Understanding the impacts of multiple environmental stressors on phytoplankton biomass is crucial for predicting marine ecosystem responses under global climate change. This study employed a sequential modeling framework integrating principal component analysis, generalized additive models, and artificial neural networks to improve predictions of phytoplankton chlorophyll a concentrations in the Taiwan Strait. Analyzing a decadal dataset, we found that a 2C rise in sea surface temperature and a 0.2 pH decline will each lead to an 11.3% reduction in chlorophyll a biomass, whereas nitrogen enrichment is expected to increase it by only 2.8%. The combined effects of these stressors will result in an 18.3% reduction, with the most significant declines occurring in high-chlorophyll areas during algal blooms. Compared to simpler models, our approach improved accuracy by reducing overestimation biases, particularly under acidification scenarios, highlighting the need for advanced, multivariate models in forecasting phytoplankton dynamics under global changes.
Scientific Significance Statement
Phytoplankton are critical to marine ecosystems and global biogeochemical cycles, yet predicting their responses to the combined stressors of warming, acidification, and eutrophication remains a major challenge. Traditional models struggle with multicollinearity and nonlinear interactions among environmental variables. This study introduces an innovative sequential modeling framework that integrates principal component analysis, generalized additive models, and artificial neural networks, addressing these challenges by combining the strengths of each method in a series. This approach not only improves predictive accuracy, particularly during algal blooms, but also reveals that combined stressors lead to an 18.3% decline in phytoplankton biomass, underscoring their vulnerability under future climate scenarios. By bridging methodological advancements with ecological discovery, this work provides a powerful tool for understanding and forecasting marine ecosystem responses to global change.
Tong Z., Guo J., Liu Y., Lin L., Chen J., Liu X., Huang B., Laws E. A. & Xiao W., in press. Novel sequential modeling framework improves phytoplankton biomass predictions in response to multiple environmental stressors. Limnology and Oceanography Letters. Article.
