Quantitative prediction strategy of UV-Vis spectroscopy of nitrate in water based on difference spectrum-hybrid prediction model

Abstract

Due to spectral influence caused by turbidity, the accuracy of nitrate quantification using UV–Vis spectroscopy remains challenging. This study proposes an integrated method combining UV-Vis spectroscopy, difference spectrum analysis, and hybrid prediction model to address this issue. By analyzing the linear relationship between difference spectrum and turbidity, a novel turbidity compensation strategy — the Mixed Difference Nitrate Method (MDNM)—was developed.Subsequently, a hybrid prediction framework integrating linear regression and threshold-based waveband selection was employed to enhance modeling accuracy. Experimental results on both standard and natural water samples demonstrate that the method achieves high accuracy and generalization ability, with an R² of 0.9982 and an RMSE of 0.2629 mg/L for standard samples, and an R² of 0.9663 and an RMSE of 0.7835 mg/L for natural water samples. The proposed method offers a simple, effective, and low-cost strategy for nitrate detection in turbid water, with significant potential for practical environmental monitoring and water quality assessment.