Statistics and global warming

The commentary “Lies, damn lies, and the grim reality of climate change” by Amy Goodman and Denis Moynihan last month used the last five words of their title to replace the word “statistics” — which, in his famous quote, Mark Twain meant to be even worse than damn lies. Though not the authors' intent, that paradoxically is precisely what “the grim reality of climate change” is: even worse than a damn lie. Why? Statistics tell a story quite different from theirs of unabated alarmism over human dependence on fossil fuels.

Though sober, the story told by statistics about global warming mutes the alarm bells. Climatologists use statistics in their study of global warming. So, why is their story so grim? To answer that question, substitute the word “misuse” for the word “use” in the preceding sentence.

In a statistics journal article last year, I showed that climatologists have been misusing statistics to claim that (a) global warming was occurring at an alarming rate and (b) its occurrence was due solely to human misbehavior. Neither of those claims is true.

What have climatologists been doing wrong? They use a statistical model that divides the measurement of global temperature at a specific time into an estimated and an error component. The estimated component is a sum of unique contributions to global temperature by different possible influences on it that can vary in extent over time, some of the influences natural and some human, like the amount of airborne carbon dioxide created by human activity at the time. In that sum, the influences are weighted statistically to minimize the variation of the error components of the measurements over time.

(Variation of the extent of an influence, of course, can occur over space as well as time, but inclusion of space in this brief explanation would make it unnecessarily cumbersome.)

So far, so good, but here is where the trouble comes. In addition to their excluding natural influences of which they are unaware, what most climatologists do wrong is to attempt to reduce the error variation or to achieve an expected result by altering possibly inaccurate measurements of some of the influences, like the amount of aerosol or the extent of cloudiness in the atmosphere. The result of that alteration, the article shows, is to create an impermissible predictive relationship between estimates and errors (errors, by definition, being unpredictable) that exaggerates estimates of the rise or fall of global temperature over time.

In the article, I also showed that correct use of the same statistical methods would demonstrate not only that the rate of global warming is no more than half of what most climatologists have claimed but also that natural rather than human activity is quite possibly the source of much, if not all, of it.

The article has had widespread distribution among not only statisticians and climatologists but also journalists, who tend to avoid the nittygritty of statistics —formulas, tables, or graphs — like the plague. You seldom see any of those forms of communication in newsprint. That is why I have avoided them here. Their correct use is certainly notable by its absence at the recent World Economic Forum meeting in Davos, Switzerland — just as the word “statistics” is notable by its absence in the title and the content of the Goodman-Moynihan commentary.