According to the passage, clouds must be represented by parameterizations because they

are observed less frequently than other components of the climate system.

respond to warming in ways that remain entirely unknown.

vary too much between regions to be described by a single formula.

form and dissipate at scales smaller than a global model can resolve.

The passage indicates that adding data to a structurally uncertain model can widen the range of estimates because the additional data may

reveal that models agreeing about the observed past diverge about the future.

contain measurement errors that propagate through the calculation.

require the model to be run at a finer resolution than before.

come from regions where cloud behaviour is unusually variable.

The author regards agreement among independent lines of evidence as carrying more weight than agreement among models because the lines

are based on longer observational records.

have been assessed by a larger number of researchers.

fail in different ways, so their convergence is not a shared assumption.

rely less on physical theory and more on direct measurement.

It can be inferred that the author would regard the principle that "a science which gathers more data ought to converge" as

correct in general but inapplicable to climate because its data are unusually noisy.

sound where uncertainty is statistical, but not where it is structural.

a misunderstanding of how scientific consensus is formed.

true of models but false of observations.

Which of the following, if true, would most undermine the author's explanation of why the range remained wide?

The 1979 report's range was arrived at by a small group over a single weekend.

Computing power available to modellers grew faster between 1979 and 2019 than the passage suggests.

Some assessment reports quoted a slightly different range from the one the passage cites.

Cloud parameterizations differing widely in their physical assumptions nonetheless yield sensitivities within a few tenths of a degree of one another.

←

Why Climate Sensitivity Has Not Narrowed

PRACTICE

100:00+4 −1MCQ Single Answer

Reading Passage

Equilibrium climate sensitivity — the eventual warming that follows a doubling of atmospheric carbon dioxide — was estimated in a 1979 report at somewhere between 1.5 and 4.5 degrees. Four decades later, after an enormous increase in observational coverage and in computing power, the range quoted in assessment reports had barely moved. Commentators have periodically offered this as evidence that the field is not making progress, on the reasonable-sounding principle that a science which gathers more data ought to converge. The inference is mistaken, and the reason it is mistaken is worth following. The range is not a margin of measurement error, of the sort that shrinks as observations accumulate. Sensitivity is governed largely by the response of clouds to warming, and clouds form and dissipate at scales far smaller than any global model can resolve; they must therefore be represented by parameterizations, compact formulas standing in for processes the model cannot simulate directly. Several parameterizations are defensible on the physics, and they imply different sensitivities. Uncertainty of this kind is structural rather than statistical, and feeding more data into a structurally uncertain model does not necessarily narrow it. It can widen it, by revealing that models which agree about the observed past diverge about the unobserved future. The narrowing that did eventually arrive, to roughly 2.6 to 3.9 degrees, came from a change of method rather than from better models. Rather than polling model outputs, assessments combined independent lines of evidence — the instrumental record of the past century, reconstructions of past warm and cold climates, and process-level understanding of individual feedbacks — and required a value to be consistent with all of them at once. The lines fail in different ways, so agreement among them carries weight that agreement among models does not. The lesson generalizes past this case. Where uncertainty is structural, progress looks less like increasing precision than like triangulation, and a range that refuses to move is not by itself evidence of a field standing still.

The primary purpose of the passage is to