Is Global Mean Surface Temperature Misleading?

In December 2020, the CO2 Coalition published a document authored by Richard Lindzen and John Christy entitled, The Global Mean Temperature Anomaly Record: How it works and why it is misleading. It should be acknowledged that this document is a political and partisan document. It is not written to advance our understanding of the GMST anomaly record. It's written to affect policy. Caleb Stewart Rossiter, Executive Director of CO2 Coalition, makes this nearly explicit. After asserting that academic publishing has been taken over by a "'climate crisis' and an 'easy to change sources' narrative, he says, "this paper, which in earlier days would have, as a matter of course, been published by a scholarly journal, now finds a home with our advocacy group." This is not an academic paper (and this by itself is not a criticism of the paper), and it would never be published in a scholarly journal, since the intended audience is general voters, not scholars. It found its home in an "advocacy group" because that's precisely what it was written for.

To me, the paper itself is hugely problematic on several fronts. It's my contention that this paper is not an honest attempt to explain why the GMST anomaly record is misleading. It's intended to mislead. I say this will full awareness that the authors of this paper are legitimate scientists who have far more expertise than I have. But this leads me to one of my litmus tests for academic integrity. I have a B.S. in Geology and a masters in Biblical Studies. I am no match for these scientists - they have far more competence and expertise than I have. So if I can easily and conclusively refute them, the reason is not likely to be that Lindzen and Christy (hereafter LC) are not competent scientists. They are either intent on deceiving or they are so strongly held by their convictions that they are not doing their homework to get the evidence correct. I see no other options here.

I was tempted to respond to this paper sentence by sentence, but this would bore even me. I want to deal instead with three major themes I see in this paper that I consider hugely problematic - first, their claim that current warming is "easily manageable," second, that GMST is a "dubious metric," and third, I want to considered whether CO2 can be blamed for current warming.

Is Current Warming "Easily Manageable?"

Thankfully, LC acknowledge that there has been about 1.2 C warming over the last 120 years. I'm going to make that a little more precisely stated. GMST is currently 1.2 C warmer than the 1850-1900 mean, where 1850-1900 is taken by the IPCC as an approximation for preindustrial levels. So the 1.2 C figure maps onto projections from the IPCC quite well. Of the major GMST datasets, NASA and NOAA go back to 1880 while HadCRUT and Berkeley go back to 1850. LC simplify this to 1900-2020, which is fine. So far, so good.

LC claim that this increase is "easily manageable," but that assertion is not supported by evidence. And it should be noted that while 1.2 C in 120 years averages to 0.1 C/decade, most of the warming has occurred following 1970. Over recent decades, warming has accelerated to values exceeding 0.2 C/decade in every GMST dataset. Rates of warming exceeding 0.2 C/decade have been quite literally undetected in either the instrumental or proxy record before the last few decades. It is therefore not self-evident at all that current warming is "easily manageable." That's an assertion that requires evidential support. However, the evidence that is given does not support the conclusion. For instance, LC compare the 1.2 C increase in GMST to non-global metrics. For instance, LC write:

"Even if a single station had recorded this [1 C] increase itself, this would take a typical annual range of temperature there, for example, from -10 to 40 degrees in 1900, and replace it with a range today from -9 to 41. People, crops, and weather at that station would find it hard to tell this difference."

This is wrong on multiple levels. First, LC are confusing changes in GMST with changes in a single  weather station. And locally, a change in mean temperature (Tavg) at a single station may or may not occur as they described. If we take a gaussian distribution of Tavg values from 1900 ranging between  -10 and 40 degrees and then shift the mean 1 degree, the resulting gaussian distribution could in principle change shape. It could become wider and the "hot" tail could become fatter. In fact, this has been observed in NH land temperatures.[2] So the resulting distribution may be from -10 to 43, and it could be that people would notice a greater frequency of hotter summer days.

Second, even if the mean shifts as LC suggest and people are not able to notice, it misses the point entirely. A 1 C shift in Tavg in any given station is far less significant than a change in GMST, and this should be obvious. As GMST increases, not every region warms at the same rate. Land is warming more rapidly than the oceans, the NH is warming more rapidly than the SH and the Arctic is warming now between 3x and 4x faster than the global average.[3] These systemic changes in climate affect the frequency of days warmer than freezing vs below freezing in the Arctic. This means that Greenland spends more days above freezing than it used to, which means in turn that the Greenland ice sheet loses ice more quickly and the resulting sea level rise acceleration affects coastlines in areas that aren't anywhere near Greenland and that that aren't warming as rapidly.

One way to think of this is that while GMST is sensitive to a given increase in CO2 concentrations, the Earth's climate system also experiences a sensitivity to an increase in GMST. What LC are claiming is that the Earth's climate system isn't very sensitive to temperature changes on the scale expected by current warming. For them, 1.2 C is easily manageable, and we need not be too concerned about future warming. In fact, they suggest that it is "likely to be beneficial." But to what extent can we estimate the sensitivity of the Earth's climate system to a given increase in GMST? It turns out we can say a whole lot more than LC are willing to say. 

What we need to investigate is to what extent the range of warming expected from AGW will affect the earth's climate system. The IPCC projects that we may experience as much as 4 C by 2100 with no mitigation, of which we've already experienced 1.2 C. So how significant is this for the Earth's climate system? One way to quantify this is to examine the difference in GMST from the last glacial maximum (LGM) to the Holocene Thermal Maximum (HTM). We can standardize the difference in GMST between these two times as an "Ice Age Unit" or IAU. Quantifying that value of change in GMST can be a little tricky. In Shakun et al 2012[4], it's about 4 C. In a recent reanalysis study Osman[5], it's about 6-7 C. One popular graphic set it at 4.5 C.

An Ice Age Unit Scale

It would seem that a good summary of these estimates would be to view one IAU as 5 ± 1 C. A difference of 1 IAU is literally the difference between Canada being under a 1-2 mile thick ice sheet and a habitable Canada with harsh winters. In other words, the Earth's climate system is highly sensitive to a 1 IAU change in global temperature. As GMST cools, the planet spends more time below freezing, allowing ice sheets to accumulate, since it's not warm enough in the summer to melt all the ice that forms in the winter. This affects albedo, sea levels, the distribution of flora and fauna, etc. We can say with some certainty that an increase in 1 IAU is likely to have drastic effects on the Earth's climate system, and since the industrial revolution the earth has already experienced 1/4 of that amount of warming. There is also  a good deal of likelihood that, with no mitigation, we will surpass 1 IAU sometime in the 22nd century.

It takes time for the Earth's system to reach equilibrium with a given change in temperature. Scientists estimate ECS to be 3 C, but these include only rapid feedbacks (on the order of decades). Over longer time frames (thousands of years), the Earth's climate continues to change, though more slowly. Albedo changes as ice sheets shrink and as boreal forests move poleward, and the Earth's climate system has not yet reached equilibrium with current CO2 levels. The last time CO2 was 400 ppm was 3 million years ago. At that time, sea levels were 15 m higher and Greenland supported only an ephemeral ice sheet.[6] Trees grew in West Antarctica a few hundred miles from the South Pole.[7] The best geologic evidence we have from paleoclimate studies is that, even if we keep CO2 at 400 ppm, when the Earth's climate system reaches equilibrium with 400 ppm hundreds of years from now, the Greenland Ice Sheet will be essentially gone along with much of the West Antarctic Ice Sheet. Sea levels will be 15 m higher. 

And we have not yet fully accounted for how rapidly temperatures are changing. GMST has experienced 1.2 C warming in about 100 years, and current warming exceeds 0.2 C/decade in all GMST datasets, calculated as 30 year trends (1992-2021). The Quaternary glaciation cycles have been about 100K years long over the last 1 million years, and during the warming phase of these cycles GMST has increased by about 4 C to 6 C over a span of about 10 to 15K years. This means that the warming phase of the glacial cycles occurred at rates averaging less than 1 C/millennium (less than 0.01 C/decade). Almost certainly there were periods within these warming phases that were significantly more rapidly than that over short time spans. But 0.2 C/decade is more than 20x more rapid than the average rate of warming as the Earth emerged from glaciations.

What should be clear is that LC have not made a case that 1.2 C warming is "easily manageable," given the sensitivity of the Earth's climate system to changes in GMST and the rapid rate of current warming. Without any mitigation, we will exceed 1 IAU sometime in the 22nd century, and that rate of warming is quite possibly unprecedented in geologic history, save perhaps for catastrophic events.[8]

Is the GMST Temperature a "Dubious Metric?"

LC acknowledge that GMST shows 1.2 C warming since 1900, and they avoid challenging the value of the anomaly as being the product of fraud or incompetence on the part of climate scientists. Their challenge to this appears to be not that the GMST anomaly is wrong, but its value is being mispresented as alarming in public discourse, when in reality (according to LC) the GMST anomaly is a "dubious metric" that is not a reliable indicator that there is a climate problem that needs to be fixed.

Superficially, I can't disagree. The mere fact that GMST is increasing at 0.2 C/decade does not mean that we're causing that increase, so it doesn't necessarily mean by itself that we can (or should) do anything about it. However, there are several unclear, misleading and incorrect claims that are made along the way here, and these impact LC's conclusions.

1. Noisiness and Natural Variability vs Climate Trends.LC claim acknowledge that the GMST Record shows "consistent trends," but it is also "noisy" with "fluctuations of a tenth or two of a degree are unlikely to be significant." Yet that's a testable claim. NASA for instance has published their confidence interval for GMST anomaly at 0.05 C or less for recent decades, gradually increasing to 0.15 C in the late 19th century.[9] Certainly LC could have investigated this and reported the actual CI for these datasets. 

Then they say, "Given the noise and sampling errors, it is rather easy to 'adjust' such averaging, and even change the sign of a trend from positive to negative." This is a bit confusing. This may well be true of individual stations. Individual stations are "quite noisy" and bias correction (adjustments) does occur that in some cases may change the sign of the trend. But the focus in public discourse is GMST, and bias correction has the net effect of decreasing the warming trend by increasing sea surface temperatures prior to 1940 to correct for sampling errors.  Adjustments to not change the sign of the  overall trend.

And of course, there is natural variability or noise that affects GMST data. The most notable source of noise is ENSO, as El Nino years trend warmer than La Nina years. However, this internal variability averages out over time, since ENSO shifts from warm to cool phases (El Nino and La Nina) every few years or so. Conventionally, climate scientists look at climate in 30 year intervals so that the climate signal emerges from the noise. LC offer no actual criticism of this practice, and it seems to be a good one.

But they do criticize using running averages that smooth GMST trends. They don't offer any actual criticism of the practice. Instead, LC criticize the way that makes the graphs look. They write that smoothed graphs "look much cleaner and presumably more authoritative than the unsmoothed results or the scatter diagrams, but this tends to disguise the uncertainty, which is likely on the order of 0.1-0.2 degrees." I have no idea why a smoothed graph looks more authoritative than a graph without smoothing. I also object to saying that this practice disguises the uncertainty, which is actually on the order of 0.05 C, not 0.1-0.2 C. Many graphs are posted with the uncertainties actually on the graph. But smoothing does allow viewers to get a better sense of the climate trend without being distracted by natural variability. This is valuable, and LC offer no criticism that using smoothing misrepresents climate data.

2. Motivations for GMST Graphs. LC go further to assign alarming motives to the presentation of graphs of GMST. LC show two graphs (Figures 3 and 4) with annual anomalies for individual stations. Showing the the station data with the GMST anomaly shows that individual stations may deviate by as much as 2 C above or below the GMST anomaly for any given year. They then show two graphs (Figures 5 and 6) displaying only the GMST anomaly. These graphs remove the noise from individual stations to show only the global mean. They then say, "In order to obscure the fact that the global means are small residues of large numbers whose precision is questionable, the common presentations plot the global mean anomalies without the scattered points and expand the scale so as to make the changes look large."

GMST is an area weighted average of station data calculated as an anomaly against a baseline (usually either 30 or 50 years). It is not the average of the anomalies of individual stations, which they acknowledge. And while there's nothing wrong with plotting both on the same graph, there's also nothing wrong with presenting one or the other without showing both. The GMST anomaly does not mean that individual locations can't be significantly above or below the trend line. The fact is that, despite the variability in station data, the global average is warming. Plotting that on a scale that shows it accurately is not misleading, nor does it obscure anything.

Let's look at a simple example. Let's take a hypothetical study of 100 students. We want to know the impact of changing teaching methods on student test scores. So for the first 10 tests, we see that the average test score is 70 with a range of 55 to 85. After employing better teaching methods, the average score improves to 75 with a range of 60 to 90. What should be clear is that the spread in scores is always about 15% more or less than the mean, but the mean still increased by 5%. You expect variability in the scores, but what you're interested in is how well the teaching practices improve the mean. Low scores, means scores and high scores all improve. Likewise, in the LC Figures 3 and 4, the low station anomalies, the mean GMST and the high station anomalies all increase with the variability in station anomalies remaining roughly constant. So in what way is it misleading to plot GMST with a reasonable scale? No actual answer is given.

3. Accuracy of Satellite vs Thermometer Record.  LC also (somewhat ambiguously) claim that the satellite measurements of the lower troposphere are more accurate than surface thermometers. They write, 

The temperature anomaly of the lower troposphere (measured by satellites) relative to the surface temperature is much better sampled and represents the more climate-relevant quantity of heat content, a change in which is a [theorized] consequence of enhanced GHG forcing.

This seems a bit misleading as well. Satellites have only been used to estimate global temperature since 1979, and their accuracy has yet to approach the accuracy of surface thermometers. The uncertainty associated with satellite measurements is about 5x larger than surface thermometers. Differences between the UAH and RSS datasets, for instance, far exceed differences between the major GMST datasets. This means that the thermometer record, which goes back to the late 19th century, is at least as accurate as satellites throughout the entire history of the thermometer record.

Moreover, the most important theorized consequence of enhanced GHG forcing is increased surface temperature. Humanity lives on the surface, and this is the part of the troposphere that affects us the most. But satellites don't measure surface temperature. I have written a couple posts on the accuracy of  the satellite record. In the first, I show that surface thermometers are more accurate (not a controversial conclusion). In the second, I argue (tentatively) that RSS is more accurate than UAH in lower troposphere trends.

Is CO2 to Blame for the Increase in GMST?

What LC seem to be primarily concerned about is policy. They are opposed to a "narrative of a global warming disaster" that is being used to justify implementing "control measures" to mitigate this coming "existential disaster." Words like "existential disaster" are difficult to quantify. What climate policy should be about, imho, is mitigating to decrease the cost of adaptation, thereby saving money and reducing suffering in the long term. I have another post where I argue for this. But here, LC describe a "narrative" that is built on a chain of reasoning leading to a conclusion that we need to implement control measures. The chain of reasoning is:

1. "Changes in this dubious metric (GMST) are almost entirely due to variations in CO2.
2. "Changes in CO2 were primarily due to man’s activities."
3. "We know precisely how to control the level of CO2, and that we know exactly how this will influence the globally averaged temperature anomaly."
4. "All of this implies the likelihood of existential disaster unless the assumed control measures are implemented"

The argument is that taken together, the probability of each of these links in the chain being true is small, so the probability that we need to implement control measures is very small. "In a logical world, it would be understood that the probability of the whole chain will be the product of the probability of each link, and so generally very small. And, of course, if any link is broken (i.e., probability zero), the whole chain is broken." I'll come back to that conclusion, but I think it's important to look at the links in the chain of reasoning and correct the statements given. Let's look at each:

1. Changes in GMST are due entirely to variations in CO2. This is is a strawman. Over the last 120 years, short term internal variability is certainly due to other factors (ENSO, volcanoes, etc). Long-term, however, the increase in GMST is due to CO2 and other GHGs, some of which has been masked by aerosols. Using round numbers, CO2 has caused ~2 W/m^2 warming, other GHGs add another ~1 W/m^2 and aerosol pollution has reduced this total by ~0.8 W/m^2 for a total of about 2.2 W/m^2. Other forcings (solar variability, volcanoes, etc) have been basically negligible over the last 120 years.  CO2 is often singled out because it's by far the largest forcing, and cooling from aerosols has largely canceled out warming from other GHGs. LC assert that "there are quite a few other factors whose common variations are as large as or larger than the impact of changes in CO2, but only offer one example - the area and height of upper level clouds - and they offer no quantification of this. Moreover, changes in clouds is a feedback to warming, the magnitude and even sign of which is some what uncertain, though recent studies suggest is more likely a positive feedback. This means that as GHGs warm the planet, changes in clouds are likely to amplify that warming, but they are not drivers of global warming.  LC offer no examples of any forcing comparable to the 2 W/m^2 increase from CO2. I strongly suspect if the could name one, they would have. So far, nobody has found any forcing near the magnitude of 2 W/m^2

Later in their paper, LC also argue that the sensitivity of GMST to increasing CO2 is small. "As noted by Lindzen and Spencer (2019), presently estimated changes in the temperature record are most consistent with low sensitivity to increases in CO2, and the related warming is likely to be beneficial." This is not a summary statement of the scientific literature. Even the paper cited is just another policy paper published for the CO2 Coalition. It is not a scientific study at all. The estimates in the scientific literature for ECS average around 3 C for doubling CO2. The most definitive study available to date, using multiple lines of evidence, agrees. "The 66% range is 2.6–3.9 K for our Baseline calculation and remains within 2.3–4.5 K under the robustness tests; corresponding 5–95% ranges are 2.3–4.7"[10] And these estimates are consistent with empirical data. In another post, I plotted GMST with radiative forcing for CO2 on the x-axis and found a slope of the best fit line to be a TCR of 2.3 C for doubling CO2. Given recent estimates for EEI, this puts ECS in the neighborhood of 3.3 C for doubling CO2. In other words, empirical data is consistent with central estimates for equilibrium climate sensitivity at ~3 C.

2. Changes in CO2 are primarily due human activity. LC acknowledge this is true over the last 200 years; virtually no climate scientist would question that human activity is responsible for virtually all the increase in CO2 over the last 200 years. From here, though, they claim, "over Earth’s history, there were radical changes in CO2 levels, and these changes were largely uncorrelated with changes in temperature." The footnote is a link to this graph.

I've posted about this before, and I don't think I need to repeat myself much here. I'll just summarize that this graph uses a very outdated schematic of temperature from Scotese' website; it's not a proper reconstruction of global temperatures. It also uses the Berner 2001 climate model, and it plots CO2 in concentration, not radiative forcing. It also fails to account for solar evolution. I find it next to impossible to believe that Lindzen and Christy don't know that the relationship between CO2 and temperature is logarithmic, or that they don't understand that the sun's luminosity has increased over geologic history. They have to know that this graph is not evidence that CO2 and GMST are "largely uncorrelated" in geologic history. Plotting Berner's model correctly with reconstructions of global temperatures within the peer-reviewed literature show the correlation quite well. I have posted about this as well.

3. We know precisely how to control the level of CO2, and that we know exactly how this will influence GMST. We can correct the wording here to make this pretty obvious. Let's state this better. It's pretty clear that we know what we're doing to increase CO2 levels - we're emitting carbon from fossil fuels and land use. We also know that if we stop doing what will increase CO2 levels, we will no longer increase CO2 levels. If CO2 levels stabilize (along with concentrations of other GHGs) they can no longer cause GMST to increase, once climate reaches equilibrium with those forcings. We also know that if we continue to increase GHG levels, we will continue to warm the planet, and we'll have to continue to adapt to the the resulting climate changes.

4. All of this implies the likelihood of existential disaster unless the assumed control measures are implemented. Again, we can correct the wording here to better reflect the actual situation and avoid overly subjective language. Adaptation to AGW is inevitable. The issue is to what extent we should  also mitigate. The argument they are responding to is simply that mitigation + adaptation is in the long run cheaper and does less harm than adaptation alone.

With all this in mind, let's restate these four links in the chain of reasoning accurately and then evaluate LC's conclusion:

1. Virtually all the increase in GMST since the mid-19th Century can be attributed to human activity (GHG emissions and aerosol pollution).[11]
2. Virtually all the increase in in CO2 above preindustrial levels is due to human activity.
3. We know that we can halt the human impact on global warming if we halt our GHG emissions.
4. Long term, adaptation with mitigation is cheaper and does less harm than adaptation alone.

LC argue that these links in a chain of reasoning are each probabilistic, and on that point I don't disagree. However, the first 3 points are not contested much in the scientific literature. Links 2 and 3 are virtually undisputed. Link 1 is almost certainly not wrong, though there's a probability that human activity has caused less than 100% of warming. If human activity turns out to be responsible for 90% of  current warming, then adaptation + mitigation is still likely the better option. If human activity turns out to be responsible for 10% of current warming, then mitigation does little good, but the probability of that being the case is extremely low.

I'm a big proponent of not committing the naturalistic fallacy. Nothing about the empirical data requires us to mitigate and adapt to AGW. We can choose to do what we want. The argument for mitigation, though, is as LC suggest, probabilistic. Given a value most of us have for doing what is most likely to achieve human flourishing while being the most cost effective and causing the least harm, mitigation with adaptation is the best option. The GMST anomaly data we have is clear, and the attribution of that warming to human activity is also clear. There is so much blatant misinformation in this paper that it simply can't be considered an honest argument against mitigation. Certainly people can talk about GMST in misleading ways (and LC's paper is proof), but GMST itself is neither "misleading" nor a "dubious metric." 

References:

[1] Richard Lindzen and John Christy. The Global Mean Temperature Anomaly Record: How it works and why it is Misleading. CO2 Coalition. Dec 2020. https://co2coalition.org/wp-content/uploads/2021/08/Global-Mean-Temp-Anomalies12.08.20.pdf

[2] James Hansen, Makiko Sato, Reto Ruedy. Perception of climate change. Proceedings of the National Academy of Sciences Sep 2012, 109 (37) E2415-E2423; DOI: 10.1073/pnas.1205276109https://www.pnas.org/content/109/37/E2415

[3] Chelsea Harvey. "The Arctic Is Warming Four Times Faster Than the Rest of the Planet." Scientific American. https://www.scientificamerican.com/article/the-arctic-is-warming-four-times-faster-than-the-rest-of-the-planet/

[4] Shakun et al, “Global Warming Preceded by Increasing Carbon Dioxide Concentrations during the Last Deglaciation” Nature 484(7392):49-54 · April 2012https://www.researchgate.net/publication/223987444_Global_Warming_Preceded_by_Increasing_Carbon_Dioxide_Concentrations_during_the_Last_Deglaciation

[5] Osman, M.B., Tierney, J.E., Zhu, J. et al. Globally resolved surface temperatures since the Last Glacial Maximum. Nature 599, 239–244 (2021). https://doi.org/10.1038/s41586-021-03984-4

[6] Stephanie Paige Ogburn, “Ice-Free Arctic in Pliocene, Last Time CO2 Levels above 400 PPM: Sediment cores from an undisturbed Siberian lake reveal a warmer, wetter Arctic.” SA
https://www.scientificamerican.com/article/ice-free-arctic-in-pliocene-last-time-co2-levels-above-400ppm/
M. Robinson, “Pliocene Role in Assessing Future Climate Impacts” Eos 89, No. 49 (2008). https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2008EO490001

[7] Rhian L.Rees-Owen. “The last forests on Antarctica: Reconstructing flora and temperature from the Neogene Sirius Group, Transantarctic Mountains” Organic Geochemistry Volume 118, April 2018, Pages 4-14
https://www.sciencedirect.com/science/article/pii/S014663801730219X

[8] Kemp, D., Eichenseer, K. & Kiessling, W. Maximum rates of climate change are systematically underestimated in the geological record. Nat Commun 6, 8890 (2015). https://doi.org/10.1038/ncomms9890

[9] Lenssen, N. J. L., Schmidt, G. A., Hansen, J. E., Menne, M. J., Persin, A., Ruedy, R., & Zyss, D. (2019). ImpImprovements in the GISTEMP uncertainty model. Journal of Geophysical Research: Atmospheres, 124, 6307– 6326. https://doi.org/10.1029/2018JD029522

[10] Sherwood, S. C., Webb, M. J., Annan, J. D., Armour, K. C., Forster, P. M., Hargreaves, J. C., et al. (2020). An assessment of Earth's climate sensitivity using multiple lines of evidence. Reviews of Geophysics, 58, e2019RG000678. https://doi.org/10.1029/2019RG000678

[11] Gillett, N.P., Kirchmeier-Young, M., Ribes, A. et al. Constraining human contributions to observed warming since the pre-industrial period. Nat. Clim. Chang. 11, 207–212 (2021). https://doi.org/10.1038/s41558-020-00965-9