Climate Sensitivity: Feedbacks Anyone?
Posted on 28 August 2011 by James Wight
Climate modeling is notoriously complex, but it all boils down to one key question: How sensitive is the Earth’s climate to perturbations like an increase in the greenhouse effect?
“Climate sensitivity” describes the amount of global warming you get from a specified forcing once all climate feedbacks are taken into account. A forcing is something that changes the Earth’s energy budget: the difference between the amount of energy entering the Earth system and the amount leaving it. If the energy budget is in balance, the Earth’s temperature is stable. A forcing creates an energy imbalance, causing global temperature change until the system gets back in balance. Forcings can be quantified in watts per square metre (W/m2), and causes can include variation in sunlight or the Earth’s orbit, surface reflectivity (“albedo”), and the greenhouse effect. Climate sensitivity is usually expressed in degrees per doubling of atmospheric carbon dioxide (CO2), a forcing of about 4 W/m2.
A recently published paper by Hansen and Sato (2011), Paleoclimate Implications for Human-Made Climate Change, examines evidence from past climates about the various feedbacks that affect climate sensitivity. A feedback is a mechanism that either amplifies (positive feedback) or dampens (negative) the initial effect. Interest is a feedback on a loan. If there were no feedbacks in the Earth’s climate system, physics tells us climate sensitivity would be 1.2°C for a doubling of CO2. In reality, a complex array of interacting positive and negative feedbacks come into play. Climate models include “fast feedbacks” like water vapor, clouds, sea ice, and aerosols (reflective particles that hang in the atmosphere), but exclude longer-term “slow feedbacks” like ice sheets (an icy surface reflects more heat than a dark surface) and greenhouse gases (warming releases gases from the oceans, melting permafrost, etc).
There is a broad consensus that fast-feedback sensitivity is 3°C for doubled CO2. In other words, fast feedbacks multiply the 1.2°C direct warming by two-and-a-half. Model estimates come with large error bars that have proven difficult to reduce as climate models have become more realistic over the decades, because modeling all the positive and negative feedbacks is so complicated. However, the paleoclimate record allows us to circumvent that problem, as past climate changes obviously included all existing feedbacks. Climatologists study past climates with climate proxies like ice cores and sediment cores.
The most accurately known climate changes are the ice age cycles of the last few hundred millennia, recorded by ice cores and ocean sediment cores. During that time the Earth oscillated from brief “interglacial” periods like today, when ice sheets are confined to Antarctica and Greenland; to long “glacial” periods when global temperature plunged by 5°C, ice sheets covered much of Canada and Europe, and sea level fell over 100 metres. The forcings largely driving these climate swings were surface albedo and greenhouse gases – themselves slow feedbacks on tiny orbital forcings sustained over long periods, but for the purpose of finding fast-feedback sensitivity they are considered forcings. The sensitivity Hansen and Sato derive is 3°C, exactly as the models predicted.
A landmark 1979 report by the National Academy of Sciences, whose lead author was Jule Charney, defined climate sensitivity as including the fast feedbacks of water vapor, clouds, and sea ice. This gave us the term “Charney sensitivity”, but its meaning varies: it can encompass all fast feedbacks, or it can exclude feedbacks Charney excluded, such as aerosols. Hansen and Sato make a distinction between the “all-fast-feedback sensitivity” (Sff) and a definition excluding aerosol feedbacks (Sff-a). They argue Sff is more useful as it can be measured with much greater precision from paleoclimate. If aerosols are counted as a forcing then the uncertainty is very large, because we don’t know how aerosols changed in glacial periods. If they are counted as a feedback then the answer is 3.0±0.5°C, consistent with but more precise than the IPCC model value of 3.0+1.4/–0.9°C.
Figure 1: Implied fast-feedback climate sensitivity from ice age transition measurements. 0.75°C per W/m2 corresponds to a fast feedback sensitivity of 3°C for doubled CO2 (Source: Hansen and Sato 2011).
But in the long run, what will be important is the climate sensitivity including fast and slow feedbacks. Slow feedbacks have received far less attention. Current models don’t take them into account, so paleoclimate is the only available tool to estimate them. Again, Hansen and Sato use the glacial-interglacial cycles. Though earlier we counted ice sheets and greenhouse gases as forcings in the cycle, both were actually slow positive feedbacks on tiny energy imbalances, lagging behind the initial temperature increase by centuries. The large magnitude of the glacial-interglacial swings tells us that slow feedbacks are large and positive on millennial timescales.
The term “Earth system sensitivity” is sometimes used for slow-feedback sensitivities, but as with Charney sensitivity, usage varies. Hansen and Sato propose a series of definitions including different combinations of feedbacks:
- Sff+sur – all fast feedbacks plus surface albedo feedbacks
- SCO2 – all fast feedbacks plus surface albedo and non-CO2 greenhouse gas feedbacks
- Sff+sf – all fast feedbacks plus surface albedo and all greenhouse gas feedbacks
Surface albedo feedbacks
Sff+sur is the long-term sensitivity to a specified greenhouse gas forcing. It is useful for cases where greenhouse gases are the initial cause, as with anthropogenic global warming. When using this definition, any greenhouse gas feedbacks must be calculated by separate carbon-cycle models.
The dominant surface albedo feedback is ice sheet area (though there is some contribution from vegetation cover). This positive feedback is not stable over geologic time, because it only works when the planet is cool enough to form ice sheets. Earth was ice-free for most of its history; in such times albedo feedbacks must have been near-zero, and Sff+sur about the same as Sff. The Earth has cooled dramatically in the last 50 million years due to geologic CO2 change. The current ice age began with the formation of the Antarctic ice sheet 35 million years ago. The Northern Hemisphere ice sheets emerged in the last few million years. As the world cooled, orbit-forced climate oscillations became greater and greater, as increasing ice increased the ice albedo feedback. The more ice on the planet, the more sensitive the climate.
The same glacial cycles Hansen and Sato used to estimate Sff also tell us about Sff+sur during that time. About half of the forcing was from ice sheet feedbacks and half from greenhouse gas feedbacks. Since here we’re defining greenhouse gases as a forcing and ice sheets as a feedback, Sff+sur must be about twice Sff, so Sff+sur = 6°C. Albedo feedbacks were approximately linear in those cycles, so the 6°C sensitivity applies to the range of climate states between interglacial and glacial.
But we are pushing the climate not towards glacial conditions, rather the opposite direction, in which the ice albedo feedback shrinks and eventually vanishes. So Sff+sur is less than 6°C for positive forcings relative to today, perhaps only ~4-5°C for climates between now and 3 million years ago. However, for a positive forcing just enough to melt the Antarctic ice sheet, there is a large nonlinear albedo feedback. Hansen and Sato explore this matter further in their 2008 paper Target Atmospheric CO2: Where Should Humanity Aim? The current global temperature and total required forcing (assuming Sff = 3°C) are about halfway between the temperature 35 million years ago and in recent glacial periods. So averaged over the climate range between today and 35 million years ago, Sff+sur works out to be almost 6°C. The ice albedo feedback is still in play.
Greenhouse gas feedbacks
SCO2 is the long-term sensitivity to a specified CO2 forcing. Levels of non-CO2 greenhouse gases are not readily measured by proxies, so in paleoclimate studies it is easier to count them as feedbacks than forcings. These gases include methane (CH4) and nitrous oxide (N2O), and they are thought to be a net positive feedback.
CO2 constituted three-quarters of the greenhouse gas forcing between glacial and interglacial. Since here we’re defining CO2 as a forcing and other greenhouse gases as feedbacks, SCO2 must be about a third higher than Sff+sur, so SCO2 = 8°C within that climate range. With ice sheet feedbacks varying as before, SCO2 to positive forcings from today’s climate must be: ~5-6°C for small forcings, almost 8°C averaged over the period of ice sheet loss, and 4°C on an ice-free planet.
Sff+sf is the “ultimate Earth system sensitivity” including all feedbacks. It is relevant to forcings which are completely external to the Earth system. In principle it applies to orbital cycles, though figuring out the calculations is tricky because the regionally varying nature of orbital forcing is not fully understood. Sff+sf cannot be accurately estimated from paleoclimate data, but must be extremely large for negative forcings from the current climate state.
Sff+sf is the only one of Hansen and Sato’s definitions which includes CO2 feedbacks. At the moment the carbon cycle is acting as a negative feedback, as oceans and vegetation are removing some of our CO2 emissions, but as global warming continues, those carbon sinks are expected to fill up and start emitting CO2. Like other greenhouse gases, CO2 was a large positive feedback in the glacial-interglacial cycles. Eventually, excess CO2 is removed from surface reservoirs by a negative weathering feedback, but this takes hundreds of millennia.
Transient climate response
All the above terms (Sff-a, Sff, Sff+sur, SCO2, Sff+sf) are possible definitions of equilibrium climate sensitivity (ECS). They all refer to the final amount of warming after accounting for climate inertia and feedbacks; the differences between them arise from different decisions about which processes to count as forcings or feedbacks. You may also sometimes hear about transient climate response (TCR), which the IPCC defines as the temperature increase at the time of CO2 doubling if CO2 increases by 1% per year. The IPCC estimates TCR is ~2°C, with a model range of 1-3.5°C.
How slow are slow feedbacks?
In the glacial-interglacial cycles slow feedbacks took millennia, but perhaps that was only because orbital forcing changed very slowly. The peak sea level rise rate occurred at the same time (within measurement error) as the peak forcing, suggesting ice sheets could melt faster if the climate changed faster. Sea level rises of several metres per century were not uncommon in past deglaciations; it is the present stability that is unusual.
Mainstream projections of sea level rise for 2100 are ~1-2 m (substantially higher than the 18-59 cm in the last IPCC report). An upper limit of 2 m by 2100 has been proposed, but this is based on the questionable assumption that glaciers will not move faster than their fastest speed in recent decades. In Antarctica, an amount of ice worth 20-25 m is rooted below sea level and held back by ice shelves. Hansen and Sato argue that satellite data show ice sheet loss occurring exponentially, with a doubling time of perhaps a decade. If this trend continues sea level could rise 5 m within a century; however, exponential ice loss is limited by the temporary negative feedback of regional iceberg cooling.
The response time of ice sheets is shorter than the negative weathering feedback which removes excess CO2. So we can expect the slow feedback response to eventually be realized: several degrees of warming will ultimately lead to tens of metres of sea level rise and double the original warming. Post-2100 sea level rise might seem a long way off, but it will be determined by policy decisions taken in the near future. Once an ice sheet begins to collapse there is no way to stop it sliding into the ocean. We would be subjected to centuries of encroaching shorelines.
What it all means
The exact value of climate sensitivity depends on which feedbacks you include, the climate state you start with, and what timescale you’re interested in. While the Earth has ice sheets the total climate sensitivity to CO2 is up to 8°C: 1.2°C direct warming, 1.8°C from fast feedbacks, 1°C from greenhouse gas feedbacks, and nearly 4°C from ice albedo feedbacks. The slow feedbacks have historically occurred over centuries to millennia, but could become significant this century. Including CO2 itself as a feedback would make climate sensitivity even higher, except for the weathering feedback which operates on a geologic timescale.
As is explained in Target Atmospheric CO2, 4 W/m2 of greenhouse gas forcing sustained long enough would ultimately return the Earth to an ice-free state, raising the global sea level by 75 metres. The preindustrial level of atmospheric CO2 was ~275 ppm, so 4 W/m2 would be the effect of doubling the CO2-equivalent of all greenhouse gases (CO2e) to 550 ppm, or increasing CO2 to ~450 ppm with other greenhouse gases responding as feedbacks. Currently CO2 is at 390 ppm and rising; CO2e levels are at 470 ppm and counting; implying significant feedbacks are already in the pipeline. However, we may still be able to prevent them if we can get the Earth back in energy balance by reducing atmospheric CO2. In practical terms, that means cutting global CO2 emissions to near-zero as soon as possible.
Contrarians often argue the paleoclimate record shows CO2 and climate change are nothing to worry about. What it actually tells us is the climate system is extremely sensitive to perturbations – and we are running out of time to prevent the global warming we started from spiraling out of our control.