Determination of nitrate concentration in water by multispectral UV analysis

This study presents the results of the development, calibration, and testing of an experimental prototype of a portable system for determining nitrate concentrations in water based on multispectral ultraviolet analysis combined with machine learning algorithms. The urgency of the problem lies in the widespread nitrate pollution of natural and drinking water caused by excessive use of nitrogen fertilizers, wastewater discharges, and natural leaching processes. High nitrate concentrations pose serious health threats, especially to infants, leading to methemoglobinemia and contributing to chronic diseases. Therefore, there is a critical need for autonomous solutions capable of providing rapid and accurate water quality analysis without complex laboratory procedures or chemical reagents. The proposed system uses four UV LEDs with wavelengths of 265, 308, 365, and 395 nm. Each channel serves a specific function: 308 nm detects nitrates, 265 nm monitors organic background, 365 nm tracks turbidity, and 395 nm is used for measurement stability control. The optical system is built on a photodiode in photoconductive mode with a transimpedance amplifier. An automatic delay algorithm was implemented to avoid measurements under unstable sample conditions based on signal fluctuations in the 395 nm channel. Differential processing of spectral data (A308−A265, A308/A265), combined with a Random Forest regression model, enabled accurate nitrate quantification up to 300 mg/L. The model achieved a high correlation (R² = 0.9806) and a mean absolute error of 0.77 mg/L. This approach effectively compensates for background interference and deviations from the Beer–Lambert–Bouguer law through adaptive mathematical modeling. The proposed system is suitable for implementation in mobile field instruments, environmental monitoring stations, agricultural infrastructure, and household water testing. Future work includes expanding the number of channels, integrating pH and conductivity sensors, and adapting the method for more complex multicomponent environments.