Causality analysis and prediction of riverine algal blooms by combining Convergent Cross-Mapping and machine learning techniques

River algal blooms are a global environmental problem that is characterized by complex causes, large impact range and long duration. This problem would be further complicated by the operation of large-scale water transfer projects. Here, we combine the Convergent Cross-Mapping (CCM) and machine learning to reveal the causes and predict the occurrence of algal blooms during a 15-year period in the Han River, the largest tributary of the Yangtze River of China and the freshwater source of the central route of the South-to-North Water Diversion Project (SNWDP), a mega infrastructure of China. We show that the random forest, a machine learning model, outperformed the gradient boosting machine model in predicting algal blooms with a 10-day lead time with the application of resampling methods to tackle imbalanced data (i.e., the algal blooms events are relatively rare compared to the non-blooms). We combine CCM causality analysis and machine learning to elucidate that the water temperature in the Han River (HR) and the water levels in HR and the Yangtze River as the dominant factors affecting algal blooms in HR with consistently high nutrient concentrations. Finally, we suggest that the operation of SNWDP weakens the response of algal blooms to the water level variation in HR, which is one of the key driving factors affecting algal blooms before the operation of SNWDP. This study provides important reference for detecting the causes of river algal blooms and exploring the impact of large-scale water transfer projects on riverine environment.
The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)