“…Although the ratio of the standard deviation (observed-estimated) to the observed mean (a'oe/pobs in Table 10.1) is not a true statistic, in that it cannot be related by, for example, a confidence level to a random distribution of uncorrelated pairs of values, it is a ratio commonly used to evaluate the success of models. Marsh and Withers (1969) indicate that, for their Reading, England, model as well as for others in the literature, the ratios 6oe/bobs are all higher than 1.1 for 6-and 24-hr averages during the heating season. From Table 10 because the former considers a band of acceptable error, whereas the latter is based on actual calculated and observed values (e.g., if Xobs = 0.03 ppm and Xest = 0.01 ppm, the skill score considers this pair a "success," whereas, for the purposes of calculating R2, this pair represents an error of 0.02 ppm).…”