Reconstructing missing BOD₅ data from COD and its implications for water quality index assessment in a large...

Area of Science:

• Environmental Science
• Water Resource Management
• Data Science

Background:

• Long-term water quality monitoring in large irrigation systems is hindered by incomplete biochemical oxygen demand (BOD5) records due to costly lab analyses.
• Approximately 42.9% of BOD5 data was missing in a 20-year dataset from the Bac Hung Hai Irrigation System, Vietnam.

Purpose of the Study:

• To develop a chemical oxygen demand (COD)-based linear regression model to reconstruct missing BOD5 data.
• To evaluate the impact of reconstructed BOD5 data on the Water Quality Index (WQI) assessment.
• To provide a practical solution for water quality monitoring in data-limited environments.

Main Methods:

• A linear regression model was developed using a 20-year dataset (2004-2024) with 3014 observations.
• Model performance was assessed using train-test split and tenfold cross-validation.
• The reconstructed data's impact on WQI was compared to original data, including uncertainty analysis.

Main Results:

• A strong, stable linear relationship was found between BOD5 and COD (R² = 0.953).
• The model accurately reconstructed missing BOD5 values with low prediction errors (RMSE = 2.94 mg L⁻¹, MAE = 1.82 mg L⁻¹).
• Reconstructed WQI showed statistically insignificant practical differences, maintaining original classification outcomes.

Conclusions:

• A simple COD-based linear regression effectively reconstructs missing BOD5 data.
• This approach preserves data distribution and relationships, offering a reliable method for water quality assessment.
• The technique provides a valuable tool for enhancing long-term water quality monitoring in irrigation systems with data gaps.