by Judith Curry
The dueling climate null hypothesis papers by myself and Kevin Trenberth are now online.
Recall we originally discussed the background for these dueling papers on the previous thread Null hypothesis discussion thread.
From the WIREs Climate Science press release:
The Human Cause of Climate Change: Where Does the Burden of Proof Lie?
Dr Kevin Trenberth Advocates Reversing the ‘Null Hypothesis’
The debate may largely be drawn along political lines, but the human role in climate change remains one of the most controversial questions in 21st century science. Writing in WIREs Climate Change Dr Kevin Trenberth, from the National Center for Atmospheric Research, argues that the evidence for anthropogenic climate change is now so clear that the burden of proof should lie with research which seeks to disprove the human role.
In response to Trenberth’s argument a second review, by Dr Judith Curry, focuses on the concept of a ‘null hypothesis’ the default position which is taken when research is carried out. Currently the null hypothesis for climate change attribution research is that humans have no influence.
“Humans are changing our climate. There is no doubt whatsoever,” said Trenberth. “Questions remain as to the extent of our collective contribution, but it is clear that the effects are not small and have emerged from the noise of natural variability. So why does the science community continue to do attribution studies and assume that humans have no influence as a null hypothesis?”
To show precedent for his position Trenberth cites the 2007 report by the Intergovernmental Panel on Climate Change which states that global warming is “unequivocal”, and is “very likely” due to human activities.
Trenberth also focused on climate attribution studies which claim the lack of a human component, and suggested that the assumptions distort results in the direction of finding no human influence, resulting in misleading statements about the causes of climate change that can serve to grossly underestimate the role of humans in climate events.
“Scientists must challenge misconceptions in the difference between weather and climate while attribution studies must include a human component,” concluded Trenberth. “The question should no longer be is there a human component, but what is it?”
In a second paper Dr Judith Curry, from the Georgia Institute of Technology, questions this position, but argues that the discussion on the null hypothesis serves to highlight fuzziness surrounding the many hypotheses related to dangerous climate change.
“Regarding attribution studies, rather than trying to reject either hypothesis regardless of which is the null, there should be a debate over the significance of anthropogenic warming relative to forced and unforced natural climate variability,” said Curry.
Curry also suggested that the desire to reverse the null hypothesis may have the goal of seeking to marginalise the climate sceptic movement, a vocal group who have challenged the scientific orthodoxy on climate change.
“The proponents of reversing the null hypothesis should be careful of what they wish for,” concluded Curry. “One consequence may be that the scientific focus, and therefore funding, would also reverse to attempting to disprove dangerous anthropogenic climate change, which has been a position of many sceptics.”
“I doubt Trenberth’s suggestion will find much support in the scientific community,” said Professor Myles Allen from Oxford University, “but Curry’s counter proposal to abandon hypothesis tests is worse. We still have plenty of interesting hypotheses to test: did human influence on climate increase the risk of this event at all? Did it increase it by more than a factor of two?”
Trenberth. K, “Attribution of climate variations and trends to human influences and natural variability”, WIREs Climate Change, Wiley-Blackwell, November 2011, DOI: 10.1002/wcc.142 http://doi.wiley.com/10.1002/wcc.142
Curry. J, “Nullifying the climate null hypothesis”, WIREs Climate Change, Wiley-Blackwell, November 2011, DOI: 10.1002/wcc.141 http://doi.wiley.com/10.1002/wcc.141
Allen. M, “In defense of the traditional null hypothesis: remarks on the Trenberth and Curry”, WIREs Climate Change, Wiley-Blackwell, 2011, DOI: 10.1002/wcc.145 http://doi.wiley.com/10.1002/wcc.145
JC comments: Read the papers, they aren’t lengthy.
Trenberth’s arguments weren’t unexpected, given his previous essay on this that was published by the AMS. Actually, I don’t see much in Trenberth’s essay that is about the null hypothesis, rather it focuses on attribution of extreme events. Allen doubts that Trenberth’s suggestion for reversing the null hypothsis will find much support, and I have to agree.
The more interesting null hypothesis debate is arguably between my position and Allen’s. Whereas I argue for nullifying the climate null hypothesis as it relates to attribution, Allen argues for preserving the climate null hypothesis.
However, Allen completely misinterprets my argument regarding the null hypothesis. In the abstract, he states “Judith Curry’s counter proposal to abandon hypothesis tests as useless is worse still.” I did not recommend abandoning hypothesis tests. As discussed in my essay, you can test a hypothesis without using a statistical null hypothesis test. My statement about the null hypothesis is made in the context specifically of attribution arguments; it is not a general proposal to abandon hypothesis tests.
As stated in my paper, climate attribution hypotheses are particularly ill-suited for null hypothesis testing. Consider the following hypothesis, H1 (from the IPCC attribution statement):
H1: “Most [>50%] of the observed increase in global average temperatures since the mid-20th century is due to the observed increase in anthropogenic greenhouse gas concentrations.”
Is Trenberth’s statement “there is no human influence on climate” useful as a null hypothesis in the context of H1? It is not, since the statement is generally accepted as trivially false and its falsification lends no support to H1. A more logical null hypothesis for H1 might be:
H0: Less than half of the observed increase in global average temperatures since the mid-20th century is due to the observed increase in anthropogenic greenhouse gas concentrations.
Attempting to formulate a null hypothesis for the IPCC’s attribution statement reveals the illogical nature of H1. The issue regarding the human influence on climate is not a binary yes-no issue, whereby humans influence climate or they do not. The key issue is the importance of the anthropogenic influence on climate relative to the background natural climate variability (both forced and unforced). The binary nature of H0 and H1 implies that the distinction between 51% of the warming attributable to humans versus 49% is somehow significant, meaningful, or important.
Allen defends the use of ‘most’ in the following way: There is nothing imprecise about ‘most’: it means more than half. As it happens, this wording was introduced to replace the (vaguer but more evocative) phrase ‘contributed substantially’ in a nice example of the IPCC review process making its conclusions both more specific and less emotive. As Curry observes, an infinite number of statements could have been made, ranging from ‘it is extremely likely that the anthro pogenic increase in greenhouse gases has caused some warming’ (not very informative, since an infinitesimally small warming is of no policy relevance) to ‘it is about as likely as not that greenhouse-gas-induced warming exceeds the total observed warming’ (which indicates the size of the greenhouse signal, but understates our confidence in attribution). Far from being a ‘poor choice’, in Curry’s words, ‘most’ was chosen for precisely the reasons she advocates: large enough to be policy relevant, while small enough for the null hypothesis ‘not most’ to be rejected at an informative confidence level.
It is useful to refer back to my recent response to the reply to my uncertainty monster paper, where the issue surrounding ‘most’ was discussed in some detail. In this response, I also criticized the IPCC for the lack of traceability regarding ‘most’ and ‘very likely.’ Allen provides a rather astonishing description of the sausage making that went into ‘most’. Quoting from my uncertainty monster reply:
Whereas X et al. disagree with our statement, the IAC Review of the IPCC seems to share our concern: “In the Committee’s view, assigning probabilities to imprecise statements is not an appropriate way to characterize uncertainty.” Assigning a ‘very likely’ likelihood to the imprecise ‘most’ is not an appropriate way to characterize uncertainty.
I continue to point out the problems of posing the attribution hypotheses and/or conclusion in the form of H1. Understanding this point is an issue in basic logic, you don’t need to understand much about climate science to see the problems with formulating H1 in this way. Not to mention the difficulty of formulating a sensible null hypothesis that is not trivially true.
This ambiguity in the attribution of warming to anthropogenic factors becomes magnified substantially in the fractional attribution of individual extreme events to anthropogenic forcing, which is the topic of main interest to Trenberth and Allen.
I’m pleased that WIRE is making these papers publicly available. WIRE has issued a press release, but I will be surprised if this exchange generates much media interest. Nevertheless, this rather arcane debate over the null hypothesis has important implications for the framing of the IPCC’s attribution arguments, and I hope that this exchange will stimulate debate and discussion on this topic within the IPCC community.