SCIENCE@home

Machine learning reveals ecological thresholds and predicts future bacterial community shifts in a chronically eutrophic estuary.

Chuting Chen,Dongyao Sun,Yanjiao Lai,Bingqian Zhu,Zhisheng Zhou,Yifan Song,Xiaodong Wang,Yan Wang,H. Yao

Published 2026 in Journal of Environmental Management

ABSTRACT

Effectively managing eutrophic estuaries under climate change requires a quantitative understanding of how microbial communities respond to multiple environmental stressors. However, predicting these responses is challenging due to complex, non-linear interactions. This study addresses this gap by developing a machine learning framework to model the seasonal dynamics of the bacterial community in the chronically eutrophic Pearl River Estuary. We aimed to identify key environmental drivers, define their ecological thresholds, and forecast community shifts under future climate scenarios. Our models revealed that temperature, silicate, and the nitrite/dissolved oxygen were the dominant drivers, collectively explaining 66-82 % of the community variance. Crucially, we identified specific environmental thresholds beyond which the abundance of key phyla, including Firmicutes, Actinobacteriota, and Proteobacteria, significantly changed. Under a high-emission scenario (SSP5-8.5), the model projects an increase in β-diversity but a potential 4.88 % decline in the relative abundance of key biomarker taxa, with the upper and lower estuary emerging as hotspots for future community restructuring. This study provides a robust, predictive tool that moves beyond monitoring to proactive management. The identified thresholds offer clear, data-driven targets for pollution control and ecosystem restoration, providing a transferable model for developing evidence-based management strategies for other complex estuarine systems globally.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-73 of 73 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1