Development of high accuracy tangent height registration for Limb Scatter measurements

The most important source of measurement error in the Limb Scatter (LS) retrieval of trace gas species is related to Tangent Height (TH) registration. Several methods have been proposed to infer TH from radiance vertical profiles, among which the RSAS and the Knee techniques. The RSAS method (which locates the inflexion point of the radiance vertical profile for a weakly absorbing channel, typically 350nm) has been used to analyze SAGE III LS data and was found to be fairly accurate (offset less than 100m and RMS errors of about 350m). The Knee method (which locates the altitude of the maximum radiance in strongly aborbing ozone bands, typically 290-310nm) has been used with OSIRIS and SCIAMACHY data and have been shown to have similar potentials when information on high altitude ozone is available. The rationale of the paper is two-fold: (1) some systematic discrepancies have been identified between these two TH techniques, with mean offset of 0.5 km, and the cause of these discrepancies needs to be investigated (2) high TH accuracy needs to be demonstrated to ensure that future LS intruments (such as the Ozone Mapping and Profiler Suite (OMPS)for Ozone retrieval) are sufficiently accurate/precise to yield the high quality gas profile information needed for trend analyses. The paper will propose a new TH method which is presently being developed for SAGE III LS data. Basically, this new method, which considers radiance vertical profiles over a relatively large spectral range in the UV(from 300nm to 370nm), is a combination of three techniques: - Stretch and Shift - Extended RSAS - Knee The Stretch and Shift compares the measured radiance profiles with the ones generated by a forward model. At low wavelengths, this method reverts to the Knee technique. The extended RSAS identifies the location of the inflexion point at each wavelength. Over the relatively wide spectral range of interest, ozone absorption varies markedly, but its effect on each of the 3 techniques can be fitted (with an MLR method) and effectively removed. The ultimate product of the new TH registration algorithm is the effective TH shift (mean over the several techniques and all wavelengths), together with the error estimate. Examples will be given in the paper, and comparisons will be shown between SAGE III LS profiles retrieved using the new TH method with SAGE II products. It is estimated that the new method RMS errors are on the order of 150-200m, which is close to the lower limit based on NCEP atmospheric temperature uncertainties.