Error propagation from line parameters to spectra simulations. Illustration on high temperature methane.

International audienceAstrophysical investigations generally need both complete and accurate spectroscopic databases. Despite continuous efforts in experimental and theoretical spectroscopic investigations, the lack of data in some spectral regions of interest is one of the main limitation of the presently available spectroscopic databases. Unfortunately information about missing data relevant to specic experimental conditions is rarely directly accessible from spectroscopic databases (focusing naturally on available data). Such information relies essentially on theoretical investigations which are equally limited to the present state of the art of modelling. The purpose of the talk is to show how multi-resolution models and rigorous error analyses can help overcoming such deciencies. Rather than ignoring the spectral regions for which high-resolution line by line modeling is presently inaccessible, low-resolution modeling can be exploited for the sake of exhaustivity which represents a major issue in many astrophysical applications. In parallel, a rigorous quantitative estimation of uncertainties, also essential for applications, represents the second key feature of this approach. The error propagation from predicted line positions and intensities onto simulated spectra will be described in detail based on multi-resolution global analyses using an effective Hamiltonian approach. The procedure makes extensive use of statistical numerical experiments. It is illustrated on methane which is present in the atmospheres of many astrophysical objects. The modeling of the methane absorption coecient in the near infrared and/or at high temperature remains a challenge mainly due to the intrinsic complexity of its rovibrational spectrum. The extremely large density of levels in highly excited polyads makes it very dicult (if not to say impossible) to envisage line by line modeling above say 10 000 cm-1 (http://hal.archives-ouvertes.fr/hal- 00277904/fr/). In fact, the present state of the art of high-resolution modeling is limited approximately to the spectral region below around 5 000 cm-1. The example of methane at high temperature (Tvib = 2000 K and Trot = 296 K) will be presented as a typical situation where transitions from several hot band systems with quite different accuracies have to be considered in the same spectral window. Typical confidence bands for the absorption coefficient and transmission signal are detailled at http://hal.archives-ouvertes.fr/hal-00473611/fr/. Prospects concerning possible improvements of the procedure and its computer implementation will be outlined