New Computer Fund

Sunday, July 22, 2012

Reanalyse This

UPDATE:  Solar cycle added at end:

In Analyse This I described an anti-cherry picking method of using linear regressions on non-linear data.  You use every possible time period for a series, perform a linear regression for each period and plot the slope of each regression.  Open Source has a LINEST(a:b) function that works perfectly for that task.  There is nothing particularly note worthy about the process, other than it is easy to do.  In fact, it is not more informative than a good eyeball.

When comparing a number of time series, it does have a few advantages.  Above I compare all of the UAH ocean data sets.  At the end, the data set start synchronizing.  At the beginning the sets are somewhat synchronized and in the middle there is the least synchronization.  That is for the 60 month or five year lengths of the series.

In this plot, I use 180 month (15year) regressions and limit the chart to Global, Tropics, Northern and Southern extents.  The first arrow shows the peak slope for the tropics.  The other two arrows point out the peaks of the other series.  The warming due to the 1998 El Nino apears to have immediately impacted the tropics and to a lesser extent the rest of the ocean area, then there appears to be a delay of about eight years before that peak is seen in the rest of the series.  

The 1998 El Nino event is a natural event.  The magnitude of the event may be amplified by CO2 forcing, but then the following El Nino events would also have been amplified if that was the case.  Note that the yellow tropics curve becomes positive in 1995.  That may be due to CO2 amplification, then it returns to neutral in 2002, again just after 2008 and 2011.  The general slope of the tropics from 1998 is negative.  That is not in keeping with CO2 amplification.  It is not too unreasonable in my opinion to consider a good portion of the ocean temperature change as being due to a natural event with a delayed response.  Looking at just global temperatures, that would not appear to be the case.  

In this chart I converted the HADSST2 data set from temperature anomaly to approximate energy based on the average ocean surface temperature.  That conversion 5.67e-8*(294.25+Tanomaly)^4, amplifies the change by the T^4 power.  So the chart shows an approach to ~425.8Wm-2, an impulse perturbation around 1998 and a decay back to ~425.8Wm-2 following that perturbation.  This is not what I would expect to be a feature of CO2 enhancement.

Unfortunately, I don't have the new HADSST3 data set divided into the same regions.  If I did, I would suspect that I may be able to tease out a time constant or two from the longer time series, if they are accurate of course.  So far though, the oceans do not appear to be responding to CO2 forcing as much as "other" factors, some very natural in origin.

This plot is using 132 month (11year) regression lengths which should show some of the common solar cycle variation between sections of the oceans.  1998 still stands out as exceptional plus the 1995 bump in the tropical ocean regression is more prominent.

For the land fans.

No comments:

Post a Comment