Quantitative Analysis of Eutrophication Marker Molecules in Atmospheric Air Based on Model IR Spectra and the SIMCA Method

Abstract

Monitoring the state of water bodies requires reliable detection of gaseous products of anaerobic decomposition. This work evaluates the SIMCA (Soft Independent Modeling of Class Analogy) algorithm for the qualitative and quantitative analysis of infrared (IR) absorption spectra of multicomponent gas mixtures containing eutrophication marker molecules (CO2, H2S, CH4, N2O, PH3) in near-surface atmospheric air. Model absorption spectra were computed via the HITRAN2020 database over the 1000–10000 cm-1 range for temperatures of 236–316 K and relative humidity of 60–80%, conditions typical of the Siberian region. The predictive model combines Savitzky–Golay preprocessing, principal component analysis (PCA), SIMCA classification with Mahalanobis distance, and k-fold cross-validation, together with an original visualization of classes in two-dimensional PCA cross-sections. Within the most informative interval, 2510–2960 cm-1, the model determines the concentrations of H2S, CH4, and N2O with an average accuracy of about 90% (88.3%, 89.2%, and 92.4%, respectively) at a noise level on the order of 10-6 cm-1, attainable by photoacoustic spectroscopy. Accuracy decreases by roughly 7% as temperature and humidity rise, so measurements are preferable near 276 K. The model remains robust with as few as 40 spectra per class, and the PCA cross-sections form a universal "concentration space," yielding an interpretable tool for practical assessment of water-body conditions.
The work was performed according to the Government research assignment for TSU, project FSWM-2025-0038.

Speaker

Tretyakov Akim Konstantinovich
Tomsk State University
Russia

Discussion

Ask question