The Analysis of Upper-Level Humidity

Abstract

The incorporation of diagnostic data obtained from surface observations into a humidity analysis at the 850-, 700- and 500-mb levels is tested using European surface and upper-air data for 22 observation times in February 1962. Sparse data conditions are simulated by withholding a portion of both surface and upper-air data. The rms errors and contingency table per cent correct scores indicate that an improved analysis is obtained in data sparse areas by weighting the diagnostic data relative to the radiosonde data. The most appropriate weighting is a function of the reliability of the diagnosis and the data density. The effects of using different initial guess fields, varying the application of the successive approximation technique (SAT) to these fields, and applying smoothing were also considered.