Near-infrared (NIR) spectroscopy is used to monitor online a large variety of processes. Hydrocarbons with their strong NIR spectral signature are good candidate analytes. For this work, the sorption data are measured in a manometric setup coupled with online NIR spectroscopy, to monitor the bulk composition. The assessment of time based results faces a baseline stability problem. The goal of this article is to study the robustness of different spectral preprocessing methods when dealing with time based data. In this study, it was found that for time based experiments it is necessary to perform drift correction on the spectra combined with a water band correction. For the calibration experiments, which only last few seconds, offset correction and drift correction performed equally well.