Recently, I made a request to Thomas Bracegirdle, junior partner of Connolley and Bracegirdle, for the model data used in two recent articles: Bracegirdle and Connolley (GRL 2007) about 20th century Antarctic models; and Bracegirdle, Connolley and Turner (JGR 2008) about 21st century Antarctic models. William Connolley is well known in the blogosphere, especially for his zeal in extinguishing heresy at Wikipedia.
The request for collated model data was very similar in form to my request last fall to Benjamin Santer, senior partner of Santer and Associates LLP, for model data used in Santer et al 2008.
Readers may remember my recall Santer’s rude refusal to provide the data, culminating in the cordial Team salutation “Please do not communicate with me any further” – I guess this means “Hasta la vista, baby” in Team dialect.
After all of this, Santer’s boss, David Bader, sent me an email purporting to “clarify several mis-impressions” – saying that, aw shucks, they had planned to put the data online all along and that my various FOI requests had nothing to do with it. See here.
Connolley and Bracegirdle (GRL 2007) had said that they had assessed “19 coupled models from the IPCC fourth assessment report archive from the simulation of the 20th century,” from the “Coupled Model Intercomparison Project phase 3 (CMIP3) multi-model dataset at https://esg.llnl.gov:8443/ “. They report the collecting several variables, one of which was sea ice, a variable of considerable recent interest. For sea ice, they said that used data from “15 coupled models”, considering “sea ice fraction (the proportion of grid covered by sea ice and not leads) rather than ice extent (the proportion of grid covered by ice of fraction at least 15%)”.
They noted that
… except for CSIRO the models have essentially zero skill. This is because, apart from CSIRO, all the models have months in which their total area falls well outside the observed range compared to satellite observations from Comiso  using the bootstrap method, which verify best against other observations in Antarctica [Connolley, 2005]. All models produce a seasonal cycle with a peak in approximately the right season, though HadCM3 is a month late and NCAR CCSM two months early. IAP FGOALS has vastly overextensive ice, extending to South America.
Bracegirdle et al (JGR 2008) examined 13 models for 21st (rather than 20th century) reporting that:
Projections of total sea-ice area show a decrease of 2.6 ± 0.73 × 106 km2 (33%).
On Mar 30, 2009. I wrote Bracegirdle on the online submission form at the British Antarctic Survey, requesting the collated monthly data used in the articles. The underlying data is online at PCMDI, but to extract the monthly data would require downloading terabytes of data and then figuring out how the monthly composites were made – time-consuming clerical operations with related risk of error, that are irrelevant to statistical analysis.
For the purpose of statistical analysis, I was prepared to use the Bracegirdle-Connolley collation – only verifying the collation if issues arose. Here is my initial request:
Dear Dr Bracegirdle,
I read your interesting articles on AR4 models and would appreciate a digital version of the collation of Antarctic sea ice model projections as used in your most recent articles.
Regards, Steve McIntyre
Bracegirdle promptly replied but not responsively, sending me a PDF of his article, rather than the requested data. I replied:
I already had a copy of the article. My interest was in the DATA: the collation of Antarctic sea ice model projections.
Thanks, Steve McIntyre
Bracegirdle cheerfully replied:
I’d be happy to supply the data. Would NetCDF format be ok?
I thought to myself that the Team seemed to have learned something from the Santer episode. (They had, but not in the way that I had optimistic ally thought when I got the above email.) A week or so later, Bracegirdle emailed:
I have attached NetCDF files of the data that were used for Fig. 8 and described in Bracegirdle et al. (2008). I had to convert to NetCDF from PP format (which is what we and the UK Met Office use). Therefore some metadata, such as variable name (look for ‘unspecified’), does not appear in the NetCDF files. Hopefully these are the data that you were referring to.
These turned out to be tiny files. They did not contain the collated monthly sea ice data used for the calculations, but small files of about 80 numbers showing 21st century sea ice concentration change (difference between 2080-2099 mean and 2004-2023 mean) for (a) DJF, (b) MAM, (c) JJA and (d) SON.
Once again, it wasn’t responsive to the request. So one more request:
This is not at all what I was looking for. You sent me data sets that are only 30K. You state in the article “Model data were retrieved from the data portal at https://esg.llnl.gov:8443/ , from which 19 of the 24 available models were found to have the data required for this assessment.”
This is the data that I was looking for.
Regards, Steve McIntyre
Bracegirdle’s responded that he couldn’t supply the data. After initially asking me “would NetCDF format be ok?” and my answering yes, he now said that they had deleted the NetCDF data and that it “would take some time (more time than I have spare!) to retrieve the data again or convert them back to NetCDF”, with his answer ultimately being the same as Santer’s.
Unfortunately we have deleted all the NetCDF files that we downloaded after converting them to PP format. It would take some time (more time than I have spare!) to retrieve the data again or convert them back to NetCDF. However, all the data are freely available at https://esg.llnl.gov:8443/home/publicHomePage.do – you just need to register. “
I’ve responded as follows:
I find it hard to believe that the British Antarctic Survey would permit the deletion of relevant files for two recent publications or that there aren’t any backups for the deleted data on institutional servers. Would you mind inquiring for me? In the mean time, would you please send me the PP format files that you refer to here for the monthly sea ice data for the 20th century models discussed in your GRL article and the 21st century models referred to in your JGR article.
Regards, Steve McIntyre
We’ll see where this goes.