Reservoir water quality simulation with data mining models

Area of Science:

• Environmental Science
• Water Resource Management
• Computational Science

Background:

• Water pollution poses significant challenges to effective water resource management.
• Data mining offers novel approaches for understanding complex patterns in reservoir water quality.
• Accurate modeling is crucial for predicting and mitigating water pollution events.

Purpose of the Study:

• To evaluate the efficacy of data mining techniques, including Genetic Programming (GP), Artificial Neural Network (ANN), and Support Vector Machine (SVM), for reservoir water quality modeling.
• To compare the performance of these data mining models against the numerical water quality simulation model CE-QUAL-W2.
• To assess the accuracy and efficiency of different models in simulating nitrate concentration in reservoirs.

Main Methods:

• Employed Genetic Programming (GP), Artificial Neural Network (ANN), and Support Vector Machine (SVM) for reservoir quality modeling.
• Utilized data generated from the CE-QUAL-W2 numerical water quality simulation model as input for the data mining models.
• Applied statistical goodness-of-fit criteria, including Mean Absolute Error (MAE), Nash-Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), and coefficient of determination (R²), for model evaluation.

Main Results:

• The Support Vector Machine (SVM) model demonstrated superior performance in predicting nitrate pollution compared to GP, ANN, and CE-QUAL-W2.
• SVM significantly reduced simulation runtime, achieving time savings of 581s, 276s, and 146s over CE-QUAL-W2, GP, and ANN, respectively.
• Excellent goodness-of-fit metrics were achieved with SVM, including R² = 0.97 and NSE = 0.92, with the lowest error indicators (MAE = 0.034, RMSE = 0.007).

Conclusions:

• Data mining tools, particularly SVM, offer a time-efficient and highly accurate method for simulating solute concentrations in reservoirs.
• The study validates the potential of applying advanced data mining techniques for enhanced water quality management and pollution prediction.
• SVM presents a promising alternative to traditional simulation models for practical reservoir water quality assessment.