Thursday, February 24, 2011

Antarctic, RO10, Steig and TempLS.

At the end of his RC post (in happier days), Eric Steig said of the OLMC10/S09 situation: "This probably means going back to the drawing board to write up another paper".

Well, this post suggests another method, using TempLS v2. TempLS was a program written last year, originally to provide alternative calculation of the main global temperature indices. It did that. It works on a different basis to most other such codes - instead of gridding anomalies, it fits a linear model to global temperatures by weighted least squares.

Version 2 extended this to fitting spatial variation parametrised by coefficients of basis functions, intended to be families of orthogonal functions like spherical harmonics. But the EOF's introduced by Eric Steig for the Antarctica analysis would do as well.

This post describes some preliminary results. There are some loose parameters which will need better definition. I have done almost no verification stats. The trends come out rather high. But the patterns are quite similar to the O10/S09 results. And their are some big plusses. One is simplicity - run times are a few seconds instead of the 40+ minutes I found with the RO10 method. And the simplicity means that one can experiment with more things - eg spatial weighting.

Sunday, February 20, 2011

Ryan's code - testing.

I set out to do, as promised, a repetition of Ryan's sensitivity test with S09 using more AVHRR PCs. But I was distracted by what seemed to me to be a simpler and more informative test, which I'll report here. It gives a quantitative comparison and also can act as a kind of calibration.

I detrended all the raw station data. That is, I estimate and subtract the trend for each station, leaving the mean over the observation period unchanged. So we have data which should return a zero trend. What do we get?

Thursday, February 17, 2011

Ryan's code - S09 with more PCs

In my last post, I was stuck. With help from Carrick (see update), I've been able to get a working copy of the code. I've revised the HTML markup file.

A full run takes about 46 min on my 32-bit Windows setup. Most of that is iridge. If I take out just the parts for S09, it's about 80 sec. So I did that, and started tinkering with the number of retained PCs. Ultimately, I want to see how much of the difference between RO10 and S09 is due to this factor.

For the moment, I have just plotted the all spliced trend maps for the period 1957-2006 for retained PCs from 3 to 7, using the RO10 implementation of Steig's method. Here they are:

Monday, February 14, 2011

Ryan's code

I'm trying to run Ryan O'Donnell's code for the Antarctic paper (O10), in the form that he posted at the CA site. The datafiles are there too. The code is called RO10%20Script.R.

Update 16 Feb. Carrick says in the comments that he used the file RO2010 Code.txt, in the same dir, and that it ran to completion. There is a readme.txt file which mentions this latter file and not the .R file. The .txt file does not have the apparently defective code that I report being stuck on below. I am currently running it, and it reaches the error pointed out by JohnA. This is at a much later stage in the running, so that is good news.

The confusion over the codes is puzzling. In that same CA thread, pax noted an error which I nad also encountered, and Ryan said that he had fixed it. In fact, he updated the .R file, not the ,txt one.
Anyway, I'll switch to the .txt file. 

  References for  Antarctica posts:
The discussion is prompted by two papers:
S09, a 2009 Nature paper by Steig et al and
RO10 (or O10), A 2010 J Climate paper by O'Donnell et al.
There has been heated blog controversy about these papers - this post is just one entry point. Here is a Real Climate post following S09's original appearance, and here is Eric Steig commenting on O10 (this led to other heated postings).




It's a long file. Eric Steig said that he found it hard to follow, preferring that it be broken up into separate files. But for turnkey effect, one file is more convenient.

When I'm trying to understand how a long program works, I often get into it with emacs and make a html version with links. Then one can make a table of contents, and link function calls to their definitions etc. And also color things. I've done that online here, and the source is available in a file (ryan.htm) I've uploaded at the document depository. It's not for running - just an aid to navigating around the code.

Friday, February 11, 2011

On words

There has been more recent discussion, partly inspired by Steig and Trenberth, on the use of a word that is also a silk measure. I don't use the word, though I don't see anything wrong with doing so. I refrain because it leads to time-wasting arguments.

The preferred alternative is skeptic. I have used that, spelt with a k. Spelt with a c, there's a danger that consistent pronouncers will make the c soft. I might even omit it in text, which would lead to trouble.

Wednesday, February 2, 2011

Tropical Cyclone Yasi

This is a monster. I just hope for the best for people in North Queensland. WUWT's latest post gives thorough coverage, although I wish they would let alone the premature whinging about people talking about AGW.