Application of Stepwise Multiple Regression Techniques to Inversion of Nimbus “IRIS” Observations

Abstract

Stepwise multiple regression analyses are used to explore the statistical (linear regression) relationships between satellite-observed earth?atmosphere emission spectra and meteorological parameters. The stepwise regression technique permits screening of a large number of potentially useful spectral observations (predictors) to isolate those few that contribute most to the explanation of the variance of a particular meteorological parameter. Such a technique is particularly useful when applied to complete spectra of the type obtained by the IRIS (infrared interferometer spectrometer) instrument on the recent Nimbus meteorological satellites. Emphasis is placed upon inferences of key meteorological parameters not usually obtained from routine inversion of satellite spectral observations. The technique is applied to a sample of Nimbus 3 IRIS spectra. The results indicate that information on atmospheric temperatures, geopotential heights of pressure surfaces, tropopause pressure, and tropopause temperature can be obtained directly from the satellite observations with the use of simple linear relationships having only a few terms each. Based upon the results of this exploratory study, suggestions are made for further development and exploitation of the stepwise regression analysis technique and its application to the problem of inferring meteorological parameters from earth?atmosphere emission spectra of the IRIS type.