DISCUSSION OF: A STATISTICAL ANALYSIS OF MULTIPLE TEMPERATURE PROXIES: ARE RECONSTRUCTIONS OF SURFACE TEMPERATURES OVER THE LAST 1000 YEARS RELIABLE

摘要

I join the authors in expressing dissatisfaction with some paleoclimate analyses. I endorse their claim that there has been underestimation of uncertainty in paleoclimate studies. The implication that additional participation of the statistics community is needed is undeniable. However, our priorities should be to contribute rich statistical analyses that (i) model the processes and data and (ii) offer useful information regarding the issues of climate change. If achieving these goals requires that we do not continue with questionable assumptions, nor merely offer small fixes to previous approaches, nor participate in uncritical debates, so be it. The authors note that it is common to assume that proxy observations are linearly related to climate variables and they proceed with this assumption. This seems untenable to me (for an extreme example see the Yellow River data in Figure 6). Even if linearity is plausible, lumping all spatial-temporally distributed data of various types, qualities, and degrees of relationship to climate variables into a variance—covariance based summarization (principal components or EOFs) with no underlying analysis gives me pause. I am not surprised by difficulties in then extracting usable information. Performing various tests and analyses based on these reductions seems of little interest; indeed, it seems to me that they serve as a distraction.