Home > Uncategorized > FAQ Launch

FAQ Launch

2010 August 5

Just thought I ought to make a permanent home for stuff like D’Aleo at Heartland: An Apple a Day. The FAQ concentrates just on the land surface, ignoring sea surface for now.

Feel free to use the comments here to give suggestion, ask questions, or remind me of links I’ve left out. After I let this mature a bit, I’ll open comments in the FAQ itself.

dali

Advertisements
  1. 2010 August 6 at 1:26 am

    Excellent stuff!

    The opening question “What surface temperature records are available?” talks about analyses and reconstructions. The question could be construed to mean station records (in other words, GHCN, SCAR READER, USHCN, and so on). Maybe the next question should be the “what surface data is used” question that is currently lower down.

    “Why isn’t more code available?”, your answer: “Climate scientists have typically published descriptions of their methods for data processing rather than releasing their code directly. This allows interested third parties to independently reconstruct their methods which is a more robust form of confirmation than simply recompiling and rerunning the original code.” doesn’t address the question. Your answer speaks to why one might not (should not?) be interested in the code, not why it hasn’t been published. Personally I suspect the real reason is closer to “Scientists can’t be bothered, and they’re not required to”. Possible with a bit of embarrassment and secrecy thrown in.

    The first sentence of the answer is good, but could still be reinforced: “It is not typical in scientific publishing for source code to be published;”.

    “What about satellite records?” it isn’t clear (and I don’t know) whether the two records are using data from the same satellites or merely the same sorts of satellites.

    “Where is the surface data?” GISS doesn’t actually have any addition data for St Helena; what it does do is shift a record “by hand” to account for a discontinuity (which you can see here: http://data.giss.nasa.gov/cgi-bin/gistemp/gistemp_station.py?id=147619010000&data_set=0&num_neighbors=1 ). (Due to a station move, by the way).

  2. 2010 August 6 at 5:31 am

    re St Helena: Doh!

    I’ve updated the easy ones. I’ll think more about the “Why isn’t more code available” answer. I don’t think I’ll go with the “its not typical” answer since that seems to vary widely depending on the field.

  3. carrot eater
    2010 August 6 at 12:45 pm

    I don’t know much about the satellite measures, but I thought RSS and UAH were using the same satellite data, up until recently, and now they use different satellites (Aqua and NOAA-xx, I forget what # they’re up to now)

    GISS also does a manual adjust of Lihue.

    for more code: USHCN code is up.
    http://www.ncdc.noaa.gov/oa/climate/research/ushcn/

    go down to ‘pairwise’ software. GHCN code will probably be up as well, once they relaunch it with the pairwise method. Beyond the fact that publishing code was never the norm in many scientific fields (maybe it was in others), you have to consider to what extent something is fully automated and objective. You can’t push a button and have the code replicate the exact same thing again, when there’s a human judgment step in there.

    On the plus side, a fully automated process is objective, fully reproducible, and extremely transparent. Don’t understand something? With a bit of effort, you can see exactly what happened and why. On the downside, scripted algorithms are literal idiots, and now and then they’ll do something that may not appear to make sense.

  4. carrot eater
    2010 August 6 at 12:51 pm

    oh, and not always releasing code has basis in more than just norms. code is IP, and science is very competitive. You say in your paper what you did, and others should be able to work it out. You don’t feel compelled to spoon feed your competitors with all your spreadsheets and intermediate calculations.

    Now there comes a time though when some codes should just be public resources, open source, and things like GISTEMP fit. But most papers being published don’t have anything like GISTEMP.

  1. No trackbacks yet.
Comments are closed.