A Study of Satellite Emission Computed Tomography

We have performed a study to assess the capability of tomography to infer the temperature structure of the atmosphere when realistic instrument geometry and noise characteristics are taken into consideration. Fleming (1982) demonstrated a significant improvement using his tomographic retrieval technique in an idealized simulation experiment. The present study rectifies a number of shortcomings of Fleming's (1982) experiment, following Fleming's (1985) suggestions. In particular, geometric effects of the instantaneous field of view of the sensor are properly accounted for in our study. In addition, realistic instrumental noise is included in our simulations. Finally, we have used realistic atmospheric cross sections and realistic geometry and simulation codes appropriate for the HIRS2 sensor. Our atmospheric cross sections are based on National Meteorological Center analyses during the Global Weather Experiment. The simulation code we used is a linearization of the HIRS2 rapid algorithm developed by J. Susskind (Susskind et al., 1982) of NASA/GSFC. We have exclusively used Fleming's (1982) row action method of solution of the satellite tomography problem, but a number of modifications to improve convergence were necessary. The tomographic approach is superior to the single angle approach in the cases studied when observational noise (instrument noise plus scene noise) is large (1.5 brightness temperature degrees in each channel). For smaller noise levels (0.75 degrees) the two approaches are comparable in the cases studied.