Modelling water quality parameters using model tree, random forest, and non-linear regression for Mula-Mutha River,...

Area of Science:

• Environmental Science
• Water Resource Management
• Data Science

Background:

• Accurate assessment of vital water quality indicators like Biological Oxygen Demand (BOD) and Chemical Oxygen Demand (COD) is crucial for environmental health, human well-being, and agricultural productivity.
• Data-driven techniques (DDT) provide automated, reliable, and rapid solutions for water quality evaluation.

Purpose of the Study:

• To employ and compare various DDTs, including Random Forest (RF), Model Tree (MT), and Non-Linear Regression (NLR), for predicting BOD and COD levels.
• To develop separate BOD-COD prediction models for the three distinct stretches of the Mula-Mutha River in Pune, India.

Main Methods:

• Utilized RF, MT, and NLR techniques to build predictive models for BOD and COD.
• Performed data analysis using violin diagrams to understand data characteristics.
• Evaluated model performance using error metrics (R, MAE, RMSE) and visual tools (Taylor diagram, scatter plot, hydrograph).

Main Results:

• MT and RF models demonstrated a strong correlation between actual and predicted BOD and COD values.
• NLR models also showed good performance, closely following MT and RF.
• The interpretability of RF (tree-based), MT (equation sequences), and NLR (single equation) enhances practical adoption.

Conclusions:

• DDTs, particularly RF and MT, are effective tools for accurate and efficient water quality assessment of river systems.
• The developed models provide valuable insights for water quality professionals and future research in environmental monitoring.
• This study highlights the potential of automated data-driven approaches for managing and protecting river water resources.