Applications of AI in River Protection

By : Dr. Shah Christirani Azhar

Rivers are essential for providing clean drinking water, driving food production, powering industries, and sustaining entire ecosystems. However, with the growth of urban development, agriculture, and industrial activities, many rivers are becoming increasingly polluted. This growing pollution situation highlights the critical need for rigorous and continuous water quality monitoring. While conventional methods, such as manual sampling and laboratory testing, can provide accurate results, they are often time-consuming and labour-intensive, and can sometimes lead to human error.

Artificial Intelligence (AI), particularly Artificial Neural Networks (ANNs), has transformed the way river water quality is monitored. ANNs can handle vast amounts of data from sensors placed along rivers, tracking key indicators such as pH, turbidity, dissolved oxygen (DO), biochemical oxygen demand (BOD), chemical oxygen demand (COD), suspended solids (SS), and ammonia nitrogen (NH₃-N). Real-time data analysis with AI enables rapid detection of pollution and accurate prediction of water quality changes with remarkable accuracy.

ANN offers one of its main benefits through its ability to understand complex relationships that conventional statistical methods often overlook. This ability makes ANNs particularly powerful for detecting early signs of pollution and predicting changes in water quality. Like a person who learns from experience, ANNs become more intelligent and accurate in their predictions as they receive more data over time. This approach helps environmental authorities to respond quickly and effectively when potential pollution is detected.

The multilayer perceptron of the ANN

 

An example of implementing ANN applications in water quality prediction is a study conducted in the Muda River Basin, Malaysia. Researchers created an ANN model to classify river water quality based on the National Water Quality Standards (NWQS), which range from Class I (extremely clean) to Class V (heavily polluted). Unlike the traditional Water Quality Index (WQI), which requires complex calculations using various sub-indices, the ANN model was able to classify water quality directly from raw parameter data, making the process faster and more efficient.

Location of the Muda River basin and the distribution of the monitoring station.

 

The ANN model with the Softmax Activation Function was optimised using six key water quality variables. It achieved an impressive 96.8% accuracy during training and performed even better during testing and validation. It was effective in classifying Classes II and III, which represent moderately clean water that is suitable for treatment and supporting aquatic life. Even in its simplified form, using only dissolved oxygen (DO), biochemical oxygen demand (BOD), and chemical oxygen demand (COD) as inputs, the model maintained consistently high accuracy. These results show that the ANN models are highly adaptable in regions with limited monitoring tools or a lack of sensors. This flexibility makes AI an asset for managing water quality in resource-constrained areas.

                                     Measured and predicted river classification.

 

Overall, using Artificial Intelligence (AI) such as Artificial Neural Networks (ANN) in water quality monitoring optimises the water quality classification process, reduces manual workload, and enables real-time assessment. By transforming continuous environmental data into actionable insights, AI acts as a proactive guardian of our rivers. In conclusion, AI is not just a futuristic concept but a practical and essential tool in environmental protection. As challenges in water management become more complex, the use of intelligent technologies such as ANN ensures faster response times, improved water safety, and better preservation of natural resources for future generations.

References

Azhar, S.C., Aris, A.Z., Yusoff, M.K., and Ramli, M.F. 2019. Water Quality Class in a River Basin using an Artificial Neural Network with the Softmax Activation Function. ARPN Journal of Engineering and Applied Sciences 14(23):8585-8593.

Anctil, F., Filion, M., and Tournebize, J. 2009. A neural network experiment on the simulation of daily nitrate-nitrogen and suspended sediment fluxes from a small agricultural catchment. Ecological Modelling 220(6): 879–887.

 

ASCE Task Committee on Application of Artificial Neural Networks in Hydrology. 2000. Artificial neural networks in hydrology, II: Hydrologic applications. Journal of Hydrologic Engineering. 5(2): 124–137.

 

Dawson, C.W. and Wilby, R.L. 2001.Hydrological modelling using artificial neural networks.Progress in Physical Geography 25(1): 80–108.

 

Gazzaz, N.M. 2012. Artificial Neural Network for Water Quality Assessment and Land Use Pattern Recognition in Kinta River Catchment. PhD Thesis, Universiti Putra Malaysia, Serdang.

 

 

 

Tarikh Input: 12/09/2025 | Kemaskini: 12/09/2025 | hasniah