Methods for the analysis of data from the Planetary Fourier Spectrometer on the Mars Express Mission

This work presents an algorithm for the scientific analysis of individual calibrated measurements from the Planetary Fourier spectrometer (PFS). The instrument, included in the scientific payload I of the ESA Mars Express mission to Mars, acquires spectra in the range between 250 and 8200 cm(-1), with a sampling step of similar to 11 cm(-1) and an effective resolution of similar to 2cm(-1). The observed radiance depends on several parameters of the atmosphere and surface of Mars as described by the radiative transfer equation. Adopting the very general formalism of Bayesian analysis, we determined which quantities are actually retrievable from individual measurements. Namely, they are: the surface temperature, the column density of dust and water ice aerosols in the atmosphere, the air temperature as a function of altitude (in the indicative range 5-45 km above the surface), the surface pressure, and the column density of water vapor and carbon monoxide. These evaluations are carried out taking into account the noise equivalent radiance (NER) of the instrument and the natural variabilities of the investigated parameters in the Martian environment, as estimated from the expectations of the European Martian Climate Dataset v3.1 (EMCD). Other parameters included in the radiative transfer equation shall be assumed as known, because they are not retrievable from individual measurements due to the instrumental NER or an underconstrained inverse problem: the surface emissivity in the thermal infrared, the optical properties of suspended dust and the analytical shape of dust concentration vs. altitude. During the development of the algorithm devoted to these studies, different approaches were evaluated on the basis of formal, computational and scientific considerations, with the aim to develop the general design of an integrated software package. The resulting code was extensively tested on a wide set of simulated PFS spectra. These spectra were computed from the atmospheric and surface conditions extracted from the EMCD, assumed to be representative of the Martian environment for different values of latitude, local time and season. Their comparison with the retrievals from simulated observations allowed us to evaluate the systematic and random errors affecting the procedures with respect to the different quantities involved. The code evaluates the surface temperature with an error in the order of I K, while the vertical air temperature profile is computed with an uncertainty less than 2 K from in the region between 5 and 20 km above the surface, increasing up to 7 K at 50 km. The column opacity of dust, measured in terms of integrated optical thickness at 1100 cm(-1), is computed with an error of around 0.13. The surface pressure determination is carried out with a typical uncertainty of 0.2-0.3 millibar. Several auxiliary tests allowed us to study the correlations between the different retrieval errors and the possible causes of incorrect PFS data interpretation. The choice of a suitable model for the dust optical properties is demonstrated to be particularly critical. This paper also presents the first discussion about application of the procedure to actual PFS Martian data. Despite the calibration issues still affecting the determination of absolute radiance in the near-infrared, the algorithm is able to achieve a satisfactory modeling of observations in a wide range of situations. (c) 2005 Elsevier Ltd. All rights reserved.