Below One Hundred and Eighty

The story is told in every lecture on data quality, and it goes like this. NASA had been watching Antarctic ozone from orbit since 1979. Its processing software was set to reject readings that looked impossible. The ozone hole produced readings that looked impossible, so the software rejected them, and the hole went undetected for years until three men with a ground instrument found it the slow way.

It is a good parable and almost none of it is right.

What the threshold was

The Total Ozone Mapping Spectrometer aboard Nimbus 7 inferred the amount of ozone above a point from the light coming back at it. Working out a quantity from a radiance means assuming the atmosphere is reasonably close to a set of profiles you have calibrated against. If the real atmosphere is far outside that set, the arithmetic stops being trustworthy, and the answer it produces is not a measurement so much as an extrapolation.

A Dobson unit is a way of expressing how much ozone sits in a column of atmosphere by imagining it all brought down to sea level pressure and measured with a ruler. One Dobson unit is a layer a hundredth of a millimetre thick. A healthy column over Antarctica in spring ran to three hundred of them, which is to say three millimetres of ozone, which is what stands between the surface and the harder end of the ultraviolet.

The processing team set a quality control flag on any retrieval below 180 Dobson units. That number was not arbitrary. Before 1983 there had never been an observation below 200, anywhere, so anything under 180 was outside the range the instrument had been calibrated for. The flag meant this reading is outside what we can interpret, not this reading is wrong.

And it was a flag, not a delete. The values were marked and kept, which is what allowed them to be reprocessed and confirmed two years later.

What the team actually did

The October 1983 data went into processing in August 1984, at the Goddard Space Flight Center, by the Ozone Processing Team led by A. Fleig and including Donald Heath and P. K. Bhartia.

They noticed at once that the number of flagged points had jumped enormously. Anyone who has run an instrument knows the first hypothesis for a sudden crop of impossible readings, and it is not that the world has changed. It is that the sensor has broken. The standard check is to compare against ground truth.

Here the story turns on an accident of access. The only publicly available real time Antarctic ozone measurements at the time came from the South Pole station, and they showed about 300 Dobson units for October 1983, which is normal. That was strong evidence that the satellite was faulty. The British Antarctic Survey's Halley records, which would have said the opposite, were not publicly accessible.

The puzzle did not go away, because the retrievals from everywhere outside the polar vortex looked entirely normal, which is not how a broken sensor usually behaves. And there was a real fault in the mix: the preliminary observations from October to December 1983 turned out to be invalid, because the wrong channels had been read on the instrument. A team trying to decide whether an anomaly is real had, sitting in the same data, an anomaly that was not.

By December 1984 they had worked through it and were confident enough that the readings were genuine to submit an abstract to a conference in Prague the following August.

The British

Halley had been measuring total ozone with a Dobson spectrophotometer since 1957, from the ground, looking up. Twenty six years of the same measurement in the same place.

Joe Farman, Brian Gardiner and Jonathan Shanklin, of the British Antarctic Survey, had values from Halley and from Faraday showing that after about two decades of steady readings, ozone in the austral spring had begun falling in the late 1970s. By 1984 the October layer over Halley was around two thirds as thick as it had been in earlier decades. In October 1983 it had gone below 200 Dobson units for the first time.

They published in Nature on 16 May 1985, and they did more than report the fall. They proposed the mechanism, linking it to chlorofluorocarbons in aerosols and refrigeration.

The Prague abstract was three months away.

After

It took until November 1985, after a workshop, for the first press report to carry the NASA results, in the New York Times. That article contains the first public use of the word hole to describe the feature, a term Sherwood Rowland had coined a few months earlier.

It also records something that gets forgotten now that the answer is settled: the cause was still open. Farman and his colleagues had argued for chlorine, but two other explanations were live. One was dynamical, an anomalous upwelling of relatively ozone-poor air from below. One was a solar cycle effect.

NASA's own paper on the Nimbus trends came out in August 1986, with the data reprocessed using additional low ozone profiles in the calibration. It showed the long term decline clearly across the polar vortex and corroborated the sub-200 October 1983 values from Halley.

What the parable gets wrong, and what it gets right

The satellite did not throw anything away. The team noticed, checked against the only ground data they could get, were misled by it, hit a genuine instrument fault at the same moment, worked it out anyway, and were beaten into print by three months.

But there is a real lesson underneath, and it is not about carelessness. It is about where thresholds come from. The limit was set at 180 Dobson units because nothing had ever been recorded below 200. Every limit of that kind is built out of what has happened so far. It works perfectly until the thing you are measuring does something it has not done before, which is the only occasion on which you actually needed it to work.

And the reason the discovery went to a hut on the Brunt Ice Shelf rather than to a satellite is that somebody had been pointing the same instrument at the same patch of sky since 1957. The satellite could see everywhere at once, and it had been doing so for four years. Halley could see back twenty six. When the question is whether a number has changed, the second kind of instrument is the one that answers it.

Sources

• Gavin Schmidt, RealClimate, "What did NASA know? and when did they know it?" (the quality control flag set on retrievals below 180 Dobson units and the absence of any observation below 200 before 1983; the dependence of the retrieval on calibrated atmospheric profiles; the processing of the October 1983 data from August 1984 by the Goddard Ozone Processing Team under A. Fleig with Donald Heath and P. K. Bhartia; the noticing and investigation of the increase in flagged data; the South Pole readings of about 300 Dobson units and the inaccessibility of the British Antarctic Survey data; the normal retrievals outside the polar vortex; the invalid preliminary observations from October to December 1983 caused by the wrong channels being read; the December 1984 abstract submitted for the Prague conference of August 1985; the November 1985 New York Times report as the first public use of the word hole, coined by Sherwood Rowland; the dynamical and solar cycle theories still credible at the time; and the August 1986 NASA publication with additional low ozone profiles in the calibration confirming the Halley values).
• Rob J. Hyndman, "How NASA didn't discover the hole in the ozone layer" (the myth that anomalous data were thrown away; the TOMS record from 1979 and the Halley ground record from 1957; and the 180 Dobson unit threshold being designed to identify instrument problems rather than genuine systematic change).
• J. C. Farman, B. G. Gardiner and J. D. Shanklin, "Large losses of total ozone in Antarctica reveal seasonal ClOx/NOx interaction", Nature volume 315, 16 May 1985 (the paper itself).
• Nature, "The discovery of the Antarctic ozone hole" (the unanticipated decreases over Halley and Faraday; steady values for about two decades before the fall began in the austral spring of the late 1970s; the October layer over Halley being about two thirds of its earlier thickness by 1984; and the suggested link to chlorofluorocarbons used in aerosols and refrigeration).