Multi-resolution error analysis of predicted absorption coefficients. Method and computer implementation. Application to the IR spectrum of methane at high temperature.

A general method for the estimation of the confidence interval of molecular absorption coefficients is presented. It combines statistical numerical experiments with analytical calculations. The method uses line parameter predictions (position and intensity) with known estimated uncertainties. The propagation of errors from line parameters to absorption coefficients or cross-sections are analyzed in detail as a function of the resolution. This work is especially intended to provide expert information for applications requiring theoretical predictions for which the present state of the art of line by line high-resolution modeling is not accurate enough to meet the experimental precision. As an illustration, it is applied to predict the absorption coefficient of methane at high temperature involving simultaneously well known lower energy levels (cold bands) and higher energy levels with much lower precisions (hot bands). Potential extensions to applications for atmospheric remote sensing of astrophysical objects are discussed