Abstract
Estimates of the natural variability of monthly mean temperature data from 107 U.S. stations are made. The natural variability of monthly means is defined as those interannual fluctuations that can be attributed to the effects of statistical sampling alone. It is variability resulting from the variance and autocorrelation associated with daily weather fluctuations. It does not indicate “climate change” but rather it is the variability within an “unchanging climate”; as such it is a measure of unpredictable “climatic noise”. Comparisons between the natural and actual interannual variability are discussed in the context of potential long-range predictability. The natural variability is proposed as a lower limit for the standard error of estimate for any long-range prediction. A characteristic time between independent estimates is computed.
