accumulated brown coal that now fuels about a quarter of German electricity generation. Spain bucked the wet trend by having a hot, dry climate in the south, and a warm and wetter climate in the north just like today, with long dry seasons.

An artist's rendition of mid-Miocene life in Spain.

Enlarge / An artist's rendition of mid-Miocene life in Spain.

European vegetation also shows that there was less temperature contrast between seasons.

Siberia was rain-soaked with 3-5 times the precipitation of today, while swamps in Eastern Russia also accumulated coal. In Arctic Canada, where these days it's treeless permafrost tundra, the MMCO changed what had been a cool-temperate forest of birch, elm, holly, and umbrella pine into a warm-temperate forest rich in beech and hickory, sweetgum, walnut, and lime trees.

Nearer the equator, early elephants and antelope roamed Arabia's grassy, wet interior, while North Africa was lushly forested where Saharan sand dunes drift today. Apes spread across the forested planet, and it was about that time when great apes (our ancestors) diverged from other apes.

But it was Antarctica that altered most dramatically.

130 feet of sea level rise

Between a third and three-quarters of Antarctic ice melted. Land liberated by retreating ice sprouted tundra and forests of beech and conifers, which can't have happened unless Antarctic summers were warmer than 10ºC (50ºF—much warmer than the -5ºC/23ºF it is today). It's not clear what Greenland was up to, but there may have been a small ice sheet in Northern Greenland that melted substantially.

Consequently, sea levels rose by a whopping 40 meters or so (~130 feet). To put that in perspective, Mid-Miocene-like sea levels today would draw a new US Atlantic coast roughly along Interstate 95 through Philadelphia, Baltimore, Richmond and Fayetteville, North Carolina, inundating the New York-New Jersey-Connecticut metro area, Boston, most of Florida, and the coastal Gulf of Mexico. Similar things would happen across densely populated lowland areas around the globe, home to a quarter of the world's people.

Forty meters is just a bit more than the latest projections for modern sea level rise of 1-3 feet by 2100, and 4.5 to 5.25 feet (1.4-1.6 meters—home to about 5 percent of the world's population) by 2300, assuming we stabilize warming to around 2ºC. The difference is, once again, partly explained by time. According to the 2017 US National Climate Assessment, 2ºC of warming would commit us to a loss of three-fifths of Greenland's ice and one third of Antarctic ice, resulting in 25m (80ft) of sea level rise—but occurring over 10,000 years.

Even so, the Miocene hints that modern sea level rise could be larger and more rapid.

Sediments offshore of East Antarctica show that its ice was highly sensitive to even small changes in CO2 levels and orbital wobbles during the Miocene, responding with fast melting rates. How fast? Edward Gasson, of Sheffield University UK, and colleagues, calculated that Antarctica may have initially raised sea levels by roughly eight feet per century, tapering off to 30-36 meters (98-118 feet) after 10,000 years. That rate is consistent with a projection by Robert DeConto of Penn State and David Pollard of Amherst, based on the Pliocene, which had a cooler climate than the Mid-Miocene and sea levels "only" about 20 meters higher than today. DeConto and Pollard inferred that modern warming of about 2.5ºC in 2100 would raise sea levels 5.7 meters (19 feet) by the year 2500—about four feet per century. This rapid change may seem extreme, but we know that at times during just the past 500,000 years, sea levels have risen by as much as 4 to 5.7 meters (13 to 19 feet) per century.

If modern sea level rise turns out to be a Pliocene-like 4 feet per century, or a Miocene-like 8 feet per century, instead of the IPCC's 1.5 feet, we're facing a very different future. Sea level rise, compounded by tidal flooding and storms, would render large amounts of coastal infrastructure and property worthless in a generation or two (a mortgage or two).

Computer models did not support such rapid melting—until now.

Ocean-driven melting that undermines and destabilizes ice sheets was critical in the Miocene and seems to be so again today. That process can trigger runaway "Marine Ice Sheet Instability" as glaciers retreat inland—counterintuitively—into deeper basins due to Antarctica's bowl-like bedrock topography. The deeper they get, the more ice melts from below because it's at higher pressure, and the thinning glaciers tend to float, so they recede farther inland and accelerate, until they form tall cliffs that break apart under their own weight ("Marine Ice Cliff Instability"), making the situation worse. Worryingly, this may already be beginning in Antarctica.

Surface melt water, which requires air temperatures above freezing, is another accelerant. It seeps into cracks and freezes, fracturing ice like a log-splitter, a phenomenon witnessed in the demise of Greenland's Jakobshavn Glacier. Again, ice surface melting is happening today in parts of Antarctica. These melt-amplifying processes have only recently been added to new computer models, and now they show that ancient rates of sea level rise are possible for our descendants.

Retreating ice amplifies global warming by exchanging bright, reflective ice with darker, more heat-absorbent water and land. As a result, temperatures will slowly rise further.

What the Antarctic Ice Sheet may have looked like during the Miocene, 14 to 23 million years ago.

Enlarge / What the Antarctic Ice Sheet may have looked like during the Miocene, 14 to 23 million years ago.

Hope in uncertainty?

Could some of the gap between the Miocene climate and our projected future just be due to the sparseness and wide uncertainties in ancient climate data?

"CO2 changes in the Mid-Miocene might be larger than the median value reported. Other drivers are not known about at all. Methane or N2O levels are completely unknown. The amount of ozone or black carbon (from fires or vegetation emissions) are similarly uncertain," Gavin Schmidt told me. "Thus, even if we had perfect global temperature proxies (which we don't), estimates of sensitivity gotten from dividing the temperature by the CO2 forcing alone are not comparable with ECS estimates for today."

And yet, despite a spread of values for CO2 levels, proxies do cluster around 500ppm for the Mid-Miocene; some studies even suggest Mid-Miocene CO2 might have been lower yet driven even warmer temperatures. A relatively warm climate is supported by geological evidence for high sea levels and by fossils around the world, including offshore Antarctica.

Was it exaggerated by orbital cycles? Although individual Miocene glacial cycles were driven by orbital wobbles just like in the last ice age, warmth and maximum ice retreat persisted through several orbital and glacial cycles, tracking the higher atmospheric CO2. So we can't pin the MMCO just on Earth's orbit around the Sun.

Confusing matters further, the Miocene world started out different from today. The early Miocene climate was warmer than our preindustrial climate, grasslands had not yet proliferated, and the oceans were connected differently, with a current flowing from the Pacific to the Atlantic through what is now Panama, while the Bering Strait was closed. Yet scientists think the currents probably didn't have much of an effect on the climate, and in many other ways the planet was quite similar to today.

So there are big uncertainties in how well the Miocene represents our descendants' future. It's also true that there is no analog for the rapid rate of modern emissions in at least the last 66 million years. You could reasonably dismiss the relevance of any ancient analog on those grounds. But bear in mind that uncertainty is a double-edged sword: it cuts both ways, not only in the comforting direction.

If all of this feels depressingly "doomist," there is hope! It lies in Earth's slow reaction time, which gives us a (limited) window of opportunity.

"Sea Level Rise: Some Reason for Hope?" by Peter Sinclair and Yale Climate Connections.

A hand in the flame

If you pass your hand through a candle flame quickly enough, you won't get burned. The same principle applies to Earth—if we minimize the time that the planet spends above preindustrial temperatures, Miocene-like sea level rise may be avoidable.

Although Greenland and West Antarctic ice is already melting at an accelerating rate, East Antarctica is—for now—relatively stable (except the Totten Glacier). So, if we can keep warming well below 2ºC, DeConto and Pollard's models suggest East Antarctica will contribute little to future sea level rise.

But this will require us to reduce greenhouse gas concentrations, going beyond achieving "Net Zero" emissions.

"Negative emissions" (actively sucking CO2 out of the air) could slowly reduce global temperatures and stabilize many sources of sea level rise during the 22nd century. According to Matthias Mengel of the Potsdam Institute for Climate Impact Research and colleagues, falling CO2 would eventually allow Antarctica to begin accumulating ice, so sea levels would begin to fall again, three centuries into the future.

But this assumes negative emissions technologies can be deployed massively by the 2030s, a scenario with "limited realistic potential." Every five-year delay could commit our descendants to an extra 1 meter (3 feet) of sea level rise by 2300. Avoiding this future also assumes we don't trigger widespread ice sheet collapse in the meantime. If that happens, it will be effectively irreversible for millennia, even with removal of CO2 from the atmosphere.

Our present window of opportunity may not be open for long—scientists are scrambling to see if ice sheet collapse is starting in one of the largest glaciers of West Antarctica."Things are changing now very, very fast relative to a lot of what we see in the geological record," said Eley. "I would love to think that we're not going to end up with some of the worst-case scenarios, but we're already, I think, on a path to hitting those sorts of levels."

"[CO2] shows 100 to 200ppm increase in the middle Miocene. We've already bumped it up 127 since preindustrial. We're halfway there," said Hren. "The uncertainties are not simply the CO2 that we're going to get to, but really how the system will respond to something that is changing so rapidly."

Howard Lee is a freelance science writer focusing on climate changes in deep time. He has a bachelor's degree in geology and masters in remote sensing, both from University of London, UK. He currently lives in New Jersey. He'd like to thank Yvette Eley and Michael Hren for their interview at the AGU fall meeting, and to Gavin Schmidt, Edward Gasson, Dana Royer, and Peter Jacobs for their emailed comments.