Modeling water quality in an urban river using hydrological factors--data driven approaches

Area of Science:

• Environmental Science
• Hydrology
• Artificial Intelligence

Background:

• Taiwan's rivers experience seasonal flow and quality variations due to extreme weather events.
• Sudden river flow changes can degrade water quality and harm ecosystems.
• Current water quality monitoring is monthly/quarterly, hindering timely pollution management.

Purpose of the Study:

• To develop a systematic analysis scheme (SAS) for assessing spatio-temporal water quality interrelations in urban rivers.
• To construct water quality estimation models using artificial neural networks (ANNs) and the Gamma test (GT).
• To enable daily-scale water quality estimation for effective river pollution management.

Main Methods:

• Utilized hydrological, water quality, and economic data from the Dahan River basin, Taiwan.
• Employed the Gamma test (GT) to identify key hydrological factors influencing ammonia nitrogen (NH3-N) concentration.
• Developed ANNs, including the nonlinear autoregressive with exogenous input (NARX) network, for NH3-N estimation.

Main Results:

• Identified four key hydrological inputs: discharge, days without discharge, water temperature, and rainfall.
• The NARX network achieved high accuracy in estimating NH3-N concentration (coefficient of efficiency: 0.926, RMSE: 0.386 mg/l).
• The model effectively captured peak NH3-N values during dry periods, crucial for pollution control.

Conclusions:

• The proposed SAS reliably models spatio-temporal NH3-N concentration using only hydrological data.
• This approach enables daily-scale estimations, significantly faster than traditional methods.
• The SAS provides a more efficient and timely tool for river managers to address pollution events.