Shifts between the alternative stable states have been a popular topic of ecological studies for over 50 years. However, identifying the mechanisms that drive these regime shifts remains a challenging task in the field of applied ecology and ecosystem management. Herein, we applied a Bayesian latent variable regression (BLR) to the dataset obtained for a shallow lake experiencing a shift from the dominance of submerged plants to the dominance of floating plants to classify the observed lake states, determine total phosphorus (TP) thresholds, and establish a steady-state relationship between the phosphorus content and the size of an aphotic zone. In addition, a structural equation model was utilized to elucidate the mechanisms driving this shift and their effects on the trophic web of the ecosystem. The BLR model exhibited high state classification rates (median/mean accuracy > 97.5%) and accurately determined the steady-state relationships between the TP thresholds and the aphotic zone size. We also found that the top-down control via piscivorous fish was stronger in the state dominated by submerged macrophytes, while the bottom-up control via TP was stronger in the state dominated by floating macrophytes. Additionally, the dominance of floating macrophytes was related to the low degree of biodiversity and simplification of the trophic web. Our results show that top predators can be used to control nutrient levels and ensure the dominance of submerged macrophytes. It also provides a new perspective on the theory of alternative states and potential tools for the management of ecosystems experiencing regime shifts.

在过去的50多年中，替代稳态之间的转换一直是生态研究的热门话题。然而，在应用生态学和生态系统管理领域，确定驱动这些制度转变的机制仍然是一项艰巨的任务。在此，我们将贝叶斯潜在变量回归（BLR）应用于获得浅水湖泊（从淹没植物的优势转移到漂浮植物的优势）的数据集，以对观察到的湖泊状态进行分类，确定总磷（TP）阈值并在磷含量与无光区的大小之间建立稳态关系。此外，利用结构方程模型阐明了驱动这一转变的机制及其对生态系统营养网的影响。BLR模型表现出较高的状态分类率（中位数/平均准确度> 97.5％），并准确确定了TP阈值与无光区带大小之间的稳态关系。我们还发现，在淹没水生植物主导的状态下，通过食肉鱼的自上而下的控制更强，而在漂浮水生植物主导的状态下通过TP的自下而上的控制更强。此外，漂浮大型植物的优势与生物多样性程度低和营养网的简化有关。我们的结果表明，顶级捕食者可用于控制营养水平并确保淹没大型植物的优势。它还为替代状态理论和用于管理经历制度转移的生态系统的潜在工具提供了新的视角。