Harmful algal blooms (HABs) can have dire repercussions on aquatic wildlife and human health, and may negatively affect recreational uses, aesthetics, taste, and odor in drinking water. The factors that influence the occurrence and magnitude of harmful algal blooms and toxin production remain poorly understood and can vary in space and time. It is within this context that we use machine learning (ML) and two 14-year (2005–2018) data sets on water quality and meteorological conditions of China’s lakes and reservoirs to shed light on the magnitude and associated drivers of HAB events. General regression neural network (GRNN) models are developed to predict chlorophyll a concentrations for each lake and reservoir during two study periods (2005–2010 and 2011–2018). The developed models with an acceptable model fit are then analyzed by two indices to determine the areal HAB magnitudes and associated drivers. Our national assessment suggests that HAB magnitudes for China’s lakes and reservoirs displayed a decreasing trend from 2006 (1363.3 km2) to 2013 (665.2 km2), and a slightly increasing trend from 2013 to 2018 (775.4 km2). Among the 142 studied lakes and reservoirs, most severe HABs were found in Lakes Taihu, Dianchi and Chaohu with their contribution to the total HAB magnitude varying from 89.2% (2013) to 62.6% (2018). HABs in Lakes Taihu and Chaohu were strongly associated with both total phosphorus and nitrogen concentrations, while our results were inconclusive with respect to the predominant environmental factors shaping the eutrophication phenomena in Lake Dianchi. The present study provides evidence that effective HAB mitigation may require both nitrogen and phosphorus reductions and longer recovery times; especially in view of the current climate-change projections. ML represents a robust strategy to elucidate water quality patterns in lakes, where the available information is sufficient to train the constructed algorithms. Our mapping of HAB magnitudes and associated environmental/meteorological drivers can help managers to delineate hot-spots at a national scale, and comprehensively design the best management practices for mitigating the eutrophication severity in China’s lakes and reservoirs.
作者： Jiacong Huang, Yinjun Zhang, George B.Arhonditsis, Junfeng Gao et al. Water Research. 2020. DOI: doi.org/10.1016/j.watres.2020.115902
