Application of a stochastic, Weibull probability generator for replacing missing data on ambient concentrations of gaseous pollutants

A stochastic, three-parameter, Weibull frequency distribution, probability generator was tested by using theoretical data. Subsequently, it was applied to replace missing Values of hourly atmospheric concentrations of trace gases that were continuously monitored at three study sites, for 2 years. The results were highly accurate and realistic. The cumulative means and the medians calculated by the Weibull method were intermediate between corresponding values calculated by uniform substitution of missing values with 'zero' or with half of the minimum detection limit of the appropriate measurement instrument used. Furthermore, the Weibull method allowed the replacement of as many as 100 missing values on either side of a measured data sub-set, without altering the overall characteristics of the true frequency distribution of the entire data set. (C) 2000 Elsevier Science Ltd. All rights reserved.