Development of statistical regression and artificial neural network models for estimating nitrogen, phosphorus, COD,...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Spectroscopy

Background:

• River water quality monitoring is essential for environmental management and policy development.
• In situ UV-Vis spectrometry offers potential for real-time water quality assessment but faces challenges in model development for complex conditions.
• Reliable methods are needed to link spectral data to specific water quality parameters, especially in eutrophic rivers.

Purpose of the Study:

• To develop and validate a robust framework for estimating multiple river water quality parameters using UV-Vis spectrometry.
• To establish reliable conversion models linking absorption spectra to nitrate-nitrogen (NO3-N), total nitrogen (TN), chemical oxygen demand (COD), total phosphorus (TP), and suspended solids (SS).
• To overcome the challenges of in situ measurements by integrating desktop and in situ UV-Vis spectrometers for model calibration.

Main Methods:

• Developed a framework integrating desktop and in situ UV-Vis spectrometers.
• Utilized desktop spectrometer data to create models for estimating NO3-N, TN, COD, TP, and SS.
• Employed Partial Least Squares Regression (PLSR), Principal Component Regression (PCR), and Artificial Neural Networks (ANN) for model development and validation.
• Validated models using in situ spectrometer data.

Main Results:

• Accurate estimation models were developed for NO3-N (PLSR), COD (PCR), TN (ANN), TP (PLSR/PCR), and SS (PLSR/PCR).
• Determination coefficients (R2) exceeded 0.6, and normalized root mean square errors (NRMSEs) were mostly within 0.4, indicating high model accuracy.
• The integrated approach demonstrated high efficiency in simultaneously monitoring multiple water quality parameters.
• The method avoids time-consuming and uncertain in situ measurements during model setup.

Conclusions:

• The developed framework provides an efficient and reliable method for simultaneous monitoring of key river water quality parameters.
• This approach enhances the potential for real-time water quality assessment and informed environmental policy.
• The study successfully addressed the challenges in linking spectral data to water quality parameters in dynamic river environments.