摘要：Understanding how phytoplankton interacts with local and regional drivers as well as their feedbacks is a great challenge, and quantitative analyses of the regulating role of human activities and climate changes on these feedback loops are also limited. By using monthly monitoring dataset (2000-2017) from Lake Taihu and empirical dynamic modelling to construct causal networks, we quantified the strengths of causal feedbacks among phytoplankton, local environments, zooplankton, meteorology as well as global climate oscillation. Prevalent bidirectional causal linkages between phytoplankton and the tested drivers were found, providing holistic and quantitative evidence of the ubiquitous feedback loops. Phytoplankton exhibited the highest feedbacks with total inorganic nitrogen and ammonia and the lowest with nitrate. The feedbacks between phytoplankton and environmental factors from 2000 to 2017 could be classified by two groups: the local environments (e.g., nutrients, pH, transparency, zooplankton)-driven enhancement loops promoting the response of the phytoplankton, and the climate (e.g., wind speed)-driven regulatory loops suppressing it. The two counterbalance groups modified the emergent macroecological patterns. Our findings revealed that the causal feedback networks loosened significantly after 2007 following nutrient loading reduction and unsuccessful biomanipulation restoration attempts by stocking carp. The strength of enhancement loops underwent marked decreases leading to reduced phytoplankton responses to the tested drivers, while the climate (decreasing wind speed, warming winter)-driven regulatory loops increased– like a tug-of-war. To counteract the self-amplifying feedback loops, the present eutrophication mitigation efforts, especially nutrient reduction, should be continued, and introduction of alternative measures to indirectly regulate the critical components (e.g., pH, Secchi depth, zooplankton biomass) of the loops would be beneficial.