Memorandum submitted by Nicholas Barnes
and David Jones (CRU 57)
We are writing with regard to the evidence your
committee has heard on the Climatic Research Unit (CRU) at the
University of East Anglia. We are software experts with long industrial
experience. We also have some expertise in climate science software
such as that discussed in evidence, particularly the oral evidence
from Professor Jones of CRU and Professor Slingo of the Met Office,
and a number of evidence memoranda. We hope that our submission
may clarify some of the issues raised by that evidence.
Nicholas Barnes has been a professional software
engineer since 1990; David Jones has been a professional software
engineer since 1994. In 1997, Nicholas founded a software consultancy
firm called Ravenbrook Limited, which he continues to run today.
David has worked for Ravenbrook since 2003. We work in various
industry sectors, including desktop data analysis, games development,
memory management, software tool integration, and robotic control.
We both have degrees in mathematics, and post-graduate diplomas
in computer science, from the University of Cambridge. We both
have life-long interests in science and in computers.
In 2008 we started Clear Climate Code, a volunteer
pro-bono project, because we saw that the clarity of software
published by climate scientists was disrupting the public debate
about climate change. You have asked questions and heard evidence
about this disruption.
The goals of the Clear Climate Code project
are:
1. To produce clear climate science software;
2. To encourage the production of clear climate
science software; and
3. To increase public confidence in climate science
results.
We aim to increase public confidence in results
by publishing and clarifying the software which produces those
results. That process will certainly uncover and correct errors.
These corrections are essential to the goal. We don't want the
public to trust incorrect results; any incorrect results should
be visibly corrected, so that the public may trust them.
Several other people around the world have joined
and helped us in our efforts. We publish all of our work at http://clearclimatecode.org/
and make presentations about it at conferences.
The main focus of Clear Climate Code so far
has been on GISTEMP.
GISTEMP is a gridded dataset of global historical
surface temperature anomalies, substantially similar to the HadCRUT3
dataset produced jointly by the Met Office Hadley Centre and by
CRU. GISTEMP produces a "Global Land-Ocean Temperature Index",
a chart that shows how global temperatures have changed since
1880. see http://data.giss.nasa.gov/gistemp/graphs/. This chart
is the analogue of HadCRUT3's "Global average temperature"
chart, see http://hadobs.metoffice.com/hadcrut3/diagnostics/global/nh+sh/.
The source code that produces the GISTEMP analysis
is published, see http://data.giss.nasa.gov/gistemp/sources/.
All the data used in the GISTEMP analysis is also published. Clear
Climate Code have been reimplementing this analysis, with an emphasis
on clarity. The intention is that anyone interested in the subject,
and capable of understanding a program, should be able to download
our software and easily follow it. Collectively we have closely
examined every part of the original GISTEMP source code and produced
a new version that produces exactly the same results as GISTEMP.
We continue to refine our version, to improve its clarity. Dr
Reto Ruedy, a mathematician at NASA GISS with responsibility for
GISTEMP, has said recently that they would like in future to use
our version.
As part of this process, we have read several
of the scientific papers that describe the GISTEMP analysis, including:
J Hansen, R Ruedy, J Glascoe, Mki Sato, J
Geophys Res 104, 30997-31022, doi:10.1029/1999JD900835 (1999).
J E Hansen, S Lebedeff, J Geophys Res 92,
13345-13372 (1987).
It is obvious to us that the original GISTEMP
software is not engineered to a high quality. But this is not
at all unusual for scientific software, or for software in the
commercial sector.
Scientists generally receive little or no training
in software development. In making our new implementation we have
found a few minor problems or "bugs" in the GISTEMP
software. We have reported these to Dr Ruedy at NASA GISS who
has in all cases fixed them and thanked us for reporting them.
The problems we have found either do not affect the published
results, or only affect them by tiny amounts, far less than the
uncertainty. For instance, one month's reading might change by
a hundredth of a degree.
As our work continues, we have also investigated
some questions raised by critics of GISTEMP, for instance whether
the warming signal of the results could be due to rounding numbers
in the GISTEMP system, or due to the urban heat-island adjustment,
or due to changes in weather station numbers in the 1990s. We
have been able to show very easily that these factors are not
significantthe effect of each factor is tiny, and the effects
are often in the opposite direction to that suggested by critics.
Regarding GISTEMP we conclude:
the source code could be clearer;
the source code is free of any major
errors;
it performs as described in the scientific
literature;
it produces a result which is substantially
the same as HadCRUT3, and other similar datasets;
the result is also substantially the
same when the factors commonly mentioned by critics are removed;
and
we have no reason to doubt the output
of the GISTEMP analysis.
By "substantially the same" we mean
that the conclusions regarding warming in the latter part of the
20th century that one can draw from the GISTEMP analysis are the
same as the conclusions one can draw from the HadCRUT3 analysis.
Of course, none of our work supporting these
conclusions would have been possible without the full publication
of the original GISTEMP software and the data which it uses.
If you have any questions regarding this submission,
or our work more broadly, we are of course available to give further
evidence, in person if necessary.
Nicholas Barnes
Director
David Jones
Consultant
Ravenbrook Limited
March 2010
|