Category: Ecology

Summary: Testing whether species-level interaction heterogeneity can widen a stable-but-reactive window while concentrating the amplifying mode onto fewer species.

May-type random community matrices are a classic benchmark for ecological stability, but asymptotic stability does not rule out large transient amplification. This GPU experiment asks whether increasing species-to-species heterogeneity can simultaneously widen the stable-but-reactive regime and localize the leading reactive mode onto a smaller subset of species.

The study measures both the width of the reactive window and the spatial concentration of the dominant amplifying mode across heterogeneous ensembles. That combination makes it possible to distinguish communities that are merely stable from communities that are stable yet highly vulnerable to localized bursts.

The broader value is mechanistic. If heterogeneity localizes reactivity while reshaping stability margins, then ecological risk may depend as much on where transient amplification sits as on whether eigenvalues alone predict eventual return to equilibrium.

Method: GPU batched May-matrix simulations sweeping species-level interaction heterogeneity while measuring reactivity and eigenvector localization.

What is measured: Width of the stable-but-reactive regime, localization of the leading reactive mode, heterogeneity dependence, and stability-reactivity tradeoffs.