Bishop Hill A Russell bug
Jan 26, 2011 Climate: Surface John Graham-Cumming emails to point me to his latest blog post, in which he outlines a small bug in the code used by the Muir Russell panel in their (kinda, sorta) replication of the CRU temperature series. This was spotted by someone called David Jones, who I think is something to do with Nick Barnes' Clear Climate Code project. The problem appears to be that the Russell version of the code doesn't weight cells by area.
The impact doesn't seem to be enormous.
The warming trend shape doesn't change, but the temperature anomaly does alter. In recent years the unweighed average is greater than the weighted. For example, grabbing 1998 to 2008 at random the differences ranges between 0.02C and 0.07C with an average of 0.06C. So the upshot of the ICCER bug is that it makes things seem slightly warming.
Reader Comments (20)
"So the upshot of the ICCER bug is that it makes things seem slightly warming."
I would have put money on the fact that any bug would produce higher teperatures!
One bug does not a summer make (Aristotle couldn't have said it better, he used swallows).
Code verification should be part of the application of quality assurance. I don't think much of that goes on in the climate field.
Glad to see the Clear Climate Code Project. Thanks for noting it.
Funny how 'mistakes' only seem to effect the results in a one way , If you did not known better you suggest that is because the bases are so loaded in that direction that is much more likely to have mistakes in it than in other ones.
It's not true that the bugs all go in one direction. For example, the error I identified in the Met Office calculation of uncertainty (error ranges) when fixed makes the error range smaller.
JG-C
I am sure you are right - but I bet a warming bug is overwhelmingly found.
What we need is someone to analyse the data and produce a graph.
My bet is that it looks like a Hockey Stick!
Interesting to observe that the independent review, which recommended greater openness, did not think to release its own code at the same time as its findings, and subsequently becomes aware of issues in its methodology, as a result of independent review.
I suspect that change in the practice of climate 'science' will be outpaced by a snail. (A dead snail).
The BBC's Paul Hudson has a post today on the warm bias of the Met Office.
http://www.bbc.co.uk/blogs/paulhudson/2011/01/global-temperature-forecast-fo.shtml#comments
He states their central prediction for 2010 was 0.55 above the 1961-1990 average (they are reporting 0.50 as the final figure) and that this was the 11th out of 12 years when their prediction overshot the final figure. I have their central prediction as 0.58 for 2010 - which I can't now find on their site - though this article confirms my figure:
http://www.independent.co.uk/environment/climate-change/next-year-forecast-to-be-hottest-on-record-1838184.html ).
Their predictions, compared to the outcomes, can be found on the following post with the final figure for 2010 being 0.50.
http://climateedinburgh.blogspot.com/2010/12/official-bias.html
Without meaning to excuse unexcuseable conduct, it is not unreasonable for errors in any calculation on any subject to move in the expected direction. Accountants know this falacy well.
If you make an error which leads the results in an unexpected direction, then you think "that's funny" and go check you work. If you find your error, you correct it and move on.
If your error takes the results in the direction you expect, then you probably do not even notice your output is wrong (unless the error is hugely too great).
Having said that, certain results in climate science have been, well slightly unbelievable, and replication by others has been, shall we say, rather difficult to achieve.
For the accountants amongst us, if the Trial Balance balances, there is a great tendency to sigh in relief and move on.
If the TB does NOT balance then you KNOW something is wrong and you are forced to retrace your steps. Even a one cent error can be the signal of two almost compensating errors, or quite frequently a string of such errors.
Accountants are human too, despite appearences, or certainly were in my day.
I followed the link to Paul Hudson's article by Cameron Rose above. If I'm not mistaken, that stems from my work. Here is a comment I will be posting to that blog, once my beeb registration goes through.
Those who follow all the arcane minor disputes will remember this as the cosine of latitude weighting. Past errors have included Ross McKitrick hitting a degree/radian issue and Michael Mann hitting a square of cosine issue for weighted variances; no prizes for guessing which of those two openly acknowledged their error and corrected it.
In this case, if there has on average been more warming away from the tropics, failure to weight would indeed tend to slightly overstate a global trend.
Why would he use the cos of the latitude of the middle of the area – cos(($lat+$grid_size/2)*$pi/180) – when it's almost as easy and much more correct to integrate over the latitude range?
That is: sin(($lat+$grid_size)*$pi/180) - sin(($lat-$grid_size)*$pi/180)
This will work better if the grid item is near the poles or overlaps the equator or is simply large.
Why do an approximate and wrong calculation when an exact and correct one is almost as easy?
Correction: not having access to the definitions being used for the grid cells, this is in keeping with the convention implied by cos(($lat+$grid_size/2)*$pi/180):
sin(($lat+$grid_size)*$pi/180) - sin(($lat)*$pi/180)
@John M
You are too modest. Clearly Harrabin and Hudson are trying to take credit for 'discovering' the warming bias of the Met. You are perfectly entitled to press them on this.
I left a message on WUWT's Tips and Notes section. This story is worth posting here, there and every other skeptical blog.
@ John M
"I followed the link to Paul Hudson's article by Cameron Rose above. If I'm not mistaken, that stems from my work."
Just for clarity, it was Paul Hudson who raised the issue of warm bias in his post on his website of 9.1.11 thus:
" Could the model, seemingly with an inability to predict colder seasons, have developed a warm bias, after such a long period of milder than average years?. . . "
I picked up on the issue using (and referencing) the information originated by you. (I was using an alter ego - I think I had had difficulty registering.) And I used the url shortener which redirected and does redirect to the WUWT post on which you commented.
I appreciate the useful work you have done.
"11th out of 12 years when their prediction overshot "
@ Cameron Rose
Null hypothesis: Met Office forecast is unbiased (equally likely to be too low or too high)
p value 0.00325
Conclusion, Met Office forecast is biased at a statistically significant level
Now where is that chap who was 'auditing' the Met Office forecasting performance after the fiasco this winter (so far) ?
Cameron Rose,
Thanks. I in no way meant to imply that I was not properly acknowledged in the comments. Personally, I don't really want any acknowledgment anyway, but if the MSM is suddenly going to "discover" something, they ought to admit they're not the first on the scene and the skeptic blogs were pointing these things out long before their sudden eye-opening.
I didn't point this out, but the Harrabin link on Hudson's report saidsomething like "the MET has now admitted to the BBC", like it was the BBC that ferreted it out.
What a bunch of heroes! :)
Hello.
Originally posted by biferi: thanks but can you give me a direct link to the Download file I went to the web site and it just is moving slow for me and the site is Gerking around with my Mouse.And can this program do a good job did you try it? the above link is the direct one all you do is click the download button and the download will startbadbak
[url=http://www.dvd-to-avi-converter.net/]dvd to avi converter[/url]
[url=http://www.best-webhosting-services.com/wordpress-web-hosting.html]wordpress web hosting[/url]