A NUMERICAL-METHOD FOR SYNTHESIZING ATMOSPHERIC-TEMPERATURE AND HUMIDITY PROFILES

A numerical technique is described for synthesizing realistic atmosphe ric temperature and humidity profiles. The method uses an ensemble of radiosonde measurements collected at a site of interest. Erroneous pro files are removed by comparing their likelihood with prevailing meteor ological conditions. The remaining profiles are decomposed using the m ethod of empirical orthogonal functions. The corresponding eigenprofil es and the statistics of the expansion coefficients are used to numeri cally generate synthetic profiles that obey the same statistics (i.e., have the same mean, variability, and vertical correlation) as the ini tial dataset. The technique was applied to a set of approximately 1000 temperature and humidity soundings made in Denver, Colorado, during t he winter months of 1991-95. This dataset was divided into four cloud classification categories and daytime and nighttime launches to better characterize typical profiles for the eight cases considered. It was found that 97% of the variance in the soundings could be accounted for by using only five eigenprofiles in the reconstructions. Ensembles of numerically generated profiles can be used to lest the accuracy of va rious retrieval algorithms under controlled conditions not usually ava ilable in practice.