Jamie Andrews

How to move climate science forward

Yesterday, I watched the Ad Hoc Science and Technology Committee hearing about “The disclosure of climate data from the Climatic Research Unit at the University of East Anglia”. This is the result of so-called “Climategate” when the University of East Anglia Climate Research Unit emails were hacked. You can watch a recording of the hearing here.

A few things struck me.

First of all, I was surprised that Lord Lawson’s Global Warming Policy Foundation was represented as the only NGO. Given that all other witnesses questioned were formally linked to the CRU or the Information Commissioner’s Office (ICO), I thought it was odd that what is effectively an anti-global warming lobbying organisation was given such credence.

Having said that, it was nice to see some of the more ridiculous claims by Lord Lawson answered comprehensively. For example, he asserts that ‘one single tree’ has been used by climate scientists who use tree rings to trace historical temperature records (outrageously reported by the Telegraph here). This was clearly rebuffed as part of the hearing, which was refreshing. Whether anyone who believes statements by Lawson and others like him will ever see the hearing is doubtful given how selective climate ‘skeptics’ seem to be in the information that they take in.

But the most interesting thing I observed in the hearing was when Dr. Phil Jones (head of the CRU and author of some of the most controversial emails) was being questioned by one of the panel about what constitutes ‘standard practice’ in academic practice. The questioner (who was generally quite hard on Dr. Jones) felt that it was wrong for the full data and computer programs behind scientific papers not to be made available at the time of publication.

Jones made it clear that all scientific methods were clearly documented in the papers, and that it wasn’t standard practice in the field to make all data available when publishing a paper (therefore clarifying that he wasn’t at fault for not doing so from the outset). But he eventually agreed that it ‘might be a good idea’ for all data and computer programs to be made available as standard practice when publishing papers (as opposed to the status quo, where data sources are cited but not actually ‘attached’ with the papers).

It strikes me that taking a coordinated approach to online dissemination of data would help with this issue. For a tiny research unit the size of the CRU (who have a full-time staff of three) it would be much easier to point directly to a URL (the modern equivalent of a footnote) with standardised and transparent data sets accessible rather than collate and re-publish all relevant data-sets themselves. This is exactly the kind of thing that Tim Berners-Lee has been advising the UK Government on (‘opening up’ data) and also why the company I work for (AMEE) has built AMEE Explorer.

The discussion of ‘standard practice’ (see around 1 hour 11 mins into the video of the hearing) seems key in the debate about how climate science can be fully transparent, and I’d like to see some leadership from academic institutions (perhaps research councils) and Government (through the ICO) about how to enable this transparency without massively increasing the Freedom of Information request burden on research units such as the CRU. Hopefully the committee will conclude something along these lines.

Activity

3 total comments, leave your comment or trackback.
  1. Hi Jamie

    Some interesting thoughts here.

    I’m not so sure of the value of making all the data publicly available as a matter of course – in the past I’ve spent days if not weeks re-formating data files so that they can be uploaded to national data archives. In the CRU case, very few people would be that interested in the raw data – the whole rationale behind CRU’s work on surface temperature time series is to produce a usable product that has been properly quality controlled. The only reason that certain people are desperate to get CRU’s data is because they’ve been told they can’t have it (from CRU, who don’t “own” it). In most cases, I’m sure if you went to the data owner with a legitimate scientific purpose, then they would give you the data – this is exactly what CRU did.

    As for code, I suppose if it had to be made public I’d probably programme a bit more elegantly but, like Jones, no reviewer (or even paper co-author!) has ever asked to see my code. Maybe it should be submitted with papers as supplementary material but I doubt anyone would be particuarly interested in it – at least in that case it would be there if there were ever disputes over papers. I’m not sure how impressed publishers would be at having to archive GBs of code though!

    I assume you saw my review of the evidence session, I don’t think Phil Jones did a good job describing how the scientific process works but I also don’t think that reproducability should mean that you have to give anyone everything you’ve ever worked on. There are issues here but I hope that knee-jerk solutions aren’t imposed because of this case.

    Andy

  2. admin
    Mar 2nd 2010

    Hi Andy,

    Thanks for your comment.

    My point about making ‘raw data’ available is not that the CRU or any equivalent academics should be responsible for doing any re-formatting (that’s exactly why we need a *standard* way, so that they can focus on science), but that now in the age of the internet we can begin to share data more easily and allow it to be accessed publicly as a matter of default. Taking inspiration from sites such as the Guardian data store – http://www.guardian.co.uk/data-store – this would hopefully make conclusions easier to interrogate, and more to the point stop giving people like Lawson fuel for their witch-burning fires.

    Re. program code, have you come across Github – http://github.com ? It’s basically a version control system for that works in a similar way to other software such as Subversion but makes it very easy to make a codebase ‘open’ online. Given that there is very little to hide once a paper has been published, it seems that we could better encourage scientific collaboration in this way (the codebase should only be opened once conclusions have been drawn).

    The idea being that once you have open data + open code, we can reinstall some rational debate about science rather than the shrill squawking about whether or not datasets or code are open. I thought it was apt that Dr Benny Peiser had to sheepishly admit that he is not a climate scientist – perhaps following some of the above suggestions would more acutely highlight his lack of expertise.

    There may indeed be risks in this approach if raw data is further construed to misrepresent warming trends, but I agree with the statement in the hearing that science must be as open as possible, and using the internet to its full potential seems to be a good way to do this. I also have faith in most humans to generally make their mind up independently when all the facts are in front of them, and I think that something must be done to halt the worrying drop in public belief in climate change.

    To reiterate, I think that opening up historical data is not something that the academics themselves should have responsibility for, otherwise we’ll get more frustration and productivity bottlenecks such as the ones catalysed by the FOI requests at CRU. But going forward we can instil a culture of openness more proactively by using web-based dissemination techniques.

    Jamie

  3. JW Moore
    Mar 3rd 2010

    Good stuff. Thanks for the link to the entire P. Jones hearing. The real scoop here — picked up by just a few observers and of course ignored altogether here in the US — was that CRU has just THREE full-time climate scientists!
    I agree totally that all work must perforce be open to all scientists (and anyone else for that matter) to ensure the integrity of the conclusions being published in the literature. The notion that peer-review confers some seal of authenticity is just one of the casualties of “climategate.”
    ~Jack


Leave a Reply


Search

Feel free to search older content using topic keywords.

Browse by Category