Abstract
Near-infrared (NIR) interaction spectroscopy enables non-contact, non-destructive, subsurface sensing of layered, structurally evolving samples. Achieving robust calibration behaviour requires a better understanding of the underlying light–matter interactions governing the signal. In this work, we investigate how absorption, scattering, structural changes, and measurement geometry influence NIR interactance spectra, using water content assessment of dried salt-cured cod (clipfish) as a case study. Bulk optical properties of the fish muscle at different drying stages, previously obtained from double integrating sphere (DIS) measurements and inverse adding-doubling (IAD) analysis, were used as input for Monte Carlo simulations of the NIR interactance signal. Comparison with measured signals shows that the spectral variation during drying is primarily governed by the formation of a surface salt layer, rather than by changes in the optical properties of the muscle. This leads to distinct photon transport regimes, which explain limitations in chemometric model robustness and define requirements for robust sampling. These results demonstrate how a physics-based understanding of NIR interaction spectroscopy can improve signal interpretation, geometry selection, and calibration robustness in complex turbid media.