Applications of statistical models to synchronous climate variables: a case study of temperature and dew point for College Station, Texas

Abstract

Concurrent hourly temperature-dew point and temperature-dew point depression data were fitted to the bivariate normal distribution for the six months of January, March, May, July, September, and November for College Station, Texas. Calculations of the U[] and U[] statistics using a significance level of 0.05 were used to determine the acceptability of fit for each month-hour. Each variable was independently tested for normality before being applied to the bivariate normal distribution using criteria limits for the coefficients of skewness and kurtosis, and the Cornu criterion (again at the 0.05 significance level). Month-hour variables which did not possess characteristics of a normal distribution were transformed by use of power transformations. Generally speaking, temperature, dew point, and dew point depression values could be transformed to resemble a bivariate normal distribution. Of the 288 cases (6 mos. x 24 hrs. x 2 moisture variables) applied to the bivariate normal distribution, only three were considered acceptable at a significance level of 0.05. Because of these results, data was further divided up into wind direction classes and subjected to the same normality tests. Results from this procedure also proved negative. Using temperature and dew point data, heat index values were calculated from each observation for the months of June, July, August, and September. A method for approximating monthly cumulative probabilities of the heat index was successfully produced for College Station using a generalized form of the beta distribution and regression techniques.