Over in Suggestions Tenuc is tickled by Roy Spencer and his for amusement only Excel spline, not that Tenuc is right, just that Roy being a gent doesn’t want to let on Phil Jones helped him. Err, oh, I’m definitely regressing. (any strangers reading this, is gentle in-house humour)
The data used here is computed from published gridded by me. (matches but more could be extracted)
10 year to the dataset ends, educational purposes only. (tends to flatten the ends slightly) There is no Met Office / Hadley / CRU hockey stick generator with me. (I assume readers know about that Met Office red face on being caught)
I’m showing RSS and UAH because they are noticeably different with UAH showing more recent warming. Perhaps this is a surprise to some readers.
Dataset differences: RSS omits poles, UAH covers, RSS omits a few high mountains, UAH covers, RSS has more high frequencies. I could plot the two as maps. (haven’t because I am tired, maybe later if people ask)
Here is Roy Spencer on the same data
This is a local copy. (original)
Over to you folks. I’ve included other data in following if anyone wants to have a play.
Datafile here including RSS and UAH tlt, tmt and tls datasets global, north and south hemisphere. This is an OpenOffice .odt, change to .ods if needed. (if necessary I will have to export to .xls and upload that as well)
Had couple of hours snooze so here is some more and I hope illuminating to a few. Only takes a few minutes to hurl the datasets at some of the software, twiddle and plot.
A couple of posts ago Nicola Scafetta ticked off Mosh over what many call regression, the matter of the old trick of withholding known data. In this case I just said do a quick 7 term model on automatic, but on data to end 2003/4 (forget which). r2 is circa 0.7 – 0.8 and makes a sane job over the period where it can see data. Won’t match fine detail, not intended.
As you can see this dataset is NOT predictive, extrapolate fails. I’ve added traces for the whole model and the long period term only (with offset).
I think rather nicely this demonstrates the simplistic carry on upwards vs. nature speaks. I add, the software is in two minds on whether to lock to 45 – 70 years but rejects this. (mentioned because I am aware things are not as simple as I show)
Above is just a high speed hack.
Interesting aside here. Probably before Nicola S. showed his recent temperature model I had done the same thing but I kept silent. Reason is finding very interesting prediction using Hadcrut3 *but* this only works on the gridded data *without* Met Office / CRU / Hadley statistics post processing. I also found it was ambiguous, either carries on level then another rise to another plateau circa 2060, or things fall. This is useless. It can give insights though. (talking extrapolate from as far back as 1960)
I’d better stop.
Posted by co-moderator