Spatiotemporal variation and influencing factors of phosphorus in Asia’s longest river based on receptor model and machine learning
Phosphorus contamination in rivers has received widespread attention. However, in areas with extensive sources of phosphorus and complex hydrogeology conditions, it is difficult to accurately evaluate the main influencing factors of phosphorus. In present study, the spatiotemporal variations of phos...
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-02-01
|
| Series: | Ecological Indicators |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1470160X25001463 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Phosphorus contamination in rivers has received widespread attention. However, in areas with extensive sources of phosphorus and complex hydrogeology conditions, it is difficult to accurately evaluate the main influencing factors of phosphorus. In present study, the spatiotemporal variations of phosphorus were analyzed at wet and dry seasons in the Yangtze River during 2020–2023. Phosphorus concentrations in the Yangtze River decreased by 9.15 % in four years, reaching a peak in summer. In addition, absolute principal component score‐multiple linear regression (APCS‐MLR) model was proved to be suitable for exploring the contribution of main phosphorus-influencing factors in Yangtze River. The contribution rate in wet season was ranked as point source pollution (40.13 %) > agricultural pollution (32.74 %) > organic pollutants (3.78 %), while the contribution rate in dry season was ranked as point source pollution (44.88 %) > organic pollutants (13.13 %) > seasonal factor (7.60 %). Machine learning models (e.g., RidgeCV, Random Forest, XGBoost) were used to establish a connection between total phosphorus concentrations and explanatory variables defining influencing factors, aiming to predict total phosphorus concentrations in the Yangtze River. The anthropogenic and natural variables, such as domestic sewage, GDP, agricultural area, livestock, rainfall, wind speed, temperature and population were selected as predictors. The Random Forest model performed well in predicting total phosphorus concentrations, with R2 value of 0.76. This study provides useful information for optimizing phosphorus pollution management and strategies for eutrophication control in the Yangtze River as well as in other large watersheds. |
|---|---|
| ISSN: | 1470-160X |