How quantitative can quantitative spectroscopy be?

Remote sensing or the Earth´s atmosphere requires an accurate spectroscopic database such as HITRAN. Especially the new satellite instruments like MIPAS and SCIAMACHY on Envisat, TES and MLS on AURA, IASI on METOP and the instruments for CO2 measurements on OCO and GOSAT require accuracies of line intensities in the percent range, the latter two even in the sub-percent range. Whereas the accuracy of line positions is often sufficient, that of linestrength, broadening and shifts is often not. In case of larger molecules where absorption cross sections are used the accuracy is often not sufficient, too. The reason are numerous error sources associated with intensity measurements which are sometimes hard to quantify. In case of high resolution Fourier-transform spectrometers, which are most commonly used for measurements of the spectrocopic database, problems are related to detector non-linearities, channeling, thermal radiation, phase errors, thermal drifts, microphonics, multipassing, and instrumental lineshape error. The knowledge of sample column amount, the stability of the sample over the measurement time, and the temperature knowledge and homogeneity are potential error sources. Furthermore, the line model and line parameter retrieval can cause errors. In order to provide a spectroscopic database suitable for remote sensing adequate sample cells, including multi-pass cells for sufficient absorption path, with sufficient temperature range and homogeneity as well as alignment stability are mandatory. Furthermore, procedures and infrastructure for generating defined gas mixtures as well as flow reactors for unstable species are needed. A sound measurement strategy including sufficient measurements with redundancy is necessary. Quality assessment procedures can provide an important tool for identifying hidden error sources and validating accuracy. In only very few cases reported uncertainties for database entries include a complete error analysis with quality assessment. Often measurements reported from different laboratories show differences outside stated uncertainties. As an example for the generation of a spectrocopic database with well defined uncertainties new FT-measurements of ν2 water line parameters carried out with the Bruker IFS 120HR at DLR will be presented