Lindzen and Choi 2011 - Party Like It's 2009
Posted on 6 July 2012 by dana1981
We previously discussed Lindzen and Choi 2009 (LC09) which used Earth Radiation Budget Experiment (ERBE) instrumental measurements of shortwave and longwave radiation fluxes and observed sea surface temperature (SST) variations in the tropics to evaluate the overall radiative feedback of the climate to SST changes. The study essentially looked at how much energy escapes into space as the tropics warm, and concluded that climate sensitivity to increasing CO2 is very low (in the ballpark of 0.5°C for a doubling of atmospheric CO2).
As we noted in our discussion of LC09, the paper contained a number of major flaws. Lindzen himself has even gone as far as to admit the paper contained "some stupid mistakes...It was just embarrassing." Lindzen and Choi attempted to address some of those issues in a new paper, Lindzen and Choi 2011 (LC11), which they submitted to the Proceedings of the National Academy of Sciences (PNAS). LC11 was very similar to LC09, but used both ERBE and Clouds and the Earth’s Radiant Energy System (CERES) data.
PNAS editors sent LC11 out to four reviewers, who provided comments available here. Two of the reviewers were selected by Lindzen, and two others by the PNAS Board. All four reviewers were unanimous that while the subject matter of the paper was of sufficient general interest to warrant publication in PNAS, the paper was not of suitable quality, and its conclusions were not justified. Only one of the four reviewers felt that the procedures in the paper were adequately described. As a result, PNAS rejected the paper, which Lindzen and Choi subsequently got published in a rather obscure Korean journal, the Asia-Pacific Journal of Atmospheric Science.
As PNAS Reviewer 1 commented,
"The paper is based on...basic untested and fundamentally flawed assumptions about global climate sensitivity"
Here we will discuss the untested and flawed assumptions in LC11 identified by the PNAS reviewers.
Flaw #1: Comparing Tropical Apples with Global Oranges
The first unjustified assumption in LC11 noted by Reviewers 1, 3, and 4 is that correlations observed in the tropics reflect global climate feedbacks. LC11 only examined data in the tropics (20° South to 20° North), but used that limited data to draw conclusions about global climate sensitivity without providing any real justification as to why tropical responses are representative of global changes.
This was also a problem in LC09, as we previously discussed. A great deal of energy is exchanged between the tropics and subtropics. For example, Murphy et al. (2010) found that small changes in the heat transport between the tropics and subtropics can swamp the tropical signal. Murphy et al. and Chung et al. (2010) concluded that climate sensitivity must be calculated from global data.
To be fair, in the version of LC11 published in the Asia-Pacific Journal of Atmospheric Science, Linden and Choi did attempt to use global CERES data from 2000 to 2008, and concluded "the use of the global CERES record leads to a result that is basically similar to that from the tropical data in this study." However, Lindzen and Choi also noted that the global CERES record contains more noise than the tropics-only data, and they do not explain why their results differ from those of Chung et al.
Flaw #2: Assuming Short-Term Local Feedbacks are Representative of Long-Term Global Feedbacks and Cherrypicking Noisy Data
Similarly, Reviewer 3 noted that LC11 focuses on short-term local changes which might not be representative of equilibrium climate sensitivity, because for example the albedo feedback from melting ice at the poles is obviously not reflected in the tropics. LC09 and LC11 only looked at short-term local effects and then compared them to long-term global sensitivity, which as Reviewer 3 noted, is yet another "apples and oranges" comparison.
Reviewer 2 expressed two main concerns about the paper, the first of which was essentially echoed by Reviewer 4.
"The first concern is that month-to-month variability of the tropics may have nothing to do with climate feedback processes. Although the paper acknowledges this in its introductory sections, the conclusion is greatly overstated as applying directly to CO2 driven climate change."
This is similar to an issue noted by Trenberth et al. (2010), who found that the LC09 low climate sensitivity result is heavily dependent on the choice of start and end points in the periods they analyse. Small changes in their choice of dates entirely change the result. Essentially, one could tweak the start and end points to obtain any feedback one wishes (Figure 1).
Figure 1: Warming (red) and cooling (blue) intervals of tropical SST (20°N – 20°S) used by Lindzen et al 2009 (solid circles) and an alternative selection proposed derived from an objective approach (open circles) (Trenberth et al 2010).
As with the Murphy et al. (2010) conclusion regarding tropical data being insufficient to draw global conclusions, Lindzen and Choi simply did not address this fundamental problem in their analysis, carrying the problem from LC09 into LC11.
Flaw #3: Insufficiently Clear Methodology
Reviewer 2 also had major concerns with the methodology in the paper, which were not explained in sufficient detail to reproduce the LC11 analysis. The results of the paper (that climate sensitivity is less than 1°C for doubled atmospheric CO2) are quite radical, since virtually all other research using many different lines of evidence finds that climate sensitivity is very likely between 2 and 4.5°C for doubled CO2.
The long and short of it is that extraordinary claims require extraordinary evidence, and LC11 does not provide anything remotely like extraordinary evidence. Reviewer 2 specifically noted:
"I am very concerned that further analysis will show that the result is an artifact of the data or analysis procedure."
Flaw #4: Failing to Address Contradictory Research
A general problem incorporated into several of the previously discussed flaws is a failure of Lindzen and Choi to address the contradictory results from the Trenberth, Murphy, and Chung et al. groups in 2010. The entire purpose of LC11 should have been to address those subsequent results which had conflicted with LC09. Instead, all reviewers voiced the same concerns LC11 did not address the problems identified in LC09. Reviewer 3 said:
"I feel that the major problem with the present paper is that it does not provide a sufficiently clear and systematic response to the criticisms voiced following the publication of [LC09]"
For example, Chung et al. (2010) attempted to reproduce the LC09 analysis using near-global data, and found net positive feedback. Trenberth et al. also tried to reproduce those results and obtained the opposite result as LC09 - yet LC11 did not explain how they could arrive at the opposite conclusion as these two other studies while using the same data.
Flaw #5: Cloud Causality
An additional flaw noted by Reviewer 1 involved LC09 and LC11's treatment of clouds as an internal initiator of climate change, as opposed to treating cloud changes solely as a climate feedback, as most climate scientists do.
"the authors go through convoluted arguments between forcing and feed backs. For the authors' analyses to be valid, clouds should be responding to SST and not forcing SST changes. They do not bother to prove it or test the validity of this assumption. Again this is an assertion, without any testable justification."
Lindzen and Choi plotted a time regression of change in top of the atmosphere energy flux due to cloud cover changes vs. SST changes and found larger negative slopes in their regression when cloud changes happen before surface temperature changes, vs. positive slopes when temperature changes happen first. LC09 and LC11 thus concluded that clouds must be causing global warming.
However, Dessler (2011) plotted climate model results and found that they also simulate negative time regression slopes when cloud changes lead temperature changes. Crucially, SSTs are specified by the models. This means that in these models, clouds respond to SST changes, but not vice-versa. This suggests that the lagged result first found by Lindzen and Choi is actually a result of variations in atmospheric circulation driven by changes in SST, and contrary to Lindzen's claims, is not evidence that clouds are causing climate change, because in the models which successfully replicate the cloud-temperature lag, temperatures cannot be driven by cloud changes.
Reviewer 2 also noted a quandry for Lindzen and Choi with regards to their assumption that clouds are driving SST changes and their attempts to estimate climate sensitivity:
"If the cloud variations are driving the SST, then these data are not appropriate for computing climate feedbacks, as they are disequilibrium forced fluctuations."
LC11 - Overhyped and Under-Supported
Ultimately the main flaws in LC11 are the same as those in LC09 - Lindzen and Choi simply did not address most of the problems in their paper identified by subsequent research, and what few issues they did address, they failed to explain why their results differ from those who attempted to reproduce their methodology.
Nevertheless, LC09 and LC11 have become extremely over-hyped. Frequently climate contrarians (for example, Christopher Monckton and John Christy) claim that mainstream climate sensitivity estimates rely wholly on models (which is untrue), whereas lower climate sensitivity results are based on observational data. When they make this assertion, they are referring to LC09 and LC11.
This is a key point for climate contrarians, whose arguments are effectively a house of cards balanced atop the 'low climate sensitivity' claim. Since the body of research using multiple different approaches and lines of evidence is remarkably consistent in finding an equilibrium climate sensitivity of between 2 and 4.5°C for doubled CO2 (whereas a 'low' sensitivity would be well below 1.5°C), climate contrarians reject the body of evidence by (falsely) claiming it is based on unreliable models, and attempt to replace it with this single study by Lindzen and Choi under the assertion that it is superior because is observationally-based.
However, subsequent research identified a number of fundamental errors in LC09 which simply were not addressed in LC11, which is why the PNAS reviewers - even those chosen by Lindzen himself - unanimously agreed that the journal should not publish the paper. While LC09 and LC11 are based on observational data, they also rely on a very short timeframe, mainly on data only from the tropics, and their methodology contains a number of problems.
Quite simply, this one paper is insufficient to overturn the vast body of evidence which contradicts the 'low climate sensitivity' argument.
The information in this post has been incorporated into the rebuttal to Lindzen and Choi find low climate sensitivity. Thanks to Kevin Trenberth and Andrew Dessler for their feedback on this post.