AI 中文总结
该研究发现生态群落接近临界稳定性时，软模主导波动的预期可能不成立，三类生态场景中均未出现预测的强成对丰度关联，说明临界态临近不一定体现在等时成对关联中。
AI 中文摘要
生态群落不断受入侵、排除和多样化过程重塑，这些过程自然地将其推向动态稳定性边界。在该边界附近，软模的弛豫速度越来越慢，在随机驱动下，预计将主导波动，有效将动力学简化为一维，并产生强正负丰度关联。因此，这类关联被视为即将发生转变的标志。本文表明，即使在任意接近临界点时，这一预期也可能不成立。原因在于，谱软度并不保证随机可见性：软模必须获得足够的环境驱动，才能主导其余模式产生的波动背景。我们在三种生态场景中验证了该机制：合成可行群落、由区域库迁入组装的局域群落，以及通过重复多样化生成的群落。在所有三种场景中，群落接近边际稳定性，却未出现单模图像预测的近乎完美的成对关联。因此，生态临界态的临近不一定能在等时成对关联中体现出来。
英文摘要
Ecological communities are continually reshaped by invasion, exclusion, and diversification, processes that naturally drive them toward the boundary of dynamical stability. Near such a boundary, a soft mode relaxes increasingly slowly and, under stochastic forcing, is expected to dominate the fluctuations, effectively reducing the dynamics to one dimension and generating strong positive and negative abundance correlations. Such correlations have therefore been proposed as signatures of an imminent transition. Here we show that this expectation can fail even arbitrarily close to criticality. The reason is that spectral softness does not guarantee stochastic visibility: the soft mode must receive enough environmental forcing to dominate the fluctuation background generated by the remaining modes. We demonstrate this mechanism in three ecological scenarios: a synthetic feasible community, a local community assembled by immigration from a regional pool, and a community generated by repeated diversification. In all three, communities approach marginal stability without developing the near-perfect pairwise correlations predicted by the single-mode picture. Thus, proximity to ecological criticality need not be visible in equal-time pairwise correlations.
