“Climate Sensitivity” is a measure of the (in)stability of the Earth's temperatures, most commonly defined as the globally averaged temperature increase to be expected from a doubling of atmospheric carbon dioxide (e.g., an increase from 300 ppmv to 600 ppmv). (See also TCR and ECS.)
The most straightforward and obvious way of estimating climate sensitivity to a doubling of CO2 is by examining the result of the “experiment” which we've performed on the Earth's climate, by raising the atmospheric CO2 level from about 311 ppmv in 1950 (or 285 ppmv in 1850) to about 408 ppmv in 2018. We simply examine what happened to temperatures when the atmospheric CO2 level was raised by 31% (or 43%), and extrapolate from those observations.
However, there are a few pitfalls with that approach. For one thing, natural global temperatures variations due to ENSO can be larger than the “signal” we're looking for, so it is important that we choose an analysis interval which avoids those distortions. For another, it would be a mistake to assume that all of the warming which the Earth has experienced since pre-industrial conditions was due to anthropogenic CO2, because much of that warming occurred when CO2 levels were still very low, and because we know of other factors which must have contributed to warming, such as rising levels of other GHGs, and probably aerosol/particulate pollution abatement.
So the key question is, how much of the warming can be attributed to rising CO2 level? In the calculations below, the assumed answer to that question is an explicit parameter, “A” (for “Attribution”).
You can calculate an estimate of TCR sensitivity, using the time period and temperature index of your choice, as follows:
The formula is very simple:
Over that period, CO2 level rose from about 317 ppmv in 1960 to about 399 in 2014. Depending on which temperature index you trust, temperatures rose by about 0.5 °C (HADCRUT3) or about 0.75 °C (GISS), or somewhere in-between (or even less, according to the satellite-based lower troposphere temperature measurements). Let's use the midpoint of the surface indices, 0.625 °C:
If T1 is 0.00, T2 is 0.625, C1 is 317 (in 1960), C2 is 399 (in 2015), and A is 50%, then:
Note #1: ECS is usually estimated to be about 1½ × TCR.
Note #2: the above discussion doesn't mention minor GHGs like O3, CH4, N2O & CFCs. To take them into account, there are two simple approaches you can use. One is to substitute estimates of “CO2e” (CO2 equivalent) for C1 and C2. The other is to adjust A to account for the fact that some portion of the warming (perhaps one-fourth) is due to other GHGs.
Other than that, the attribution factor, A, is really just an educated guess, but it is based on expert opinion. The American Meteorological Society frequently surveys meteorologists and asks them what percentage of the last 50 years' warming they attribute to “human activity” (presumably mostly GHGs). This bar chart is from their 2017 survey report:
As you can see, the “average” or “midpoint opinion” of American broadcast meteorologists is that a little over half of the
warming was caused by human activity (presumably mostly by CO2):
(.905×15/92)+(.7×34/92)+(.5×21/92)+(.3×13/92)+(.095×8/92) = 57%
So if we attribute 57% of the warming to anthropogenic causes, and 75% of that to CO2, the attribution factor, A, should be 0.75 × 0.57 = 0.43 (43%), resulting in a calculated TCR sensitivity estimate of 0.81 °C per doubling of CO2.
Note that our calculation includes the effects of both positive and negative temperature feedbacks.
For ECS, multiply TCR by 1.5, yielding 1.21 °C per doubling of CO2.
The ECS/TCR ratio is sometimes estimated as high as 1.65:1. If we use that multiplier we could get the ECS estimate up to 1.34 °C per doubling of CO2, which is still slightly below the IPCC's “low end” estimate of 1.5 °C per doubling.
On the other hand, if the ECS/TCR ratio is only 1.25:1, then ECS = 1.25 × TCR = 1.01 °C per doubling.
Even if 100% (rather than 57%) of the warming since 1960 is attributed to anthropogenic causes (and 75% of that anthropogenic warming is attributed to CO2), TCR still comes out to only 1.41°C per doubling, and ECS = 1.5 × TCR = 2.12°C.
It is very difficult to approach the IPCC's “midrange” ECS estimate of 3°C per doubling, or the CMIP5 models' average assumption of 3.2°C per doubling, using this sort of analysis.
For a different approach to estimating climate sensitivity, which results in an estimate of ECS rather than TCR, I can recommend
this blog post by “SteveF” at
The Blackboard climate blog:
…and this follow-up:
Other researchers have reported quite different results. For example, Barrett (mostly circa 2011, archived here) calculates a TCR sensitivity of 1.84 ±0.11 °C. (Of course, to the extent that the 20th century's warming was natural, or due to unaccounted for anthropogenic forcings, the TCR sensitivity would be lower.)
More recently, Lewis & Curry 2016 reported a TCR sensitivity of 1.34 (0.91–2.44) °C.
Christy & McNider (2017) (or preprint) attempted to account for volcanic and ENSO distortions (like Santer et al 2014). They measured an underlying rate of climate warming of 0.096°C/decade, and calculated a lower tropospheric TCR climate sensitivity of +1.10 ±0.26 °C per CO2 doubling. The paper is quite long, but here's a readable discussion.
Atmospheric Physicist Richard Lindzen discusses an innovative approach to estimating climate sensitivity, based on measurements of evaporation rate changes, starting at 35:46 in this video. The result is a sensitivity estimate of about 0.8°C per doubling of CO2.
In 2014, the American Physical Society held a debate/discussion between three experts on each side of the issue, about whether climate sensitivity is high or low. They used an ECS of 2.5°C as the dividing line. Three of the experts (the lukewarmists) argued that ECS is below 2.5°C, and the other three experts (the alarmists) argued it is above 2.5°C. The APS has the transcript on their web site (or here), and I've written to them, asking for a copy of the recording of the event. ↑
Additional resources related to this topic can be found here: