Support vector machines in water quality management.

Area of Science:

• Environmental Science
• Data Science
• Water Resource Management

Background:

• Surface water quality monitoring is crucial for environmental protection.
• Optimizing monitoring programs can lead to cost and resource efficiencies.
• Machine learning offers advanced tools for analyzing complex environmental data.

Purpose of the Study:

• To optimize surface water quality monitoring using machine learning.
• To classify sampling sites (spatial) and months (temporal) for data reduction.
• To develop a regression model for predicting biochemical oxygen demand (BOD).

Main Methods:

• Support Vector Classification (SVC) for spatial and temporal grouping of monitoring sites and months.
• Support Vector Regression (SVR) for predicting biochemical oxygen demand (BOD).
• Model performance evaluated using misclassification rates, correlation coefficients, and root mean squared errors.

Main Results:

• SVC models successfully grouped 10 sites and 12 months into 3 clusters each with acceptable misclassification rates.
• SVR models demonstrated high predictive accuracy for BOD with correlations up to 0.952 and low RMSE.
• Nonlinear models (SVM, KDA, KPLS) outperformed linear methods (DA, PLS).

Conclusions:

• SVC and SVR models are effective for optimizing water quality monitoring programs.
• SVC facilitates significant data reduction (92.5%) for future monitoring.
• SVR provides a reliable tool for predicting water BOD using key variables.