One of the interesting features of the Earth Trends Modeler is the ability to compute Empirical Orthogonal Teleconnections (EOT). The intention of the technique (described in more detail at the end of the end of this post) is to uncover the major underlying patterns of variability in the analyzed series over space and time.
The first EOT in monthly anomalies in sea surface temperature from 1982-2007 (above) is the familiar El Nino / La Nina (ENSO) phenomenon -- not surprising, since it is unquestionably the dominant pattern of interannual variability in sea surface temperature (SST). However this post concerns the second EOT (below) which presents a less familiar pattern.
The second EOT is the largest pattern of space-time variability in SST anomalies that the technique can find in the residuals from ENSO (i.e., after the effects of ENSO have been removed). The image above shows the spatial pattern and the graph below shows the temporal pattern. Areas with high positive values in the image are strongly associated with the temporal pattern (and vice versa).
The space/time portrait of EOT2 is dramatic. First, most of the world's oceans show warming (i.e., a positive association with increasing anomalies in temperature after the mid 1990's). Second, the warming is most pronounced in the Atlantic. The image below shows the temporal graph of EOT2 along with an index to the Atlantic Multidecadal Oscillation (AMO) superimposed (in red) -- a low frequency Atlantic SST oscillation with what is thought to be a 65-70 year cycle associated with the thermohaline circulation.
Clearly the temporal evolution of EOT2 is a VERY close match to the AMO index (r = 0.71). However, the AMO is also the subject of significant controversy. The issue is the extent to which the pattern of Atlantic warming in recent years is attributable to a natural climate cycle (the AMO) or to global warming. Look for more on this issue in the next posting.
About the technique:
The Empirical Orthogonal Teleconnection technique (as implemented in the Earth Trends Modeler) searches for the single location in space (a pixel in this case) whose temporal profile can best describe the temporal evolution of all other locations. The profile at that location becomes the first EOT. A residual series is then created such that the effects of the first EOT are removed from the original image series. The process is then repeated to find the next EOT, and so on. In our implementation, after all the designated number of EOT’s have been found, a multiple regression is run with the original image series as the dependent variable and the EOT’s as independent variables to create partial R images as a spatial portrait of the EOT.
EOT's are similar to obliquely rotated Principal Components. They are independent in time, but not necessarily in space. They have the advantage that they are easily understood and are associated with a specific location.