The global remote sensing of tropospheric ozone profiles is a critical environmental measurement to be performed by future satellite experiments. We have applied the method of nonlinear least squares in conjunction with an efficient and accurate line-by-line radiative transfer model to directly retrieve vertical profiles of tropospheric ozone from simulated dear sky, nadir-viewing radiances covering the entire 9.6-mu m ozone band. The simulations have been generated for the specifications of the tropospheric emission spectrometer (TES), a Fourier transform spectrometer with 0.032 cm(-1) resolution (half width at half maximum (HWHM), unapodized) being developed for NASA's Earth Observing System. Profile retrieval errors for background tropospheric ozone levels are characterized as a function of measurement noise, spectral resolution, and vertical resolution based on a linear error analysis and an initial guess profile with minimal constraint, hence negligible potential profile bias at all altitudes. The main conclusions of the study are that (1) for the TES experiment design, ozone profiles are retrievable to approximately +/-5% (1 sigma) for a vertical resolution of 5 km in the middle and upper troposphere, (2) the stratospheric portion of the profile must be retrieved directly from the measured nadir spectrum, (3) for equal measurement times and considering the effects of both systematic error and source radiance noise, an optimal spectral interferometer resolution exists that is close to the TES resolution (0.032 cm(-1); HWHM unapodized), and (4) ozone boundary layer retrievals are highly dependent on the contrast between that layer and the surface. The results from an ozone retrieval utilizing simulated radiances from an atmosphere defined by a radiosonde observation at Ascension Island are considered in the context of the solution of the nonlinear problem and the linear error analysis, The ozone retrieval analyses presented in this paper are principally concerned with the effects of measurement error and thus represent the optimal retrieval capability for the assumed design. However, the availability of the full high-resolution spectrum will enable the detection and mitigation of the systematic errors.
