A quantitative prediction strategy for UV-vis spectroscopy of nitrate in water based on a difference spectrum-hybrid prediction model.

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 a hybrid prediction model to address this issue. By analyzing the linear relationship between the 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^-1 for standard samples, and an R² of 0.9663 and an RMSE of 0.7835 mg L^-1 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.