What question did this study set out to answer?

The central aim is to understand how conventional water quality parameters relate to pollution levels in the Luoqing River.

February 13, 2026Open Access

Machine learning-based correlation analysis of conventional water quality parameters and composite pollution index in the Luoqing river of the South China Sea Coastal Zone

Key Points

The central aim is to understand how conventional water quality parameters relate to pollution levels in the Luoqing River.
Analyzed water quality data from three monitoring sections along the Luoqing River.
Evaluated seven conventional parameters using correlation analysis and machine learning models.
Constructed a Composite Pollution Index (CPI) to quantify pollution.
Monthly variability was high, but interannual patterns remained stable.
Downstream sections had higher pollution levels with complex parameter interactions.
Machine learning models predicted CPI with high accuracy (R² ≈ 0.994).

Abstract

Rivers serve as critical conduits for transporting pollutants and nutrients to coastal zones, directly influencing coastal ecosystem functions and marine environmental health. Understanding spatiotemporal variations and driving mechanisms of river water quality is therefore critical for effective watershed management and coastal zone protection. This study analyzed water quality data from three monitoring sections (Suqiao, Xialiang Village, and Longxi) along the Luoqing River in Guilin, China, collected during January 2023–March 2025. Seven conventional parameters—water temperature, pH, dissolved oxygen (DO), permanganate index (COD Mn ), total nitrogen (TN), total phosphorus (TP), and ammonia nitrogen (NH 3 -N)—were evaluated using Spearman correlation analysis, principal component analysis (PCA), and machine learning models (CatBoost, Random Forest, and XGBoost). The results showed pronounced monthly variability but relatively stable interannual patterns, indicating dominant control by seasonal hydrological and biogeochemical processes. Spatial differences were evident, with downstream sections exhibiting higher pollution levels and more complex parameter interactions, reflecting intensified anthropogenic influences. PCA extracted two principal components explaining over 70% of the total variance: PC1 associated with nutrient and organic pollution (TN, TP, NH 3 -N, COD Mn ) and PC2 representing physicochemical conditions (pH and DO). A Composite Pollution Index (CPI) was constructed based on NH 3 -N, TN, TP and COD Mn to characterize overall pollution levels. Machine learning models achieved high predictive performance for the CPI (R 2 ≈ 0.994). The shapley Additive Explanations (SHAP) and partial dependence analyses identified TN as the primary controlling factor, with synergistic enhancement between TN and COD Mn under high nitrogen conditions. Overall, Luoqing River water quality is characterized by a nitrogen-centered pollution structure, providing a scientific basis for targeted watershed management and coastal ecosystem protection.

Demander à l'IA

Bookmark

View Full Paper