Mysteries in Climate Statistics

In this talk, I discuss some apparent mysteries in the analysis of climate statistics. The first of these is the origin of the problem where, for example: 1) a paper is published claiming to find period "X" in a given data set with a reasonable basis for the claim; 2) several more papers appear stating that they looked for period "X" in similar data, or earlier or later data from the same location, and either found no evidence, or found it with a 180-degree phase---. The second problem is that of "long memory" processes: sea level, Nile river flow, and Northern Hemisphere temperature have all been used as examples in the statistics and climate literature. A third problem is the "perverse" statistics of data such as daily temperature measurements. As an example, the spectrum of such a series often is more variable than standard theory predicts. The common features that explain these appear to be overly simplified statistical models, nonstationarity, and solar forcing.