Sunday, July 19, 2009

Yeah, climate contrarianism is getting a bit old

I won't disagree with Nate on that note. Nate's a statistical wonk, which is a vice I share to an extent. What makes global warming so damn difficult to discuss is the tremendous amount of variability present in the data, whether we think in terms of long-term global trends or even if we think about local day-to-day temperature fluctuations. A couple of coolish summer days in Minneapolis really wouldn't be terribly unusual, and as Nate shows, just looking at that locality the last month has been a strikingly average month for that particular metropolitan area. Of course it is crucial to keep in mind that whatever happens at a particular locality tells us little or nothing about what is going on globally. Just to give you an idea, here's a graph provided by NOAA of temperature anomalies for June of this year (h/t to clif):

There are a few islands of blue amidst a sea of red - a mildly poetic way of saying that globally June was warmer than usual. Keep in mind that the graph merely portrays one month. Here's another that I found late last month that gives us an idea of the long-term global trend:

As I noted at the time:
Image found at Paul Krugman's blog. Krugman goes on to say:
...temperature is a noisy time series, so if you pick and choose your dates over a short time span you can usually make whatever case you want. That’s why you need to look at longer trends and do some statistical analysis.


What this tells me is that annual temperature is indeed noisy: there have been many large fluctuations, indeed much larger than the up-and-down in the last decade or so. But the direction of change is unmistakable if you take the longer view. The fitted line in the figure is a 3rd-degree polynomial, but any sort of smoothing would tell you that there is a massive upward trend.

Of course, trend-spotting is no substitute for causal modeling; and the models are getting truly scary in their implications.
John Cole also looks a bit at the "climate contrarians" who've been naysaying for too damned long.

In particular, I thought the Krugman post was useful in reminding readers that in order to understand what is happening with the planet's climate, one is required to search for long-term trends using what is rather noisy data (that is, data in which there is a great deal of variability). In the case of global warming, it appears that the signal, even with all the noise, is coming in crystal clear.
Merely convenience sampling a few days-worth of temperatures in one location (or perhaps a year or two globally) will lead one to make whatever sort of inference one might want to draw, but the inference is unlikely to be worth much. What I think happens with the climate contrarians is that they capitalize (in some cases) or fall victim to (in other cases) fairly normal cognitive biases. One of those biases falls under the name "availability heuristic" - we tend to overestimate the probability of occurrences that are most easily recalled. The most recent day's weather will be more available to conscious awareness than what might have happened a few days, weeks, or years earlier; and of course extremes will get remembered more easily than events that are more or less normal. For someone in the upper Midwest, summer temperatures that apparently are too cool for summer skinny-dipping in Lake Michigan would be very salient and will be given more weight than deserved. Then, there is the tendency for we humans in everyday life to search only for data that confirm our particular pet hypotheses rather than search for data that might debunk our particular pet hypotheses. We do a very incomplete scan of the available data, and having found a few data points we stop searching. Gordon Allport once remarked that from a thimble full of data, we make inferences that are as large as a tub (I'm paraphrasing his quote). Clearly this is a bias that can lead to errors. On top of that all is simply that our own particular sets of beliefs are so active (even if we're not aware of it; a phenomenon that in social cognition gets referred to as chronic accessibility) that they shape our perceptions of the present and distort our memories of the past. Hence, for a climate contrarian, the end result will be an exceptional awareness of any unusually cool weather in their locality, as well as an exaggerated tendency to recall unusually cool weather in the recent and not-so-recent past.

The problem with these particular biases in thinking is that they hinder accurate perception and judgment. When dealing with something as complex as the global climate, it is crucial to use tools that will allow us to get past our own particular cognitive limitations - the aforementioned biases and our limitations in synthesizing vast amounts of data. This is where statistical modeling along with accurate descriptive statistical data come in. These tools enable us to get past our own limitations and see more clearly what would have otherwise gone undetected.

Although I am reasonably convinced that the overall trend is toward a warmer planet, there are going to be variations on how locations are affected, and of course there will be cold days and hot days, as well as relatively cold years and relatively hot years. Keep in mind that all the variation is noise - never lose sight of the signal.

No comments:

Post a Comment