I think I understand your point about the lousy performance on the Hurst coefficient that the historical records are too short. There seems to be much in the way of long-term processes not adequately included in the GCMs.

I see that most geologists have a long-term view. They seem to be most critical of much of what passes for climate science. In that regard I have many questions that don’t have good answers. Thanks for your blog.

No, I not accept climate predictions as reality; I’m a better statistician than that. I observe, interpret, evaluate and conclude.

Thanks for the paper reference. Have you read the actual paper? I have, and I don’t agree with your summary “climate models can’t predict anything”. The authors of this paper attempt to verify, or in their terms, falsify, IPCC model predictions of hydrological cycles at local scales. They selected recent historical records from several locations, then selected the four nearest model grid points, and compared the historical and reconstructed model records. An interesting approach. This is basic model verification and, contrary to what the authors state, is performed routinely with some of the GCMs. They appropriately cite and criticize some erroneous statements to the contrary by colleagues who (1) did not formulate the models, and (2) should know better.

But there are problems with this study. First, the reference to the Hurst exponent (HK in their paper) as an example of the limits of long-term predictability of climate is wrong. The Hurst exponent (H) indeed dictates limits to the long-term predictability of weather, but their historical records are simply too short to reliably calculate Hurst exponents. And no, Edward Lorenz did not discover this 40 years ago. Lorenz worked with Navier-Stokes descriptions of weather, in the process re-discovering what we now call today, mathematical chaos. (And I do know what I’m talking about; look here). Their arguments regarding H become even stranger when they explain that 0.5 mean time independent, whereas 1 means fully dependent. No. Since you seem to be somewhat versed in these matters, let’s agree that H=0.5 is derived from a proper stationary series (stable mean over time) with independent increments. H=1.0 is also derived from proper stationary series, but with increments that have long correlations; still time independent. So I’m not quite sure that these authors are clear on their mathematics.

Now, however, I have to say that I am not in complete disagreement with their overall conclusion, that the climate models are rather poor. But we know that, and that is why (1) we use many variations and many models, in order to assess among-model variation, and (2) we constantly try to improve the models. However, these models are designed to predict long-term climate, not short term local variability, which is what the authors focus on. If you take longer-term view of climate, such as that taken by geologists such as myself, climate is not the chaotic system envisaged by the authors. Their falsification tests are, in my opinion, invalid. Rather than living in a world where we attempt to guess at what next year’s climate is going to be, humans and human society have evolved and developed in a rather predictable climate regime. The fact that their own models cannot reflect this should be viewed as a serious shortcoming.

If you haven’t already, I suggest that you read my other blog postings, “Ignoramus et ignorabimus”, and “Just a “theory”?”

D. KOUTSOYIANNIS, A. EFSTRATIADIS, N. MAMASSIS & A. CHRISTOFIDES “On the credibility of climate predictions” Hydrological Sciences–Journal–des Sciences Hydrologiques, 53 (2008).
(http://www.atypon-link.com/IAHS/doi/abs/10.1623/hysj.53.4.671)

The Abstract: “Geographically distributed predictions of future climate, obtained through climate models, are widely used in hydrology and many other disciplines, typically without assessing their reliability. Here we compare the output of various models to temperature and precipitation observations from eight stations with long (over 100 years) records from around the globe. The results show that models perform poorly, even at a climatic (30-year) scale. Thus local model projections cannot be credible, whereas a common argument that models can perform better at larger spatial scales is unsupported.”

But I suppose Edward Lorenz found that out over 40 years ago.