Carl Wunsch of Harvard and MIT wrote this wise note unmasking the Uncertainty Monster to NYTimes writer Andy Revkin "on the range of papers finding oceanic drivers of short-term climate variations. He began by noting the challenge just in determining average conditions":
Part of the problem is that anyone can take a few measurements, average them, and declare it to be the global or regional value. It’s completely legitimate, but only if you calculate the expected uncertainty and do it in a sensible manner.The system is noisy. Even if there were no anthropogenic forcing, one expects to see fluctuations including upward and downward trends, plateaus, spikes, etc. It’s the nature of turbulent, nonlinear systems. I’m attaching a record of the height of the Nile — 700-1300 CE. Visually it’s just what one expects. But imagine some priest in the interval from 900-1000, telling the king that the the Nile was obviously going to vanish…
Wittgenstein framed the answer to this scary scientific conundrum as the last line of his Tractatus, something calculated to terrify politicos working both sides of K-Street and denizens of Madison Avenue :Or pick your own interval. Or look at the central England temperature record or any other long geophysical one. If the science is done right, the calculated uncertainty takes account of this background variation. But none of these papers, Tung, or Trenberth, does that. Overlain on top of this natural behavior is the small, and often shaky, observing systems, both atmosphere and ocean where the shifting places and times and technologies must also produce a change even if none actually occurred. The “hiatus” is likely real, but so what? The fuss is mainly about normal behavior of the climate system.The central problem of climate science is to ask what you do and say when your data are, by almost any standard, inadequate? If I spend three years analyzing my data, and the only defensible inference is that “the data are inadequate to answer the question,” how do you publish? How do you get your grant renewed? A common answer is to distort the calculation of the uncertainty, or ignore it all together, and proclaim an exciting story that the New York Times will pick up.A lot of this is somewhat like what goes on in the medical business: Small, poorly controlled studies are used to proclaim the efficacy of some new drug or treatment. How many such stories have been withdrawn years later when enough adequate data became available?
Whereof one does not know, thereof one must be silent.