Sequential double EPO for compensation of temperature and turbidity interference in UV-Vis spectroscopic...

UV-Vis spectroscopy is widely used for the determination of nitrate nitrogen (NO₃-N) and chemical oxygen demand (COD) in water. However, the determination results are susceptible to interference from temperature and turbidity, leading to reduced detection accuracy for NO₃-N and COD. Existing studies have primarily focused on correcting individual interference factors, but research on the effects of temperature-turbidity coupled interference on the spectra of NO₃-N and COD mixed solutions, as well as corresponding compensation methods, remains relatively scarce. To address this issue, this paper proposes a Sequential Double EPO strategy, which achieves stepwise separation and correction of interference information by sequentially constructing subspaces for temperature and turbidity interference. First, this study evaluated the compensation capabilities of EPO for single-type temperature and turbidity interference under individual interference conditions. Subsequently, the impact of temperature-turbidity coupled interference on the spectra of mixed solutions was analyzed to determine the interference correction sequence under dual interference conditions. Second, based on this analysis, the Sequential Double EPO compensation strategy was proposed to sequentially separate and compensate for temperature and turbidity interferences. Finally, partial least squares regression (PLS) was used to establish quantitative prediction models for NO₃-N and COD. Experimental results indicate that the Sequential Double EPO strategy can compensate for temperature and turbidity interferences, effectively reducing spectral bias and improving spectral consistency. Compared with other methods, the PLS models established using SDEPO-compensated spectra achieved coefficients of determination (R²) of 0.9160 and 0.9441 for NO₃-N and COD prediction, respectively, with RPD values of 3.4866 and 4.2040, demonstrating superior predictive stability and model robustness.