Tuesday, February 17, 2009

Update to MM07 and S07 Analysis

[It is 2/19/2009 and the conclusions of this post have changed. I'm not sure what I did before but I now get a somewhat different result than when I first put this up. ]As I was writing my previous post about MM07 and S09 Dr. Schmidt was responding to me at realclimate. Dr. Schmidt says that he picked Top/Right of the four choices which would be the same as MM07. I'm sure he is right, but the results look more like one of the two bottom choices from my analysis.

Anyway the point isn't that important because Dr. Schmidt agreed with me that the best way was to average the four boxes to match the 5x5 degree surface grid. He recalculated the result and put it in his supplementary information. He reported no significant change in results. I actually get a somewhat different result with the regression using RSS data yielding significant correlations with the economic factors.

I have rerun the UAH and RSS analysis using the 5x5 grid, and the results are below. The UAH result matches MM07, but for me the RSS result doesn't completely match S09. In the case of the RSS case some of the significance to economic factors drop, but not as much as stated in S09. I don't know what is causing the difference, as the process for converting the RSS set to temperature trends isn't documented in S09.

[update later on 2/19] As I was looking back at the S09 SI I realized that S09 used an updated surface temperature set in addition to using the RSS data. The RSS data alone doesn't really change the results from MM07 but the two together just push the correlations for some of the economic data above the .05 significance level. I show the results both ways in an attached PDF. The result with the updated surface data and the RSS data looks like S09.

I note that the updated surface data still gets essentially the same results with the 5x5 UAH data.

I have to say that until I understand why the RSS data has such a different standard deviation than both the UAH trop data and the surface data I'm not sure that it is valid to conclude there is an issue with MM07 from the fact that the model doesn't work well with the RSS data. As I mentioned in my previous post the model works well without any tropospheric data at all.

Meanwhile the issue of spatial correlation keeps coming up, so I guess I will look into that as well.

Updated code is here.

Here are the results using RSS and UAH and the 5x5 grid.

No comments:

Post a Comment