Sunday, January 16, 2011

Eli joins the Trend Wars

Well, not really, but anyone with a wind up computer knows that Tamino is the king of trends and Lucia is a wanna be. Eli, the bunny, thinks trends are interesting for visualization. Gazing at the latest versions of global temperature trends while listening to the usual chorus of nothing has changed since 1998, it occurred to the Rabett that that was the wrong way of thinking about things.

Eli has been peripherally involved in some of this, pointing out that it is best to measure a baseline in a stationary period. While it makes no difference if the trend is linear, it does affect the numerical representation for non-linear trends, and yes, there are such periods in the instrumental temperature records. Yet, with greenhouse gas concentrations going up almost exponentially and the forcing being approximately logarithmic, linear is not a bad place to start.

For this exercise Bunny Labs is gonna use the NOAA instrumental surface temperature record for no particular reason, GISS gives you the same answer pretty much. Anyinterested bunny is perfectly welcome to use HADCRUT or whatever, Eli is only so interested.

Looking at the last 50 years it is pretty clear to carrot eaters (well, maybe no, but you stare at data long enough and you learn to spot this stuff, that there was a linear trend starting in ~ 1965 and extending to ~1996. Part of the reason this may be non obvious is that most graphs of temperature anomalies are wider than high, you have to blow up the vertical axis to see this clearly. Linear trends are easy to fit, and that is what Eli did
Equally clearly there is a large positive deviation ABOVE the trendline starting with the big 1998 El Nino. (FWIW, the fast down and up deviation in Dec 2002 (down) and Jan (up) 2003 is like stinky, gotta look into that. Since GISS mines NOAA and then adjusts it, this appears also in GISS). To first order it looks like global temperatures have been very high since 1998, well above the trend line with that exception. Eli thought about how to bring this out and came up with the following metric, average the deviation from the trend line over five years. Being lazy Eli is plotting this at the date of the end of the period, so don't tell him about packing the end of a series, this is scribble not JGR which is why the trend anomaly anomaly below starts in 1970.

Ask not, dear bunnies, whether the trend in the anomaly over the last decade is zero, ask why the anomaly for the past ten years was so far above the trend. Well, go off and read Tamino for at least part of the answer. By applying crude adjustments for volcanic and El Nino/La Nina forcings, T was able to linearize the trend, returning it to its post 1965 value, or at least went a long way towards that

This is, when a hare thinks about it, a graphical version of 10 of the last 12 years (again a WAGNER, close enough and too lazy to look it up, someone will come up with the numbers) have been among the hottest on record, but it takes out the gottcha tossed at Phil Jones.


Anonymous said...

I found an 11-year moving average filter on annual-mean Temperature Anomalies gives a very interesting view. Most of the sub-decadal fuzz (ENSO, occasional volcano, etc) is suppressed, leaving the long-term "incoming tide that raises all boats". (Of course, you can't get filtered values for the first 5 and the last 5 years of your unfiltered dataset.)

--GP Alldredge

Robert Way said...

Hey Eli,
Skeptical science has a post on trends also up.