This is a joint post between myself and Eric Winsberg, Professor of Philosophy at the University of South Florida. Eric has just published, together with Naomi Oreskes and Elisabeth Lloyd, a paper called Severe Weather Event Attribution: Why values won’t go away. The paper discusses the issue of how one might assess the anthropogenic influence on an extreme weather event. This post describes what was presented in the paper and tries to justify why there may be value in approaching this issue from more than one perspective.
Extreme weather event attribution
One way that we can gain confidence in our understanding of anthropogenic influences on climate is to carry out detection and attribution studies. The basic idea is to consider an extreme event, or a pattern of extreme events, and to establish the probability of that extreme event occuring under current conditions. This can then be compared to what would be expected had we not undergone greenhouse warming, and other anthropogenic changes, to determine some kind of risk ratio.
One advantage of this approach is that it largely avoids false positives; it will only assign some probability of an anthropogenic influence if there is some clear detection and if some of this can indeed be attributed to anthropogenic influences. However, this also means that there will almost certainly be circumstances where we do not assign a probability of an anthropogenic influence when, in fact, such an influence does indeed exist. From a scientific perspective this might be fine; we would simply be waiting for a sufficiently strong signal to emerge. However, this could potentially lead us to under-estimating the impact of anthropogenically-driven climate change.
A complementary approach is to consider a storyline. For example, given that an event has occured, how might climate change have influenced this event? If the air was warmer, then we may expect enhanced precipitation. If sea surface temperatures are high, then we may expect a tropical cyclone to be more intense. The focus here tends to be on the thermodynamics (i.e., the energy) and to take the dynamics as given (i.e., the event happened).
It turns out, though, that the story-line approach has been rather controversial, with many who favour more formal detection and attribution being highly critical. They argue that it could lead to more false positives and that taking the dynamics as given ignores that dynamical factors could actually work to make some events less likely. Essentially, they argue that the storyline approach may over-estimate anthropogenic influences, potentially mistaking natural variability as being anthropogenic.
The problem, though, is that although the two approaches are complemetary, they’re not actually quite addressing the same issue. The detection and attribution approach is essentially trying to determine how anthropogenic-driven climate change influences the probability of a specific class of event. The storyline approach, on the other hand, is more looking at how anthropogenically-driven climate change might have influenced an event that has actually occured. There is no real reason why we should prefer one approach over the other; they can both play an important role in aiding our understanding of how anthropogenic influences impact extreme weather events.
The criticism of the storyline approach seems to have two main strands. One is that the more formal detection and attribution approach avoids the reputational harm that may occur if climate scientists make claims that later turn out to be wrong. The other, is that the storyline approach involves decisions that are likely to be influenced by value-judgements. Given that the detection and attribution framework relies on probabilities, it may be somewhat closer to value-neutral that the storyline approach, but it’s not completely value-free. There are always going to be judgements associated with things like model assumptions and how to present the results.
Also, the judgement that detection and attribution is preferable to the storyline approach is fundamentally value-laden. It’s a judgement that avoiding false positives is preferable to potentially presenting false negatives. Just as incorrectly associating climate change with an extreme event could lead us to investing in infrastructure that turns out to be unnecessary, under-estimating the link between climate change and extreme events could lead to harm that could have been avoided.
From a scientific perspective, the detection attribution approach may well be preferable. However, the storyline approach seems very valuable from a public perspective. It allows us to consider how climate change may have influenced a specific event and also allows us to discuss how it may impact similar events in future. Without it, you run the risk of articles like this one that uses the lack of a detectable trend to conclude that there’s no solid connection between climate change and the major indicators of extreme weather. The storyline approach would almost certainly not lead to such a conclusion.
Overall, it’s hard to see why we shouldn’t be using both approaches. They’re complementary, address slightly different aspects of the link between climate change and extreme events, and should ultimately tend to be consistent. If an expected influence doesn’t emerge, then we’d have to either re-think the storyline, or double check the detection and attribution analysis. However, the storyline approach can also allow us to stress that the lack of a detectable trend doesn’t necessary imply no link between anthropogenically-driven climate change and extreme weather events; absence of evidence is not the same as evidence of absence.
We would clearly like to quantify the impact of anthropogenically-driven climate change on extreme events, but we would also like to avoid concluding that there is no link when in fact such a link is expected and will ultimately become evident. Using both detection and attribution and the storyline approach can help to present an overall picture that best represents our understanding of the link between anthropogenically-driven climate change and extreme weather events.
Severe Weather Event Attribution: Why values won’t go away – Winsberg, Oreskes & Lloyd, Phil Sci Archive, 2019.
Extreme events and anthropogenic emissions – post I wrote a while ago that – I think – makes a similar point to the point being made in Eric’s paper.