At 12:04 PM 6/13/1998 -0700, Jeff Dozier wrote:
>Richard,
>
>Some colleagues in Computer Science at University of Washington are working
>on systems for data lineage. They are interested in the story of the early
>TOMS data, where we missed the Ozone Hole because the values from TOMS
>(although correct) were so far outside the range of climatology.
>
>I'd like to point them to some helpful references so they can use this as a
>specific example, and to help them build systems that help us better unravel
>large data sets. If you could point to some papers that describe the
>situation well, I would be appreciative. You can just 'reply all' to this
>e-mail.
>
>Thanks,
>
>Jeff Dozier (EOS Project Scientist, 1990-92)
>
>
Dear Jeff,
We have never actually published anything telling our side
of the story on the discovery of the ozone hole - I'm not sure
where you would publish such a thing. Certainly not GRL.
But also we were afraid that such an article would sound defensive.
What we have done is respond to a number of questions over the
years. I was questioned by a German author some years back, to
which I replied giving the sequence of events on our side. I
really liked the flavor of the article he wrote and thought that
he accurately summed things up. I have now sent it to a number
of people who have asked why we didn't "discover" the ozone hole.
It is reproduced below.
Rich McPeters
=================================================================
Robustness of Statistical Gossip and the Antarctic Ozone Hole
By Friedrich Pukelsheim, Corresponding Editor of the IMS
Bulletin
I do not remember where I heard the story for the first time,
but it read somehow like this. NASA, the scientific outfit with
the biggest collection on atmospheric data, would have been
first in discovering the Antarctic ozone hole had they not used
modern statistical methods. However, the computer code to
analyze the data was based on robust statistical procedures
which suppressed the unusual low ozone readings. When British
and Japanese scientists published their findings on the
existence of an ozone hole over the Antarctic, NASA went back to
reexamine their data and found that they could have recognized
the ozone hole much earlier had they given proper attention to
the unusual low ozone readings. My inquiries about these
stories, at Oberwolfach and on other occasions, were met with a
curious or malicious smile, depending on my partner, some of
whom had heard the story from other sources. But nobody was
able to provide any facts.
The story resurfaced when I was shown the article In eisiger
Dunkelheit bilden sich Wolken uber der Antarktis by Dr. Caroline
Mohring from the science staff of the national German newspaper
Allgemeine Zeitung of 14 November 1987. The first two paragraphs
are as follows [my translation].
In Icy Darkness Clouds Doom Over the Antarctic
The ozone hole signals changing conditions of life on earth
By Caroline Mohring
FRANKFURT, 13. November. They would not trust their
instruments. Such low readings contradicted all expectations.
The British scientists from the Halley Bay station skeptically
checked their instruments, repeated the measurements in the
following year and safeguarded their results in various ways:
Finally, in 1985, J.C. Farman and collaborators went public with
their finding that in the beginning Antarctic spring the ozone
in the upper atmosphere over the south pole dwindles away. The
stratospheric ozone layer which, in about 20 kilometer height,
saves the earth from harmful ultraviolet radiation temporarily
develops a hole, they reported.
Japanese scientists had made similar observations, but
published their paper In 1984 in a little red journal. Only the
NASA satellites which also monitor the atmospheric ozone content
seem to have missed the hole. However, a revision of the data
proved that the satellites had indeed reported decreasing ozone
readings - while the computer code had deliberately cast out
values that deemed not possible as measurement errors.
Reconsideration of the same data set proved: since the end of
the seventies the September ozone content over the Antarctic
clearly decreases - and this effect increases from here to here.
When I approached Dr. Mohring for more details it turned out
that they had moved offices and on that occasion had cleared
away what they thought obsolete. However she directed me to
Professor Paul Crutzen from the Abteilung Chemie der Atmosphare
of the Max-Planck Institut fur Chemie in Mainz. Dr. Crutzen
responded that he knew about the "story" without being able to
provide further detail. He proposed to further enquire with
NASA.
In April 1990 I received the following letter from Dr.
Richard McPeters, Head of the Ozone Processing Team at the
Goddard Space Flight Center, providing a full account of the
story. The upshot is that the unusual low ozone readings had
been looked at separately - as they should have been - and had
been double-checked against measurements from another station -
as they should have been. The data set from the second station
was much more in line with the experience available at that
time, and hence questioned the unusual low readings.
Unfortunately it was this second data set used for double-
checking that was in error. Here is his accounts
Dear Dr. Pukelsheim:
This is in reply to your letter asking about NASA's role in
the discovery of the Antarctic ozone hole. Unfortunately,
everyone "knows" that NASA did not discover the ozone hole
because the low values were "thrown out" by the computer code.
This myth was the result of a statement made by my colleague,
Dr. Richard Stolarski, in reply to a question during an
interview on the science program NOVA in which he was asked why
NASA did not discover the ozone hole first. Dr. Stolarski was
not directly involved in ozone processing at that time and his
answer was not correct.
NASA scientists were studying the unusual low ozone values in
July of 1984, almost a year before the publication of the Farman
paper. Our software is designed so that data are never just
thrown out. Rather, questionable data are "flagged" as not
being of best quality. TOMS makes 5 x 10**7 individual
measurements of ozone each year, and of these measurements some
small fraction will be bad because of encoding errors,
transmission errors, or possibly instrument effects (the effect
of random noise is greater at large solar zenith angles).
Screens are put in the software to detect out-of-bounds
conditions and flag them. One such screen was on ozone such
that ozone amounts less than 180 DU were flagged as possibly
being in error. This was a reasonable check since no reliable
measurement of ozone this low had ever been reported before
1983. In July 1984 we were processing the data from October
1983, the first year in which the ozone hole was sufficiently
well developed to drop below our 180 DU threshold. This was
noticed in our quality control screening as a sudden increase in
flags for ozone too low. Since this could have been the result
of an instrument problem, we compared our measurements with the
only Dobson ground station data then available, that from the
Amundsen-Scott station at the South Pole. (Data from the Halley
Bay station are not sent to the Canadian AES for archival.)
Unfortunately, because of an error, the South Pole Dobson
station was reporting ozone values of 300 DU when our satellite
instrument was reporting less than 180 DU. As noted in a paper
by Komhyr et al. (1986), "...data previously reported for
October-December 1983 have been identified as erroneous and
uncorrectable (observations were incorrectly made on A', C', and
D' rather than on A, C, and D wavelengths)."
Because of this error, we were necessarily very cautious in
accepting our own data as valid. But after careful evaluation
we could find no problems with the data and decided to report
them. In late 1984 (months before the publication of the Farman
et al. paper in the May 1985 issue of Nature) we submitted an
abstract for the IAGA/IAMAP meeting in Prague, Czechoslovakia,
August 1985, reporting these low ozone observations. In
conclusion, our failure to be first to report the Antarctic
Ozone hole was largely the result of an unfortunate coincidence
of erroneous Dobson values reported at the South Pole. (And note
that Dr. Farman was likewise very cautious in checking his data
before publication.) The myth that our computer code "threw out
the data" is unfortunately very hard to correct without
appearing defensive.
Sincerely,
Dr. Richard McPeters
Head, Ozone Processing Team
NASA/Goddard Space Flight Center
Code 916
Greenbelt, MD 20771 USA
REFERENCES
Bhartia, P.K., Heath, D.F., Fleig, A.J., Observations of
anomalously small ozone densities in south polar stratosphere
during October 1983 and 1984, International Association of
Geomagnetism and Aeronomy, 5th General Assembly, IAGA/IAMAP,
Volume 2, Abstracts (without Division 1) and Index of
Authors, Prague, Czechoslovakia, 5-17 August, 1985, page 416.
Komhyr, W.D., Grass, R.D., Leonard, R.K., Total ozone decrease
at South Pole, Antarctica, Geophys. Res. Lett., 13,
1248-1251, 1986.
Farman, J.C., Gardiner, B.G., Shanklin, J.D., Large losses of
total ozone in Antarctica reveal seasonal CIOx/NOx
interaction., Nature, 315, 207-210, 1985.
Mohring, C., In eisiger Dunkelheit bilden sich Wolken ubcr der
Antarktis, Frankfurte Allgemeine Zeitung, 265, 14. November
1987, page 3.
========================================================================