Last word on climate skeptics
I guess I better post my analysis of the M&M critique of modern paleoclimatology, started here, continued here, here and here, before it’s completely irrelevant. The replies have started to get published, as RealClimate and the WSJ report. The M&M GRL paper, available here, is their first and only paper to appear in a science journal. It spends about half its space on their phantom hockey stick theory, and the other half introduces a Monte Carlo analysis for reassessing the error bars for the reconstructed temperature series. I also skimmed their latest contribution to Energy and Environment, but didn’t see any new issues. This will be my last post on climate skeptics, I promise.
They repeat their flawed analysis of PC1 reconstruction. As seen before, it’s not that anything that they say is so wrong, as you can indeed get a different PC1 by changing the definition, it’s just irrelevant to the reconstructed climate. Unobservables, would say a physicist. Temporary local variables, would say a computer programmer. No matter what you call it, it shouldn’t change the answer, and in fact doesn’t.
So what about the data analysis, Monte Carlo and the error bars? I could go through this in detail and see whether it makes sense, but to be honest, at this point I lost interest. M&M have dropped their reconstruction with warmer values, they no longer assert that the hockey sticks in temperature are created by coding errors in the procedure, and essentially accepted the MBH values as an unbiased estimate of past climate, and are now quibbling with the error estimate. We’re now at a point where everyone agrees on the statement, “The 1990’s are the warmest decade in the past X hundred years, and maybe even longer”, but for different values of X. MBH98 says X=6, and extends it to 10 in a later paper, while M&M, with their revised error bars, don’t want to accept a value larger than 4. The thing is, there’s no policy implications to X, no question of the validity of anthropogenic impact, only mere curiousity about past climate from a particular dataset. I’m not even convinced that all such disputes can be resolved in principle: an error bar is, to some extent, a sociological construct, one way of communicating aspects of the uncertainties of evidence, but certainly not the only way. Different people could validly emphasize different aspects of uncertainty, or demand different levels of evidence before accepting a claim. A better assessment of the uncertainties of knowledge comes from how later studies, using different methods, stack up against prior studies. In this case, we have one: Moberg 2005 (RealClimate discussion | McIntyre’s perspective), using borehole data analyzed by wavelets, produces a curve which shows much higher variability in climate than previous studies including MBH98. Higher variability, but do they agree within uncertainties? Yes, they do. And are the 1990’s still warmest? Oh yeah. New data is better than any error estimate.
I see this whole thing as an example of the robustness of science. Even if you begin with people with radically different preconceptions, through contact with reality and the data, they will end up with compatible pictures of reality. I’m not claiming that M&M have completely converged with the mainstream, because they would surely disagree with that, only that the process is well underway. Science is the part that doesn’t change when you change the people.