[Update 1-23-14 1600 GMT: Judith Curry explains her logic. I've added it to the end of this post, with my commentary.]
Several times during the past few months, I’ve heard generally incredulous statements such as, “How can the IPCC increase their confidence in anthropogenic global warming at the same time their model projections are diverging farther and farther from reality?”
Specifically, the Intergovernmental Panel on Climate Change, in their Fifth Assessment Report (AR5) Working Group One (WG1) Summary for Policymakers, said:
“It is extremely likely (>95% probability) that human influence has been the dominant cause of the observed warming since the mid-20th century.”
(Note: probability percentages are added from the IPCC’s definition of likelihood terms. Emphasis added.)
They continued with a more precise, specific statement:
“It is extremely likely (>95% probability) that more than half of the observed increase in global average surface temperature from 1951 to 2010* was caused by the anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together.”
This is an increase in confidence since the previous report (out in 2007), which said:
“Most of the observed increase in global average temperatures since the mid-20th century is very likely (>90% probability) due to the observed increase in anthropogenic greenhouse gas concentrations.”
This in turn is an increase in confidence since the Third Assessment report (out in 2002), which said:
“Most of the observed warming over the last 50 years is likely (>66% probability) to have been due to the increase in greenhouse gas concentrations.”
And so on.
Yet, we’re in the middle (or perhaps the end, or perhaps the beginning) of a hiatus in the rise of global temperatures. The evidence seems to be mounting that natural variability is more important than the IPCC reports had previously contemplated, yet the IPCC’s confidence in anthropogenic global warming grows stronger.
The people who ask me such questions seem to think that the IPCC’s increase in confidence is politically motivated, that the process requires a unidirectional progression of confidence in such a key area, and that this provides reason enough to account for the IPCC going against the mounting evidence. Indeed, to them this seems like a no-brainer that proves that the whole IPCC process is corrupt.
Now, I don’t know the IPCC’s reasoning process, to the extent that a complex organization can even be said to have a reasoning process. However, those who think that the IPCC is acting unscientifically in this particular matter are failing to recognize something particularly important: despite the growing divergence, the evidence supports increased confidence in the IPCC’s statement.
Judith Curry’s Argument
How can this be? To see how, I’ll analyze Prof. Judith Curry’s recent written testimony for the hearing on Obama’s Climate Action Plan at the Senate Committee on Environment and Public Works on January 16, 2014. Her testimony provides the most comprehensive, scientifically accurate, and seemingly logical example of the argument that the IPCC’s increase of confidence is counterfactual. The complete written testimony is available here; I’ll be extracting the key arguments.
Curry’s testimony focus on three issues; the issue at hand is her first focus.
“The AR5 statement of ‘extremely likely’ implies that the overall arguments have strengthened. However, several key elements of the AR5 WGI report point to a weakening of the case for attributing most of the warming to human influences, relative to the previous assessment AR4 (2007):
• Lack of warming since 1998 and the growing discrepancies between observations and climate model projections
• Evidence of decreased climate sensitivity to increases in atmospheric CO2 concentrations
• Evidence that sea level rise during 1920-1950 is of the same magnitude as in 1993-2012
• Increasing Antarctic sea ice extent”
Curry summarizes the first point as follows:
“• After expecting a global mean surface temperature increase of 0.2C per decade in the early decades of the 21st century based on climate model simulations and statements in the AR4, the warming over the past 15 years is only ~0.05C.
• The IPCC AR5 bases its surface temperature projection of 0.10 to 0.23C per decade for the period 2016-2036 on expert judgment, which is lowered relative to the climate model results that predict substantially greater warming
• The IPCC does not have a convincing or confident explanation for the current hiatus in warming.”
All of these summary bullets are correct. No disagreement there.
Curry summarizes the second point as follows:
“…[T]he AR5 reflects greater uncertainty and a tendency towards lower values of the ECS than the AR4. The discrepancy between observational and climate model-based estimates of climate sensitivity is substantial and of significant importance to policymakers — sensitivity, and the level of uncertainty in its value, is a key input into the economic models that drive cost-benefit analyses and estimates of the social cost of carbon.”
This summary is correct. No disagreement there either.
Curry’s third point involves sea level rise, and what is causing it. I find her discussion of sea level rise to be weak, but it’s irrelevant to the issue of what’s causing the increase in global temperatures. Rather than get sidetracked, I’ll leave it at that.
Curry’s fourth point involves sea ice changes (loss in the Arctic, gain in the Antarctic) and the lack of understanding of the importance of various contributing factors to those changes. I agree with her discussion here too, but like her third point, it’s irrelevant to the issue of what’s causing the increase in global temperatures.
Okay, so far, no fundamental disagreements, and in the areas that are relevant for the issue at hand, I completely agree. So where’s the problem?
Let’s look at her three-paragraph summary. Paragraph 1:
“Multiple lines of evidence presented in the IPCC AR5 WG1 report suggest that the case for anthropogenic warming is weaker than the previous assessment AR4 in 2007. Anthropogenic global warming is a proposed theory whose basic mechanism is well understood, but whose magnitude is highly uncertain. The growing evidence that climate models are too sensitive to CO2 has implications for the attribution of late 20th century warming and projections of 21st century climate.”
“If the recent warming hiatus is caused by natural variability, then this raises the question as to what extent the warming between 1975 and 2000 can also be explained by natural climate variability.”
And finally, paragraph 3:
“The stadium wave hypothesis predicts that the warming hiatus could extend to the 2030’s. Based upon climate model projections, the probability of the hiatus extending beyond 20 years is vanishing small. If the hiatus does extend beyond 20 years, then a very substantial reconsideration will be needed of the 20th century attribution and the 21st century projections of climate change.”
I do disagree here that a “very substantial reconsideration” will be needed. Maybe it will, maybe it won’t. It all depends on the cause of the hiatus.
The Crux of the Matter
But here’s why none of this means that the warming over the past 60 years is less likely anthropogenic. Suppose we make a generous assumption for the role of natural variability. Suppose we assume that the stadium wave hypothesis is correct. As Curry alludes to, the stadium wave hypothesis posits that there is a regular, propagating, approximately 60-year oscillation in the climate system that, among other things, affects global temperatures. Let’s suppose that it has a pronounced effect, as big an effect as Curry thinks it might have, and is thus responsible not just for the present hiatus but also the hiatus centered on the 1950s.
In this best-case-for-natural-variability scenario, natural variability is strong enough to bring global warming to a halt during the negative portion of the natural oscillation, and the present halt is likely to continue for at least another decade. Conversely, during the positive portion of the natural oscillation, natural variability and anthropogenic forcing are working together to drive temperatures upward.
Curry alludes to the 1975-2000 period in her testimony. During this period, if we make natural variability as strong as it can be, natural variability would have accounted for about half of the observed warming, and anthropogenic forcing the other half. Since (using GISTEMP) global temperatures went up by about 0.5 C during this period, one would say that natural variability and anthropogenic forcing each accounted for about 0.25 C of warming. That’s 0.1 C per decade, over 25 years.
Conversely, consider a hiatus period such as the 1940s to 1970s. During this period, natural variability is posited to be just as strong, and it would have tended to cause 0.25 C of cooling. Anthropogenic forcing, meanwhile, was counteracting that, producing 0.25 C of warming, yielding zero net change in global temperatures. (During that same period, solar variations would have caused warming while volcanic activity would have caused cooling. For the sake of this example, let’s assume they separately cancel each other.)
Now suppose we want to isolate the effect of anthropogenic forcing. If the stadium wave is the primary manifestation of multidecadal natural variability, we can do this by simply considering a period equal to one complete cycle of the stadium wave, or 60 years. In 2010, the stadium wave would be in about the same phase as it was 60 years earlier, in 1950. Whatever contribution it’s making to global temperatures in 2010, it made the same contribution 60 years ago. That means that the overall temperature change over the past 60 years is entirely anthropogenic. At a pace of 0.1 C/decade, that would be 0.6 C in 60 years, which is what is shown by GISTEMP.
Too fast for you? Let’s try again, based on general principles. If natural cycles are regular and repeatable, the net temperature change over one complete natural cycle will be approximately zero. The warming during part of the cycle is cancelled by cooling during the other part of the cycle. What’s left is the long-term rise caused by man.
Why does the IPCC conclude that the long-term rise is caused by man? The primary logic is simple, really. Of all the things driving long-term changes in the climate system, the biggest by far over the past 60 years is greenhouse gases. Second on the list is particle pollution, or aerosols, which partly counteract the greenhouse gases. Over the past 60 years, natural forcings (sun, volcanoes) have also had a cooling effect. So arguments over the relative importance of different kinds of forcing don’t really matter for explaining the past 60 years of temperature rise: the only large one on the positive side of the ledger is greenhouse gases.
Of course, it’s not enough to say that greenhouse gases point temperature in the right direction. The magnitudes have to match, also. Here, too, the hiatus increases confidence that there’s not some unknown but significant positive forcing agent other than greenhouse gases that’s driving temperature. The smaller the rate of warming, the smaller the possibility that a separate, additional cause of warming is being missed, and that, therefore, greenhouse gases account for most or all of the total amount of warming.
It’s All a Matter of Timing
Curry is correct that the hiatus and increasing model-observation divergence is evidence that the models are too sensitive to forcing agents, including carbon dioxide. She is also correct that, the stronger and longer-lasting the divergence, the greater the evidence that much of the warming pre-2000 was natural. Likewise, the stronger the evidence that models might be overestimating future warming. Exploring those issues requires evaluation of all the available evidence, not just the last 60 years of temperatures, and will have to wait for another blog entry.
But here’s the thing. If, over 60 years, natural variability averages out to zero, it doesn’t matter how strong natural variability is compared to man-made climate change, what’s left over is the man-made part. Thus the IPCC can and should consider it to be extremely likely that human influence dominates the net rise in temperature over the past 60 years.
Note that such a statement only works if you look at the temperature change over one complete natural cycle. Over a shorter period, particularly one in which a natural cycle would have contributed to net warming, confidence would necessarily be lower.
Now look back at the TAR statement. It considered the temperature change not from 1951 to 2010, but from 1951 to 2000. Over that interval, under the natural cycle hypothesis, much of the warming would have been natural. Thus, if one were to make a similar assessment today, one might indeed conclude that the dominance of man-made effects over 1951-2000 is only highly likely, or perhaps merely likely, much as the IPCC assessed things at the time.
For this is another thing that the critics of these statements miss: they apply to different periods. It is possible, indeed it’s strongly supported by the evidence, that the relative contribution of man is larger over the 1951-2010 period than the 1951-2000 period, or especially the 1975-2000 period.
If you’re a believer in strong natural variability, and you’re looking to criticize the IPCC, you might complain that they don’t reduce their estimates of climate sensitivity enough, or that they don’t adequately discuss the increased evidence of the importance of natural variability in affecting temperatures from decade to decade. But none of this, and none of the evidence presented by Curry or by anyone else I’ve seen, reduces the likelihood that the temperature change over the past 60 years or so was mostly anthropogenic.
*I interpret the increase of temperature from 1951 to 2010 to mean not a linear regression rate of temperature rise, but the absolute change of temperature from the value typical of years around 1951 to the value typical of years around 2010, smoothed to remove short-term El Nino effects and the like. On its face, the statement refers to the amount of temperature change, not the rate of temperature change.
Judith Curry’s latest blog post focuses on my analysis, and includes an explanation of her reasoning. Here’s the relevant part, with my comments:
The logic behind my reasoning is this.
The way the IPCC’s attribution argument is laid out, there are two possible contributing causes to climate change since 1950: anthropogenic forcing, and natural variability. The sum of these two contributing causes is 100%; the chief issue of interest is the relative percentage contributions of anthropogenic forcing and natural variability.
The IPCC AR5 makes an extremely confident statement that ‘most’ of the warming is attributed anthropogenic forcing, and I understand ‘most’ to cover a range of 51-95%. The IPCC implicitly recognizes that the attribution issue is uncertain by not giving a distribution of values or a ‘best estimate.’ Rather, they provide a bounded region that covers ~44% of the possible territory.
[N-G: It can be a bit misleading to express this in terms of percentages. When most people see percentages, they imagine small positive numbers that collectively add up to 100%. However, different agents of climate change can have positive (example, increasing greenhouse gases) and negative (increasing aerosols) contributions. Pick a random time interval, and natural variability is just as likely to make a negative contribution as a positive one.
Curry's assumption that the IPCC means 'most' in this context to cover a range of 51-95% is flat-out wrong. Here's the full quote from the Summary for Policymakers: "It is extremely likely that more than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by the anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together. The best estimate of the human-induced contribution to warming is similar to the observed warming over this period." They explicitly state that their best estimate for the human-induced contribution is about 100%, which is outside the range that Curry assumes they mean!
If you're going to impute a range for them, at least have their most likely value somewhere near the midpoint of the range. For the sake of argument, a reasonable range would be 51-135%.]
In the absence of a compelling theoretical reason for the lower bound of 51%, I infer that the IPCC does not regard a high likelihood for values of anthropogenic forcing less than 60%; otherwise it would be difficult to argue that the lower bound would not be breached. At the time of the AR4, in talking with IPCC lead authors I had the sense that most of the scientists thought that the anthropogenic contribution was >90% (this is my subjective assessment; I would be interested in other documentation on this).
[N-G: The documentation is in the statement I just quoted above, from the Summary for Policymakers.]
So for the sake of starting somewhere in my argument, lets start with this breakdown for the IPCC AR5 statement on attribution since 1950:
[N-G: For the IPCC, it's roughly anthropogenic=100%, natural=0%, according to the SPM.]
The elements of greatest certainty in this argument are the components of anthropogenic forcing. The elements of greatest uncertainty are the sensitivity to CO2 (and the fast thermodynamic feedbacks), natural internal variability, and then there are also the unknowns such as solar indirect effects.
The uncertainty in sensitivity to CO2 forcing is acknowledged by the IPCC by the lowering of the bottom bound in their ‘likely’ range; this points to lowering the proportion of anthropogenic component of the attribution. The ‘pause’ is raising awareness of the importance of natural internal variability. Other data reflects the importance of natural internal variability, which N-G didn’t pick up on as being relevant: the importance of natural variability in 20th century sea level rise as indicated by the mid-century bump; and the importance of natural internal variability in determining both arctic and antarctic sea ice extent. With regards to arctic sea ice decline, I spotted this statement in AR5 Chapter 10:
Comparing trends from the CCSM4 ensemble to observed trends suggests that internal variability could account for approximately half of the observed 1979–2005 September Arctic sea ice extent loss.
So . . . this evidence all acts to lower the contribution of anthropogenic forcing, and increase the contribution of natural variability, and I would argue that this lowering should not be regarded as trivial.
[N-G: Sorry, I still don't see the connection. We know that global temperatures rose rapidly from 1910 to 1940, and that that rise was at least partly natural, so what does the knowledge that sea level too may have risen rapidly during that period add to the mix? Nothing. We also know that the smaller the geographical area (as in the Arctic or Antarctic vs. the globe), the bigger the influence of natural variability.]
Yes the 44% provides lots of wiggle room, but it is not unreasonable to infer, based on the evidence provided by the IPCC that the anthropogenic component has dropped below 70% and even below 60%.
[N-G: Another mistake. The Summary for Policymakers is ...wait for it... a summary. It is intended to summarize the evidence in the larger WG1 report. The judgment that "...it is extremely likely that more than half..." etc., incorporates the evidence in AR5 that acts to lower the apparent contribution of anthropogenic forcing. Modifying the summary statement on the basis of evidence that's already in the report is double-counting.]
In this light, putting an extremely confident lower bound of 51% seems insupportable. Particularly in light of the unknowns such as solar indirect effects.
[N-G: I agree with Curry that "extremely likely" would be hard to justify if the central estimate of the contribution from greenhouse gases was somewhere around 55%-65%. However, as I have shown, their central estimate is around 100%, even considering the evidence about sea level rise and natural variability near the poles.]
N-G’s reasoning about how to think about the relative contributions of natural vs anthropogenic anthropogenically is something like the way I have approached this, but not quite. I look at it the following way. Consider two periods: 1975-1998 (warming), and 1998-2013 (hiatus). Play with the percentages of natural variability (assuming warming in the first period and cooling in the second period) and anthropogenic forcing, accounting for the relative lengths of the two periods, and see what percentage breakdown works for both periods. And what the implications are of the hiatus extending another 10 years. You will not get numbers that exceed 75% for anthropogenic.
If I were given a 44% range to work with, I would put the range at 28-72% anthropogenic. My range overlaps with the IPCC in domain 51-70%, but I also allow for numbers below 50%. The main uncertainties seem all in the direction of increasing the contribution from natural variability.
[N-G: The goalposts have moved. The IPCC's attribution statement is for the period 1951-2010, and it was for 100% anthropogenic. If we analyze a different period that includes all of the warming part of a natural cycle and half of the cooling part of a natural cycle, the anthropogenic contribution would indeed be less than 100%. Also, as noted before, the IPCC's range is not 44%.]
Bottom line: with the growing recognition of the importance of natural variability, it is increasingly difficult to defend the bottom bound of 51% for anthropogenic forcing, hence I find the increase in confidence to ‘extremely likely’ to be unjustified.
[N-G: The final total, as I see it: One error of fact (the IPCC's best estimate for anthropogenic contribution is ~100%, not <95%), one error of logic (double-counting evidence that the IPCC has already included), and one case of the answer being different because the question is different (1951-2010 vs. 1975-2013). Thank you, Dr. Curry, for explaining your reasoning in sufficient detail that the specific places where we diverge are clear.]