Reading the media reporting of a new scientific paper released on 22 July, it was easy to get the impression that some “worse-case” climate warming possibilities are now off the agenda. “So this is good news?” a friend emailed. “No” was my answer.

It would be a grave mistake, and an illustration of how media reporting can get complex climate stories wrong, to find good news in this research.

The research in question is “An assessment of Earth’s climate sensitivity using multiple lines of evidence”. Climate sensitivity is the amount of warming to be expected from a doubling of the amount of carbon dioxide in the atmosphere. The most recent Intergovernmental Panel on Climate Change (IPCC) report found the range was 1.5 to 4.5 degrees Celsius (°C).

The new paper uses multiple lines of evidence — observations over the last 150 years, climate models, and climate history reaching back almost four million years — to constrain the estimated range for climate sensitivity.

One major finding was that very low estimates of climate sensitivity of less than 2°C — which climate sceptics like to emphasise — are inconsistent with the evidence. The Washington Post headlined with “Major new climate study rules out less severe global warming scenarios”, reporting that the research “found that Earth’s global average temperature will most likely increase between 2.3 and 4.5°C” compared to the IPCC range of 1.5-4.5°C. And it reported that there “would be a 6 to 18 percent chance of exceeding the upper bound defined by the study 4.5°C.” Which was good, balanced reporting.

Other coverage gave emphasis to the claim that some high-end estimates might now be off the table. But if the IPCC range was up 4.5°C, and the new paper range is up 4.5°C, what has changed that is “good news”?  Some research papers over the last two years, reporting on preliminary results from updated climate models for the next IPCC report, had suggested that climate sensitivity could be much higher, due to unresolved issues with cloud feedbacks. But scientists such as NASA’S Gavin Schmidt had warned that this work-in-progess should be taken with a grain of salt: “claims that climate sensitivity is much higher, or that worst cases scenarios need to be revised upwards, are premature”.  His warning was prescient.

And Science Daily reported one of the paper’s authors, Gabi Hegerl, saying that “these estimates make it improbable that climate sensitivity is at the low end of the IPCC range and confirm the upper range” (emphasis added).

Even though the new paper had found the upper bound of 4.5°C was the same as that of the IPCC report, media reporting included:

  • “It presents both good and bad news. The worst-case climate scenarios may be somewhat less likely than previous studies suggested. But the best-case climate scenarios — those assuming the least amount of warming — are almost certainly not going to happen”, reported E&E News.
  • Carbon Brief reported that “the silver lining to this cloud is that our findings also suggest that very high Equilibrium Climate Sensitivity (ECS – see below) estimates are unlikely”.
  • The Conversation headlined with “The climate won’t warm as much as we feared – but it will warm more than we hoped”; at arsTechnica, it was “Major study rules out super-high and low climate sensitivity to CO2”.

The impression across this reporting was that some of the more high-end warming estimates are now off the table. So should we be comforted that “our findings also suggest that very high ECS estimates are unlikely” or that “the climate won’t warm as much as we feared”?

What none of the media reports said — as far as I can ascertain — is that this new study is only about short-term warming, not the “slow” warming where the full impacts of climate change on ice sheets and carbon stores are taken into account. And that leads to a very different story.

Two types of climate sensitivity

Climate sensitivity is generally understood as the temperature rise subsequent to a doubling of the level of carbon dioxide (CO2), once the planetary system has returned to equilibrium (balance); that is, once all the perturbations caused by the increase in incoming energy have worked their way through the system, including hotter atmosphere and oceans, melting ice sheets, changed vegetation and so on. Whilst some of the impacts are fast and can be observed on short to decadal timeframes (atmospheric warming, changes to clouds, increased water vapour), others such as the melting of the polar ice sheets or large-scale loss of frozen carbon stores (permafrost) are more relevant on century to millenia timeframes. These “slow” feedbacks (self-reinforcing loops) can take many human generations in time to manifest.

IPCC reports have focused on what is generally called Equilibrium Climate Sensitivity (ECS).  The 2007 IPCC report gave a best estimate of climate sensitivity of 3°C and says it “is likely to be in the range 2°C to 4.5°C”. The 2014 report said that “no best estimate for equilibrium climate sensitivity can now be given because of a lack of agreement on values across assessed lines of evidence and studies” and only gives a range of 1.5°C to 4.5°C.

What the IPCC reports failed to make clear is that the ECS measure omits key “slow” feedbacks that a rise in the planet’s temperature can trigger. These include the permafrost feedback and other changes in the carbon cycle which can release large amounts of greenhouse gases to the atmosphere, a decrease in the ocean’s carbon-sink efficiency, and the melting of polar ice sheets and hence changes in the Earth’s reflectivity.

Climate sensitivity which includes these feedbacks  — known as Earth System Sensitivity (ESS) — does not appear to be acknowledged in the 2014 IPCC report at all. Yet, there is a wide range of literature which suggest an ESS of 5–6°C, for example by the Geological Society and a lifetime of work by former NASA climate science chief James Hansen and his co-researchers, including “Climate sensitivity, sea level and atmospheric carbon dioxide”.

The long-term (slow feedbacks) or Earth System Sensitivity can be inferred from the paleoclimate record. Over the last million years, the Earth’s climate has see-sawed between atmospheric CO2 of around 180 parts per million (ppm) in glacial periods and 300 ppm CO2 during interglacial or warm periods (see illustration, drawn from the Hansen et al paper referenced above). This is an increase in CO2 levels of around 65%, that is, less than a doubling.  Yet during such periods, the temperature see-sawed by up to 5°C. Now, there are some other factors to take into account that have an influence in triggering this process, principally changes in Earth’s orbit (Milankovitch cycles). Nevertheless there is a wide range of recent literature that points to ESS of around 5-6°C, such as “Well below 2C: Mitigation strategies for avoiding dangerous to catastrophic climate changes”.

So does the recent climate sensitivity paper deal with Equilibrium Climate Sensitivity (fast feedbacks only) or Earth System Sensitivity (including slow feedbacks)? The evidence is clear that it does not include slow feedbacks.  For example:

  • The paper notes that Earth’s Equilibrium Climate Sensitivity (ECS) is defined generally as “the steady-state global  temperature increase for a doubling of CO2, has long been taken as the starting point for  understanding global climate changes. It was quantified specifically by Charney et al. as the equilibrium warming as seen in a model with ice sheets and vegetation fixed at present-day values” (emphasis added).  In other words, slow feedbacks are excluded.
  • The paper uses a term S to denote “effective climate sensitivity”, which is derived from the traditional ECS definition, but based on the impacts at 150 years after an increase in CO2 levels, rather than the longer period in which equilibrium has been restored. The paper notes that it “does not formally exclude any feedback process, but the 150-year time frame minimizes slow feedbacks (especially ice sheet changes)”  (emphasis added), and “Our S incorporates only feedbacks acting on time scales of order a century”.  Given that slow feedbacks generally are relevant on longer than century timescales, it is again clear that they are not included.
  • And S is compared with “Earth System Sensitivity, (which) by contrast, reflects the slower feedback processes such as changes to the carbon cycle and land ice” (emphasis added).
  • And again: “In this assessment we consider well-mixed gases (CO2, CH4, N2O) to be specified forcers, since in the modern era, they are effectively under human control. Thus we do not include climate-driven variations of these gases (e.g., carbon cycle feedbacks)” (emphasis added).

  • In fact, in that section of the paper using data from climate history back millions of years, the role of slow feedbacks are excluded in order to gain insight into the role of fast feedbacks only: “The paleoclimate data come from intervals where the climate was different to today, but fairly stable for several thousand years, meaning that slow feedback processes need to be taken into account. By treating these slow processes as forcings rather than feedbacks, we are able to make inferences about S. Both the temperature changes that are used, and the slow feedback influences that are removed, are constrained using indirect proxy records” (emphasis added).

Consequences

So what does all this mean?

  1. The new paper has not reduced the higher end of the range for short-term sensitivity; in fact it has reaffirmed the 4.5°C boundary from the most recent IPCC assessment report.
  2. The paper under discussion is concerned with climate sensitivity including fast feedbacks only (Equilibrium Climate Sensitivity), so it has nothing new to say about the eventual warming caused by increased levels of greenhouse gases once slow feedbacks are taken into account (Earth System Sensitivity).
  3. The estimates of Earth System Sensitivity including slow feedbacks remain unaffected, which are likely in the 5-6°C range.
  4. Using that sensitivity, just the present level of CO2 of 415ppm, if maintained, would be enough with the slow feedbacks to eventually increase the global average temperature by 2.8-3.2°C. This would be a catastrophic level of warming, including sea-level rises in the tens of metres, and swathes of the planet too hot for human habitation. This is consistent with the paleoclimate record of past climates, for example, here and here and here and here. By contrast, using the ECS measure of 3ºC, the resulting warming would be 1.7°C, around half a degree hotter than now, so understanding what climate sensitivity assumptions are behind a future warming projection is important!
  5. Whilst “slow feedbacks” such as the “release of greenhouse gases from thawing permafrost, reduced ocean and terrestrial CO2 removal from the atmosphere” are generally considered to operate on slow timescales, a report as far back as 2007 was able to conclude “there is some evidence that such feedbacks may already be occurring in response to the present warming trend” because the rate of warming driven by human actions is faster than at any time in the past.  In fact if we look at more recent evidence from the polar regions, it is clear that the permafrost feedback is already non-negligible, and that both Greenland and the Antarctic ice sheet are exhibiting increasing rates of ice mass loss.
  6. In other words, warming of well less than 2°C can trigger slow feedbacks. In a personal communication on 19 March 2018, James Hansen told me that: “Global temperature has risen well out of the Holocene range and Earth is now as warm as it was during the prior (Eemian) interglacial period, when sea level reached 6-9 meters higher than today. Limiting the period and magnitude of temperature excursion above the Holocene range is crucial to avoid strong stimulation of slow feedbacks. A danger of 1.5°C or 2°C targets is that they are far above the Holocene temperature range. A long-term global average temperature of even 1.5°C could spur ‘slow’ climate feedbacks and is not an appropriate goal.”

And that is a story that the media coverage of this important new paper failed to tell.