Attribution Science Relies On Statistical Sleight-Of-Hand To Link Weather Events To Climate Change

Doomsayers wasted no time identifying the destruction wrought by Hurricanes Helene and Milton as “proof” of climate change. [emphasis, links added]

It is the latest example of an increasingly common phenomenon: attributing individual weather events—a flood in Vermont, a heatwave in the Pacific Northwest, a cold snap in Texas, and many others—to climate change.

Linking individual weather events that have occurred since time immemorial to long-term changes in climate has spawned a new academic field called attribution science.

However, because it is impossible to prove a specific weather event was the result of human-caused climate change the way a criminal is proved to have committed a crime, attribution science uses old-fashioned statistical legerdemain to link human-caused climate change to individual events.

The first example of attribution science dates back 20 years to an academic paper that purported to show the European heatwave of 2003 was linked to climate change. Doing so has now become routine.

A European group called World Weather Attribution now provides almost-instant analyses of weather events worldwide, virtually all of which are caused or made worse by burning fossil fuels.

Attribution science uses climate models to determine the statistical likelihood of an individual event without human-caused climate change, that is, without greenhouse gas emissions from fossil fuels and agriculture.

That counterfactual likelihood is then compared to the observed probability of similar events. The approach is identical to the “but-for” analyses lawyers often use to prove damages in commercial litigation cases.

For example, a climate model might determine the likelihood of a catastrophic flood, such as the one that devastated Montpelier, Vermont, in July 2023 to be once every 500 years.

That value would be compared to the observed likelihood, say once every 100 years. (The last major flood in Montpelier took place in 1927.)

The ratio of the two values – in this case, 0.20 – is used to calculate the “fraction of attributable risk,” which is just one minus the ratio, or 80%.

In other words, 80% of the increased risk of a severe flood in Montpelier could be attributed to human-caused climate change.

For events like Hurricanes, which occur every year, a slightly different approach is used.

Because there is no evidence that the number of hurricanes has increased over the last century or more, rather than estimating the change in the likelihood of the number of hurricanes, the change in the likelihood of the severity of hurricanes is estimated.

However, severity is an imprecise concept. Although hurricane severity is measured based on wind speeds, a hurricane’s physical and economic damages depend on many factors, including where it makes landfall, how fast it moves over land, whether the storm surge takes place at high tide or low tide, and so forth.

Consequently, a less powerful hurricane, such as Sandy in 2012, may have greater monetary damages than a far more powerful hurricane.

Statistical attribution studies depend entirely on the accuracy of climate models—and their assumptions—to calculate counterfactual likelihoods.

For rare events, those calculations are compounded because rare events are, well, rare. Hence, there may be little historical data on which to estimate accurately how frequently they will occur, much less determine the frequency using a counterfactual climate model.

And even for more frequent events, such as the hurricanes and tornadoes that the U.S. experiences every year, separating the multiple factors that can affect their severity is subject to all sorts of statistical sleight-of-hand.

Sixty years ago, a journalist named Darrell Huff wrote a short, humorous book titled “How to Lie with Statistics,” which presented a litany of statistical techniques that could be – and still are – used to distort reality.

He may not have foreseen attribution science, but he would surely smile at its use today.

Read more at RealClearEnergy