Perils and Pitfalls of Event Attribution
Perils and Pitfalls of Event Attribution
Some of you may have noticed a story that originated in the Green Blog by John Rudolf on the New York Times website (March 9, 2011) about the Russian heat wave in the summer of 2010. The news story reports on a paper to appear in Geophysical Research Letters by Randy Dole and co-authors who conclude that in the historical record there is evidence of similar events of comparable intensity. It follows, they argue, that the Russian heat wave cannot be attributed to climate change – rather it is a very rare event. (Paper at GRL website, NOAA Description of Dole et al. article, Jeff Masters blog and analysis) For a variety of reasons I followed how this story propagated around the blogs and news services for the next 24 hours. It was picked up by many sites including, quickly, by the (according to comment writers on this blog) mysterious Steven Goddard (any more on that story?).
As it happens, I am writing an article for Earthzine with Christine Shearer on how scientists and the media engage each other about extreme events (Shearer blog on WU). When it is ready, I will proudly announce it and provide a link. That article will focus on a sociological analysis of extreme weather and the media. This blog will touch on a couple of the issues we raise in that article, but mostly it will be a scientist's point of view on the discussion of the value of pursuing the attribution of single events to climate change in a context largely described by public discourse.
Event Attribution: A public question that arises after every new extreme event is: can this event be attributed to climate change? At this point in time, I cannot imagine the answer to that question ever being, convincingly, yes. Scientists often rely on the statement: no single event can be attributed to climate change, but this event is not inconsistent with climate change. I have used that answer; perhaps, I repeat the mantra (Pakistan: A Climate Disaster Case Study). On thinking about that answer, it is more than useless. But then thinking about the question, it is, depending on your point of view: a natural question, a naïve question, an ill-posed question, or a leading question.
Why do I say that I cannot imagine the answer to such an event attribution question being convincingly yes?
Dole et al. study attribution, and they do it magnificently. Their strategy is to do a physical, statistical, and process analysis of historical information. If they find like events in the historical data, then that makes it impossible to attribute the event, wholly and solely, to climate change. This implies an odd metric: an event that is “caused” by climate change must be different than any event that has been previously measured. Do we have to have some Day After Tomorrow event where physical principles are suspended and the world moves to a whole new set of behavior?
The probability that looking through all of the observations, all of the history, that you are going to find a “like event” is high. I say “like event,” because there will be some differences no matter what. Of course, it has been hot in Moscow before, so there is some atmospheric pattern that yields “hot in Moscow.” We find like events and then, maybe, the current event is 10 degrees hotter and two weeks longer; it’s a obvious record. But is it climate change?
More likely than a obvious record, there will be another event that is similar, about the same, but not quite. Then it becomes the same question as, was Henry Aaron better than Babe Ruth? Aaron hit more home runs, but there are lots of other differences that experts point to and argue about: length of season, quality of pitching, … . Throw in Barry Bonds and Mark MacGwire; they hit a lot of home runs. Well maybe the physics (or physiology) of Bonds and MacGwire are different? Is climate change weather on steroids?
Suppose you look through the record and find that the current event is 10 degrees warmer and 2 weeks longer. Is it climate change? Do you know whether or not that if you had just one more year of observations, that you would not find out that that next year had a similar hot period. What about similar events in the medieval warm period? The data system was relatively sparse 100 years ago; maybe we just missed the event. So even if we find an event that is more intense, more persistent, then we have the problem - have we really observed the historical extremes? Have we observed all natural variability? This will always challenge the public and political discourse on event attribution - always.
More likely than finding an event that is extraordinarily different, we find an event that’s about the same length of time, but one degree warmer. Is the thermometer good enough? Are the instrument sites good - have they changed? What about the urban heat island? What about regional water management projects? Good scientific investigation and analysis can account for these issues, but in any event they are sources of differences, which as in the Aaron versus Ruth argument, are irreducible. Perhaps an extreme record can be established, but then, would that be climate change?
It is hard to see how playing the game of defining extreme events and then attributing that event to “climate change” can ever be won. It is often possible to isolate with statistical certainty descriptions that the emissions of greenhouse gases have influenced an event, but that represents one of those paths of nuanced explanation. Such nuanced explanation, again, assures there is not a definitive "yes" in the public and political discourse. In fact, it seems like it is a game that necessarily leads to controversy, and controversy is the fuel of talk radio, blogs propagating around the world, and the maintenance of doubt.
But what about that question of attribution? Let’s say you find an event that is rare, that is extreme, but not a new record - does that really say that the event today, right now, is not climate change?
In a very basic, old fashioned way, weather and climate are different descriptions of the same thing. They depend on how we, somewhat arbitrarily, define how we want to organize the observations. Crudely, we average weather to make climate. Since we work from the premise that climate change will be slow, for the most part the same type of weather events will make up the old (natural) climate and the new (changed) climate. Over time, the frequency of events will change, what were rare events in the old climate, might just be less rare events in the new climate. I pose, however, that even in a world that is on average four degrees warmer than today, there will be a seventy two degree, sunny day in the spring in Washington D.C. Do we then march through the days 50 years from now and say, “old climate,” “new climate?” The idea of isolating a single event on a single day or a persistent event and asking if it is caused by climate change – does that make sense? Is it even meaningful given the definition of climate? How did we arrive at the question of climate change being a causative of a weather event?
I want to restate the previous paragraph in a different way. Let’s assume that climate is averaged weather. Then climate is defined by a mean, a standard deviation, and a set of more sophisticated parameters that describe statistical distributions. What we have come to call the natural climate is defined by certain values of the mean and measures of deviations from the mean. The future, changed, warmer climate will have different values of the mean and the measures of deviations. With the presumption that the warming of the climate is incremental, then the majority of the events in the warmer climate will be like the events in the “natural” climate. Therefore, just because a like event existed in the "natural" climate does not mean that the current event is not part of the "changed" climate. There are NOT two climates - a natural one and a changed one - with our job being to determine if we have flipped from one to another. When we say that there will be more extreme events in the changed climate, it does not necessarily mean there will be a relentless unwavering string of records. There will, perhaps, be more events that have been previously rare. But, it is not climate change causing weather events.
As you study climate change, it becomes clear that talking about independent isolated events is not especially productive when trying to address attribution questions. Climate is an average, or perhaps better, an accumulation of weather events. As such it is important to consider how a large number of events act in concert, in correlation, in cohesion.
One other point that I want to make: The practice of isolating a single event and attributing that event to climate change, is one of the most effective ways of opening up scientific investigation to effective scientific criticism. (see Pielke, Sr. et al. 2007) A single-event attribution claim is an open and appropriate invitation to those with knowledge of or interest in local information to investigate the attribution claim. Almost inevitably this leads to identification of more sources of uncertainty, which like the Aaron versus Ruth argument, are irreducible. This necessarily contributes to controversy, and controversy is the fuel of talk radio, blogs propagating around the world, and the maintenance of doubt.
This entire process of event attribution is one place where scientific investigation of the climate interfaces with the media. Therefore, it is also a place where, by definition, scientific investigation interfaces with the political argument. My analysis above suggests that, as framed by the public discourse, the pursuit of the path of event attribution and the explicit or implicit linkage of that attribution to climate change is scientifically questionable. This stands in contrast to the scientific pursuit of extreme events in historical context and the evaluation of whether their frequency of occurrence is changing. Politically or in terms of informing the public, the primary product of the pursuit of event attribution is to build and maintain doubt. The exception to doubt maintenance would be if a definitive, metaphorical smoking gun was discovered. But what is the probability of such a smoking gun being discovered in this process? A different perspective is needed on the role of extreme events in climate and the attribution of such events to global warming. As climate scientists, we have to think about what these studies mean to the body of our field’s communication of climate change.
And here is
Faceted Search of Blogs at climateknowledge.org