Spectral Reconstruction in the Background Subspace for COD Quantification under Multidimensional Interference

Over the past several decades, absorption spectroscopy has been widely used to detect aquatic constituents. Chemical oxygen demand (COD) is an important indicator of water quality, reflecting the extent of organic pollution. However, under natural water conditions, multidimensional interference including air bubbles, turbidity, and fluid motion introduce nonabsorptive spectral artifacts, causing nonlinear spectra distortions and reducing the accuracy of COD prediction. In this study, we systematically investigated the effects of bubbles, turbidity, and flow conditions on spectral morphology and intensity distribution. And we propose a background subspace–based spectral reconstruction method using Principal Component Analysis (PCA). We first construct a low-dimensional spectral subspace from COD-free samples to characterize the background of the water matrix. Spectra containing COD are then projected onto this subspace, and COD-related absorption features are extracted from the residuals between the original and reconstructed spectra. As a result, absorbance spectra containing only COD-related information are achieved without requiring explicit quantification of bubbles or flow velocity. Our method effectively eliminates spectral artifacts and improves spectral consistency and signal-to-noise ratio. Unlike previous studies focusing on turbidity alone, this approach accounts for multiple interferences, including bubbles and flow, enabling reliable in situ optical monitoring in aquatic environments and regulatory contexts.