3 ms·

There was this project but it seems to be not active any more. http://clearclimatecode.org/ http://clearclimatecode.org/ Is is sad how much of very important s

There was this project but it seems to be not active any more. http://clearclimatecode.org/ http://clearclimatecode.org/ Is is sad how much of very important software is so maintenance starved.

thu2111 7y ago
I don't think it's maintenance starved exactly. It seems more like a priorities issue. One issue Nic Lewis has repeatedly highlighted is how many climatology papers contain subtle statistical errors. He finds them, papers get retracted or adjusted, but his more general point doesn't land - that climatologists should be collaborating more with professional statisticians. Well, arguably that's true of quite a few scientific fields, but I guess climatology is by now much higher impact than even economics or psychology. They could also easily collaborate with computer scientists and professional software engineering firms. Many software specialists would love to help I'm sure, even do it pro bono. But climatologists see no reason to engage such people: it seems they don't even realise their work falls short of the highest standards, or they don't care. That then opens up routes for skeptics to attack them, because surely a field whose work is used to justify such massive public spending and policy goals should have the highest standards? A lot of stuff is just footgun shooting. Lewis investigated some code for a model that underlies a lot of climate papers published over the past 20 years. It was hard to get and then turned out to be written in a proprietary programming language of the sort where you can't even find out how much it costs: Lewis tried and the firm that makes it wouldn't even get back to him. He started on a project to rewrite the code in R, but it's apparently a huge effort. The NOAA code is at least in FORTRAN. The whole field gives off a strong vibe of "here are my results, programming is easy, trust us that we got it right". Of course professional programmers know programming isn't easy at all.

There was this project but it seems to be not active any more. http://clearclimatecode.org/ http://clearclimatecode.org/ Is is sad how much of very important software is so maintenance starved.

thu2111 7y ago

I don't think it's maintenance starved exactly. It seems more like a priorities issue. One issue Nic Lewis has repeatedly highlighted is how many climatology papers contain subtle statistical errors. He finds them, papers get retracted or adjusted, but his more general point doesn't land - that climatologists should be collaborating more with professional statisticians. Well, arguably that's true of quite a few scientific fields, but I guess climatology is by now much higher impact than even economics or psychology. They could also easily collaborate with computer scientists and professional software engineering firms. Many software specialists would love to help I'm sure, even do it pro bono. But climatologists see no reason to engage such people: it seems they don't even realise their work falls short of the highest standards, or they don't care. That then opens up routes for skeptics to attack them, because surely a field whose work is used to justify such massive public spending and policy goals should have the highest standards? A lot of stuff is just footgun shooting. Lewis investigated some code for a model that underlies a lot of climate papers published over the past 20 years. It was hard to get and then turned out to be written in a proprietary programming language of the sort where you can't even find out how much it costs: Lewis tried and the firm that makes it wouldn't even get back to him. He started on a project to rewrite the code in R, but it's apparently a huge effort. The NOAA code is at least in FORTRAN. The whole field gives off a strong vibe of "here are my results, programming is easy, trust us that we got it right". Of course professional programmers know programming isn't easy at all.