This paper addresses problems arising from in situ measurement of gas content and temperature. Such measurements can be considered indirect. Transmittance or natural radiation of a gas is measured directly. The latter method (spectral radiation measurement) is often called spectral remote sensing. Its primary uses are in astronomy and in the measurement of atmospheric composition. In industrial processes, in situ spectroscopic measurements in the plant are often made with an open path Fourier Transform Infrared (FTIR) spectrometer. The main difficulty in this approach is related to the calibration process, which often cannot be carried out in the manner used in the laboratory. Spectral information can be obtained from open path spectroscopic measurements using mathematical modeling, and by solving the inverse problem. Determination of gas content based on spectral measurements requires comparison of the measured and modeled spectra. This paper proposes a method for the simultaneous use of multiple lines to determine the gas content. The integrated absorptions of many spectral lines permits calculation of the average band absorption. An inverse model based on neural networks is used to determine gas content based on mid-infrared spectra at variable temperatures.