Abstract
The distributions of weekly averages of diurnal temperature maxima, minima, means, and ranges are found to be non-normal, although the errors in using the normal distribution do not impair the usefulness of derived estimates of probability. The central tendency and the variance are estimated by harmonic regression. This enables the estimation of probabilities for any week from as few as five parameters. A three-term harmonic curve fitted to individual years is adequate to describe the course of temperature.

This publication has 0 references indexed in Scilit: