Zhou Xi-Yin, Xve Baolin, Khu Soon-Tiam, Yinglan A, Wang Yuntao, Wang Guoqiang
School of Systems Science, Beijing Normal University, Beijing 100875, PR China.
Faculty of Geographical Science, Beijing Normal University, Beijing 100875, PR China.
J Hazard Mater. 2025 Aug 15;494:138701. doi: 10.1016/j.jhazmat.2025.138701. Epub 2025 May 23.
Algal blooms, driven by complex environmental and climatic factors, present significant ecological and economic challenges in aquatic systems worldwide. This study investigates the outbreak characteristics, driving mechanisms, and early warning approaches for harmful algal blooms (HABs).In analyzing bloom characteristics, we evaluate the total bloom duration, individual event peaks, and the rate of concentration change, identifying distinct patterns across different lake systems. We further classify these events into clusters based on duration and peak concentration, offering insight into the heterogeneity of bloom dynamics across varied water bodies. The driving mechanisms of algal blooms are elucidated through causal and network pathway analysis, revealing the direct and indirect effects of environmental variables on chlorophyll-a concentrations. Across lakes, water temperature and nutrient concentrations (TN, TP) emerge as primary drivers, while factors like nitrogen-phosphorus ratios act as limiting constraints in some systems, particularly in Dongtinghu and Dianchi. The interactions between these drivers highlight the complexity of algal bloom outbreaks, with both positive and negative feedback loops playing critical roles in bloom formation. To improve the predictive capacity for HABs, we propose a Proactive Early Warning Model of Algal Blooms (PEWAB), integrating causal discovery using the Peter and Clark Momentary Conditional Independence (PCMCI) algorithm with a Graph Convolutional Network (GCN). This hybrid approach captures the spatial and temporal dynamics of environmental factors, allowing for the identification of significant causal pathways and time-lagged dependencies. The GCN-based model processes these causal relationships to forecast future chlorophyll-a concentrations, providing a robust proactive early warning system for managing and mitigating the impacts of algal blooms. Our findings contribute to an understanding of the environmental drivers behind algal blooms and offer a proactive, data-driven framework for early detection and prevention, crucial for preserving aquatic ecosystems and public health.
由复杂的环境和气候因素驱动的藻华，给全球水生系统带来了重大的生态和经济挑战。本研究调查了有害藻华（HABs）的爆发特征、驱动机制和预警方法。在分析藻华特征时，我们评估了藻华的总持续时间、单个事件峰值以及浓度变化率，确定了不同湖泊系统中的不同模式。我们还根据持续时间和峰值浓度将这些事件分类为不同的簇，从而深入了解不同水体中藻华动态的异质性。通过因果关系和网络路径分析阐明了藻华的驱动机制，揭示了环境变量对叶绿素a浓度的直接和间接影响。在各个湖泊中，水温以及营养物质浓度（总氮、总磷）是主要驱动因素，而氮磷比等因素在某些系统中起到限制作用，特别是在洞庭湖和滇池。这些驱动因素之间的相互作用凸显了藻华爆发的复杂性，正反馈和负反馈回路在藻华形成中都起着关键作用。为了提高对有害藻华的预测能力，我们提出了一种藻华主动预警模型（PEWAB），该模型将使用彼得和克拉克瞬时条件独立性（PCMCI）算法的因果发现与图卷积网络（GCN）相结合。这种混合方法捕捉了环境因素的时空动态，从而能够识别重要的因果路径和时间滞后依赖性。基于GCN的模型处理这些因果关系以预测未来的叶绿素a浓度，为管理和减轻藻华的影响提供了一个强大的主动预警系统。我们的研究结果有助于理解藻华背后的环境驱动因素，并提供一个主动的、数据驱动的早期检测和预防框架，这对于保护水生生态系统和公众健康至关重要。
