Development and assessment of spectroscopic techniques and advanced analysis methods for characterization of bio-based fertilizers

This thesis investigates non-destructive spectroscopic techniques for rapid characterization of bio-based fertilizers (BBFs) and BBF-amended soils. BBFs contain valuable nutrients, but their efficient agricultural use requires fast, reliable quantification methods. The research evaluates Vis–NIR, MIR, and FTIR-PAS spectroscopy combined with chemometrics to predict total and plant-available nutrients across diverse BBFs. Results show that MIR generally outperforms NIR due to stronger and more resolved spectral features, while multi-sensor fusion and wavelength selection improve prediction robustness. FTIR-PAS demonstrates high accuracy for total nitrogen and phosphorus and shows promise for estimating plant-available phosphorus, although performance depends strongly on soil type and environmental stability. To address limited training data, physical and digital data augmentation approaches (CPDA and DSM) were developed, improving elemental predictions but remaining insufficient for plant available forms. Overall, spectroscopy proved effective for rapid nutrient assessment, supporting sustainable nutrient management. Remaining challenges include sensor sensitivity, sample preparation effects, limited dataset diversity, and the need for improved metrics and larger open-access spectral libraries worldwide.