Date of Award

Spring 1998

Document Type


Degree Name

Doctor of Philosophy (PhD)


Mechanical & Aerospace Engineering


Mechanical Engineering

Committee Director

Surendra N. Tiwari

Committee Member

Robert L. Ash

Committee Member

William P. Chu

Committee Member

Gary E. Copeland

Committee Member

A. Sidney Roberts, Jr.


The understanding and acceptance of remotely sensed atmospheric data depends strongly on the steps taken to characterize experiment error and validate observations through comparisons to other independent measurements. A formal error analysis of the Stratospheric Aerosol and Gas Experiment II (SAGE II) water vapor operational inversion algorithm is performed and compared to previous results. Random measurement errors were characterized by segmented least-squares profile fitting of the slant path absorption which found the error to be uncorrelated in the stratosphere with estimated variances significantly smaller than expected from 18-30 km. Estimates of null space error were developed from radiosonde hygrometers in the troposphere and from SAGE II in the stratosphere. Systematic model bias errors are significant only in the troposphere where they reach 20% at the surface. Systematic errors associated with radiative transfer modeling are similar to previous analyses.

A comparative error budget study between the operational inversion algorithm and several other algorithms was conducted with formal error analysis and by examining the error characteristics of two years' data inverted with each algorithm. Four other algorithms were considered; onion peel, Mill-Drayson, Mill-Drayson with stratospheric profile smoothing, and a sparse grid non-linear least-squares fitting method. Stratospheric random errors were largest for the onion peel due to the lack of stratospheric profile smoothing while the Mill-Drayson with smoothing was identical to the operational. The Mill-Drayson algorithm exhibited random error reduction greater than expected from the form of the contribution function with stratospheric random errors approaching operational levels. The sparse grid contribution function was found to be relatively insensitive to grid point density and computationally intensive.

SAGE II upper tropospheric observations are compared to radiosonde climatologies and in situ radiosonde reports. SAGE II clear sky climatologies are shown to be half the level of the clear/cloudy sky radiosonde climatologies while correlative comparisons display nearly the same amount of bias. Much of the bias is attributed to the least sensitive hygrometers with SAGE II agreeing quite well with the most accurate and responsive hygrometer. Incorporating isentropic trajectories into the pair matching process greatly increases the number of correlative points but does not materially affect the comparisons.