West Antarctica: still warming

O’Donnell et al. have three main criticisms of our work. First, that the reconstruction we reported was not homogenous. That is, the first part of the reconstruction (1957 through 1981) is based entirely on a linear combination of weather station data (since there are no satellite data during that period); while the second part (1982-2006) is derived simply from the satellite data. O’Donnell et al argue that it would be better to use the only weather station data for both periods, since these data are a priori considered more reliable. (There are all sorts of potential problems with the satellite data, the chief one being that there is a ‘clear sky’ bias.) That is, one wants to calibrate the data during 1982-2006, and then use that calibration to model the temperature field for both the early and the later periods, using only the weather stations.

Second, that in doing the analysis, we retain too few (just 3) EOF patterns. These are decompositions of the satellite field into its linearly independent spatial patterns. In general, the problem with retaining too many EOFs in this sort of calculation is that one’s ability to reconstruct high order spatial patterns is limited with a sparse data set, and in general it does not makes sense to retain more than the first few EOFs. O’Donnell et al. show, however, that we could safely have retained at least 5 (and perhaps more) EOFs, and that this is likely to give a more complete picture.

Third, O’Donnell et al. argue that we used too low a truncation parameter when doing the ‘truncated least squares’ regressions. In general, using too low a truncation parameter will overly smooth the results, and tend to smooth both temporal and spatial information. The problem with using too large a truncation parameter is that it creates problems when data are sparse, resulting in numerical noise (overfitting). O’Donnell et al. try to get around this problem by using cross validation — that is, trying a bunch of different truncation parameters, and using the ones that give the maximum r2, RE and CE statistics.

There are a number of other criticisms that O’Donnell et al. make, such as whether it is okay to infill the weather station data at the same time as doing the calibration against the satellite data (as we did) or whether these have to be done separately (as O’Donnell et al. did). These are more technical points that may or may not be generally applicable, but in any case do not make a significant difference to the results at hand (as O’Donnell et al. point out).

Let’s assume, for the moment, that all of these ideas are on the mark, and that the main reconstruction presented by O’Donnell et al. is, in fact, a more accurate picture of Antarctic temperature change in the last 50 years than presented in previous work. What are the implications for Antarctic climate? How would they differ what was concluded in Steig et al. (2009)? The answer is: very little.

The spatial patterns of annual trends, and how they evolve through time, is similar in both papers. In particular, O’Donnell et al. find, as we did, that the entire continent was warming, on average, prior to early 1980s (Figure below from their main “RLS” reconstruction). As we said in our paper, this would tend to support the idea that cooling in East Antarctica is a recent phenomenon at least in part attributable to recent trends in the Southern Annular Mode (SAM), which is itself forced (at least in part) by stratospheric ozone depletion.

Page 2 of 7 | Previous page | Next page