Level 0-1 Processing
This stage of the processing unpacks the level 0 or raw data, calibrates the data, removes instrument effects, develops source functions from the solar scan data, and registers the data with pressure and altitude. Registration is mostly done by using temperature versus pressure profiles to do a CO2 channel signal simulation and then comparing the simulation to measurement. The altitude range above 30 km is mainly used to avoid any serious effects due to aerosol contamination. The unpacking separates the data into files that correspond to various instrument modes. The level 0-1 processing therefore provides a set of corrected and calibrated signals that are ready for retrieval calculations.

Level 1-2 Processing
This step uses transmission profiles, different signal profiles from the gas filter channels, and solar source functions from the solar scans to retrieve temperature, pressure, and mixing ratios of HCl, HF, CH4, H2O, O3, NO, NO2, aerosol and temperature versus pressure. The retrieval method incorporates a simple "onion peel" procedure stabilized at the top and bottom of the profile with a scalar optimal estimation formulation [Connor and Rodgers, 1989]. The forward model for the gas filter channels (HF, HCl, CH4, NO) is a rigorous line-by-line code which is necessary for the effective high spectral resolution of these channels. All spectral dependence, including thermal and Doppler shift effects, is explicitly modeled. Along-path mixing ratio gradients are also included in the forward model for the diurnally active gases NO, NO2, and O3.

The radiometer channels are modeled using the emissivity growth and Curtis-Godson approximations using correction tables. These models have been validated against a line-by-line transmission code to better than 99% accuracy, and they are extremely fast, allowing a vector implementation of the optimal estimation equations. Again, full thermal and spectral dependence of the instrument is rigorously modeled, in this case, through a large set of transmission tables.

Most major interfering gases are retrieved as primary gases in other channels. Non retrieved interference is minor (such as N2O in HCl) contributing <1% error. However, interference from the Mount Pinatubo aerosol layer causes a major effect on the radiometer channel retrievals below the top of the aerosol layer. We have devised a correction approach which is based on retrieval using the gas filter channels, coupled with a Mie-scattering model to determine the aerosol extinction at the radiometer channel wavelengths. This approach works very well based on comparison with correlative measurements.

Level 2-3 Processing
The primary level 3 or "mapped" products include pressure versus longitude cross sections on selected days, pressure versus latitude for selected time periods, and polar orthographic projections. The latitude cross sections are generated on time scales of a couple of weeks or more and seasonally. The accuracy and precision of these products are high and therefore they provide very reliable means to study trends over time.