One of the oft-cited claims about anthropogenic global warming (AGW) is that ~97% of climate scientists accept that exists. The planet is heating up due to carbon emissions and we are causing it. This is a decently well-supported claim (Politifact broke it down in an nuanced way). What we know for certain is that a high percentage of actively publishing climate scientists do think that humans are causing the planet’s temperature to increase, and as you survey the scientists that publish most often (which roughly correlates with higher expertise), the percentage of scientists who think climate change is caused by humans approaches the oft-cited 97% figure.
While a scientific consensus should closely map the researchers’ opinions and perspectives on a field they know intimately, the consensus is better illustrated through what the aggregate of the individual studies in a field says. One analysis was performed by John Cook in 2013 in Environmental Research Letters, where the team analyzed the abstracts* of over 11,000 studies. This study found that among the abstracts that endorsed an opinion on whether or not humans were causing climate change, 97.1% endorsed the view that AGW exists (while ~2.1% rejected AGW and ~0.8% weren’t sure).
A more recent study** by Rasmus E. Benestad and others (including John Cook) looked into some of the remaining 2% of papers that rejected AGW. Since there was a relatively small sample of papers that rejected AGW, the group looked into 38 of them beyond merely the abstract to determine common themes and any problems they may share. As it turns out, may of the papers suffer from a lot of a lot of common mistakes and themes that impugn the assertions that climate change is false.
The analysis found that the most common error among these papers was that they failed to have proper analytical setup. The studies were founded on questionable or false assumptions, which led to faulty conclusions. Among these problems were assumptions that astronomical cycles were a significant component to AGW, or developing climate models with incorrect parameters. Another common problem was that multiple authors made faulty conclusions by implementing false dichotomies, stating that if a factor besides humans were a contributor to AGW, then humans are not complicit at all. This is unsurprising, as these are common arguments you find from denialists when it comes to global warming. Perhaps these flawed papers are the primary literature where they are getting their information from.
This group also tried to replicate some of the mathematical models these flawed papers implemented, and found flaws in these models as well. In order to fit data to a mathematical model in a meaningful way you should have a proper justification for the type of model you choose, as the underlying physics behind scientific models lead to certain mathematical behaviors and functions. It’s important to justify your model and the reasons behind your choice; even if you create a model with a good fit to your data it’s meaningless if it holds no explanatory power. For example, many data sets can be fit to a polynomial, but this doesn’t necessarily mean that the underlying natural phenomena behind the data follow anything polynomial-like. Also, if you include enough variables in your model, you can get a good fit regardless of whether the model is a good choice or not, resulting in a mistake called “overfitting”. This study found multiple examples of poor fitting, where the questionable studies did not sufficiently evaluate their models or test for consistency.
One particularly egregious example was a poor fit by use of Fourier analysis. Any physics student and many other science students know that literally any finite function can be fit to a sum of periodic functions, known as a Fourier series. Humulum et. al (2011) claimed that they were able to reduce the changes in Earth’s temperature to three cyclical natural timescales (2804, 1186, and 556 years) using this type of series fitting. Again, it is unsurprising that they were able to get a fit, since Fourier series are designed to fit to any function. When this critical study by Benestad looked at the same data, but increased the timescale to account for temperatures outside the fitted region, they were unable to reproduce the same cycles.
The authors of this study point to some of the common causes of why these papers got through peer review, stating that it’s most likely that these authors were out of their depth within the field of climate research. Furthermore, some of the journals that published these papers were not climate change journals, and may have been unable to find proper reviewers that could analyze these papers with the right kind of rigor.
Interestingly, many of the speculations and conjectures posited by the papers studied here were incompatible with other papers that also held the non-consensus view denying AGW. This reminds me of Flat Earthers and other conspiracy theorists who don’t necessarily gather together consistently with other deniers, but still often band together anyway with a shared distrust of the mainstream view.
The authors conclude by stating that we need to consider all views in research, including the non-mainstream ones, but we need to do it transparently and clearly, with robust methods. Many of these papers cut out inconvenient data, and many of the mistakes made were simply due to those without experience. The authors note that data in these papers have drastic impact on our society’s decisions, and they need to be scrutinized diligently and honestly in order to make sure our actions in the future are as informed as possible.
It’s obviously necessary to scrutinize these non-mainstream views before they are disseminated to the public. Earlier this year, the Heartland Institute sent out propaganda to schools to sell the idea that there isn’t much consensus on climate change. The authors of the Heartland report tried to dismiss papers such as Cook’s original ~97% study, and attempted to find some papers that refuted the consensus, or at least cast doubt on certain established findings. Many of the same authors and a few of the papers used within the Heartland report to prop up a denialist narrative were shown to have flaws in this Benestad study. If our educators end up teaching material from a political think tank heavily invested in muddying the waters for political gain by using incompetent or similarly biased authors, that could cause many problems down the road.
Like the authors, I think academic debate is healthy and important among experts, including views that don’t always align with the mainstream. But we have to make sure the people collecting the data are doing it the right way, that they are transparent about the data they are using, and we must make certain that they are reviewed and picked apart by similarly credentialed experts. I do think that we need to emphasize the urgency towards taking action even in the face of uncertainty, especially if AGW is as much of a threat that the other 97% of papers seem to imply. This study shows that there are many systemic flaws in studies that buck the consensus, and perhaps there is a reason for that. We have good reason to believe that climate change is a threat, we are causing it, and we need to do something about it soon.
*the abstract is a short segment, usually one paragraph, that summarizes what research occurred in a paper as well as its most important findings in a concise way.
**This was published in November 2016, but for some reason news outlets are just now getting around to reporting on this item. Perhaps there was a lot of news in November 2016 that overshadowed items like this.