Talk:Distributed computing climate change model gives bleak results

From Wikinews, the free news source you can write!
Jump to navigation Jump to search

Recently, authors of the project admitted that they have made a fatal error in the program, so the project should be completely recalculated "Message From Climateprediction.net PI Re: Recent Problem" —  Thus, the article looses any ground.

Fatal error? hardly. A single header in a single file which will require a re-run of the models. Yes, it may cost a couple months of calculations. - Amgine | talk en.WN 04:19, 20 April 2006 (UTC)[reply]
But you belive that the results will be the same? :)
I would retract the findings before I re-run the model, because my (or yours) beliefs have nothing to do with what is called "scientific results".
I agree with the anonymous commentor. The problem is not that it takes months to rerun calculations; the problem is that the current result is meaningless. Therefore the description of the results of experiments, as is, is misleading. Cema 04:55, 20 April 2006 (UTC)[reply]

It struck me greatly that nobody from the commentators above had not apparently attempted to contact the authors of the paper in question before posting (especially the one with a pathetic statement about beliefs vs scientific results). Well, I did. What follows below in italic is the essence of the authors comment:


The paper in Nature was based on simulations carried out as part of the first climateprediction.net experiment which looked at the response of the model to doubling the atmospheric CO2 concentrations. We now have well over 100000 simulations of this sort without any errors of the sort you refer to.

In February 2006 we launched an experiment in collaboration with the BBC which was designed to simulate the period 1920 to 2080. The error you refer to applies only to these simulations i.e. only simulations started in the last 3 months.

So in answer to your question, the recent error has no relevance at all to the results presented in Nature 433, 403-406 (2005) so there is no reason to send an erratum.


I believe that this point nicely completes the discussion.