This is a guest post from Ryan O. Ryan was principle author of our submission on Antarctic temperature reconstructions which used PCA based methods for combination of satellite and surface data. This post has some very loose basis in that work. When I first learned of ‘teleconnections’ it was at Climate Audit while looking at some paleo papers. I thought it was a joke but quickly figured out that it was real and based on the potential for weather patterns in distant regions to be related/correlated. For instance, if you have stable prevailing winds over long distances and time. (i.e. E-W over the US). You might find some kind of connection with the weather in Chicago to the farther west- Iowa but less to to the south or north. This is about as simple as it can be stated, b/c some suggest negative teleconnections, polar ones, and all kinds of wild stuff.
PCA analysis in data with spatial correlation will bring out patterns similar to vibrations on a drum head. Because the patterns in the drum head aren’t perfectly symmetric from climate data, it’s easy for the human mind to attach spurious climate meaning to them. In this post Ryan used VERY general properties of satellite data to create an otherwise random synthetic data set and then created nearly identical PC patterns to the actual climate data as measured by UAH.
————
Ryan O–
One of the favorite concepts popularized by climate science is that of the “teleconnection”, where large-scale dynamics are inferred from inspection of patterns resulting (primarily) from EOF or PCA analysis. These patterns are assumed to have physical meaning, either directly or indirectly. The concept is used in certain paleoclimate reconstructions (such as MBH 98) where proxies are assumed to “teleconnect” to the average earth temperature. It also appears in some climate-related EOF/PCA texts and in analyses of atmospheric pressure, precipitation, ocean current, and a myriad of other related topics. Given how pervasive “teleconnections” are used to study our planet, one might assume that they have a solid theoretical foundation.
As with other things in climate science, one would be wrong.
The issue of teleconnections and physical meaningfulness has come up here before – most notably with the work on the Steig Antarctic temperature reconstruction. The primary point of contention there was that Steig limited his analysis to the first 3 eigenvector pairs by using the “physical meaning” argument and a sampling error argument from North (1982). Steve McIntyre showed that very similar (and physically non-meaningful) patterns result from EOF analysis on an object shaped like Antarctica with exponential correlation functions and precisely zero physical dynamics.
Read the rest of this entry »