Nitrate nitrogen monitoring method for lake and reservoir waters based on hyperspectral features and its application

Aiming at the problems of poor timeliness, high cost and easy secondary pollution of the traditional detection method of nitrate nitrogen (NO₃⁻-N) in the eutrophication monitoring of lakes and reservoirs in the Yangtze River Basin, the present study proposes a new method of in situ monitoring based on hyperspectral features at the minute level.Taking the bay of Xiangxi River, a typical tributary of Three Gorges Reservoir, as the study area, a buoy hyperspectral spectrometer (200-1000 nm) was used to simultaneously acquire the absorption spectra of the surface water and the measured NO₃⁻-N concentration data in the laboratory.After Savitzky-Golay filtering and first-order differential preprocessing, it was found that there was a characteristic absorption peak of NO₃⁻-N at 210-220 nm, and its first-order differential spectrum was significantly correlated with the concentration in the 200-250 nm band (positive correlation in the 200-217 nm, negative correlation in the 217-250 nm, and negative correlation in the peak at 225 nm with r≈0.82).By systematically comparing the partial least squares (PLS) regression models constructed at discrete wavelengths, in the 200-250 nm sensitive band and in the full band (200-900 nm), it was determined that the model inversion accuracy in the 200-250 nm band was optimal: correlation coefficient of the test set was r=0.94, RMSE=0.08 mg/L, and MAE=0.07 mg/L, which is 18% higher than that of the full band model.18%.This band can effectively capture the NO₃⁻-N UV absorption feature and partially strip out the interference of bromide ions and nitrite.Cross-regional validation showed that the model was applicable to another typical tributary of the Three Gorges Reservoir, the Shennong Creek (r=0.74, RMSE=0.2 mg/L), and was directly applied to the Wensi Lake and Meiziya Reservoirs in Yichang City, with an accuracy of RMSE=0.09 mg/L, which confirms its good generalisation ability.Relying on the high-frequency collection advantage of the buoy system (10 times/minute), this method achieves the minute-level continuous dynamic monitoring of NO₃⁻-N in the Three Gorges Reservoir area, breaks through the bottleneck of the time limit of the traditional method, and provides a reliable technical means for the on-line early warning and fine management of the water quality of lakes and reservoirs in the Yangtze River Basin.