CR 06:33-43 (1996)  -  DOI: https://doi.org/10.3354/cr006033

Statistical characteristics of U.S. Historical Climatology Network temperature distributions

Guttman NB

The U.S. Historical Climatology Network (HCN) adjusted data are designed to aid in the investigation of climate change. The primary goal in the development of the adjustment procedures was to produce temporally homogeneous at-site data, with respect to averages, that could be used in the investigation of regional, long-term trends. The data, however, have been (and likely will continue to be) used for applications other than the detection of climate trends, and some of these other uses involve statistical properties of the data distributions. This study examines the change in distributional characteristics that result from the site location and urbanization adjustments to maximum and minimum temperature data in the HCN. It is shown that at most locations distributional shapes are changed when the data are adjusted. For climate change studies the differences may not be important, but for other studies involving distributional characteristics they could be important. In particular, probabilistic assessments of unusual climatic conditions that are used for design and planning purposes are highly dependent upon the shape of the frequency distributions. Changing the skewness and kurtosis, i.e. the shape of a frequency distribution, often changes the probability densities in the tails of the distribution, and therefore impacts the probabilistic assessments and the subsequent decisions that are based on these assessments.


Climatic temperature data · Statistical characteristics · Probability assessments


Full text in pdf format