Now when some of these data and programs are leaked, would it be possible to make an open-source project to rewrite the code in Python, with unit tests etc.?
Could someone build their scientific carreer in climate science on fixing the software? I guess it would depend on whether the other people would cooperate: share their data and code, and use the fixed code.
It is our opinion that the GISTEMP code performs substantially as documented in Hansen, J.E., and S. Lebedeff, 1987: Global trends of measured surface air temperature. J. Geophys. Res., 92, 13345-13372., the GISTEMP documentation, and other papers describing updates to the procedure.
This ccc-gistemp project seems, however, to be stalled, in that it hasn’t released anything since 2010 (though there were a few blog posts in 2011). This doesn’t seem to be because they achieved everything they hoped to; the last information on their website indicates that there’s plenty more to do.
(This was made possible not because anyone leaked or stole anything, but because NASA released the GISTEMP code.)
Good point. I don’t know. The situation might have improved already, in terms of people releasing their code (though probably not in terms of the code being of great quality). At least, from what I’ve seen, releasing data and code is now somewhat more common. For instance, when critiquing a recent paper by Mann, Steve McIntyre wrote:
Michael Mann has had a paper on the Atlantic Multidecadal Oscillation (AMO) accepted by Geophysical Research Letters: “On forced temperature changes, internal variability, and the AMO”. The abstract and access to Supplementary Information is here . Mann has made a preprint of the paper available, here . More importantly, and very commendably, he has made full data and Matlab code available.
I don’t know if code quality has improved since the rather embarrassing 2009 leak.
The code for the Mann 1998 paper mentioned in the post was written in Fortran, and McIntyre found it on an old FTP server of Mann’s. He rewrote the code in R. At least people are writing code in languages like R and Matlab and maybe Python now.
Some researchers still use unconventional languages for their mathematical programming, making their results more difficult for colleagues to check even if they do release their code. See for instance:
If I understand it correctly those kind of computer models need large amount of computer power to run.
I would also guess that writing the software decently is more than a one man job.
I don’t know much details, but I imagine you could test the program with a subset of the data. And maybe finding a faster algorithm could help, too. If the programs are as horrible as they were described, there is a chance they are not optimal. Maybe it’s possible to start with some small parts, and gradually add more functionality.
I think it’s a quite challenging task. It needs knowledge about climate science. It needs knowledge about software architecture and organizing a big and complex project. It needs knowledge about statistics and efficient algorithms for them.
It might be interesting to think about funding sources for such a project.
Given the political nature is there someone interested in funding a project that might not provide the answer skeptics want but that also criticises the establishment of climate scientists in some sense?
Could you motivate climate skeptics to donate money via kickstarter or indiegogo for such an open source project?
What about Google.org?
Are there companies that would pay good money for better predictions about the weather in 5 years?
Now when some of these data and programs are leaked, would it be possible to make an open-source project to rewrite the code in Python, with unit tests etc.?
Could someone build their scientific carreer in climate science on fixing the software? I guess it would depend on whether the other people would cooperate: share their data and code, and use the fixed code.
Not only would it be possible, but there already exists one (http://clearclimatecode.org/) which has recreated GISTEMP in Python, found some bugs, but find that their code produces near-identical results to original GISTEMP. They say:
This ccc-gistemp project seems, however, to be stalled, in that it hasn’t released anything since 2010 (though there were a few blog posts in 2011). This doesn’t seem to be because they achieved everything they hoped to; the last information on their website indicates that there’s plenty more to do.
(This was made possible not because anyone leaked or stole anything, but because NASA released the GISTEMP code.)
Good point. I don’t know. The situation might have improved already, in terms of people releasing their code (though probably not in terms of the code being of great quality). At least, from what I’ve seen, releasing data and code is now somewhat more common. For instance, when critiquing a recent paper by Mann, Steve McIntyre wrote:
http://climateaudit.org/2014/05/19/manns-new-paper-recharacterizing-the-atlantic-multidecadal-oscillation/
I don’t know if code quality has improved since the rather embarrassing 2009 leak.
The code for the Mann 1998 paper mentioned in the post was written in Fortran, and McIntyre found it on an old FTP server of Mann’s. He rewrote the code in R. At least people are writing code in languages like R and Matlab and maybe Python now.
Some researchers still use unconventional languages for their mathematical programming, making their results more difficult for colleagues to check even if they do release their code. See for instance:
http://climateaudit.org/2013/04/22/non-centring-in-the-forest-2006-study/
If I understand it correctly those kind of computer models need large amount of computer power to run. I would also guess that writing the software decently is more than a one man job.
I don’t know much details, but I imagine you could test the program with a subset of the data. And maybe finding a faster algorithm could help, too. If the programs are as horrible as they were described, there is a chance they are not optimal. Maybe it’s possible to start with some small parts, and gradually add more functionality.
I think it’s a quite challenging task. It needs knowledge about climate science. It needs knowledge about software architecture and organizing a big and complex project. It needs knowledge about statistics and efficient algorithms for them.
It might be interesting to think about funding sources for such a project.
Given the political nature is there someone interested in funding a project that might not provide the answer skeptics want but that also criticises the establishment of climate scientists in some sense?
Could you motivate climate skeptics to donate money via kickstarter or indiegogo for such an open source project? What about Google.org?
Are there companies that would pay good money for better predictions about the weather in 5 years?