Category: Ecology

Summary: Estimating the self-regulation strength where transient amplification disappears in trophically coherent nonreciprocal food-web operators before full asymptotic stabilization.

Ecological communities can be asymptotically stable yet still show dangerous transient amplification after a disturbance. This experiment asks for the earlier suppression point in a trophically coherent food-web model: the self-regulation strength at which positive symmetric reactivity vanishes, even before the larger stability transition is reached.

The script builds dense symmetric reactivity operators for nonreciprocal food webs and runs GPU eigensolves entirely on CuPy when hardware is available. It uses iterative deepening and bisection on self-regulation strength, scaling matrix size up to the practical VRAM limit.

That distinction matters because transient amplification can be ecologically important on its own. The result therefore targets the threshold for eliminating reactive bursts, not just the later point where all linear instability indicators disappear.

Method: GPU dense symmetric eigensolves with iterative deepening and bisection on self-regulation strength in trophically coherent food-web reactivity operators.

What is measured: Critical self-regulation threshold, disappearance of positive symmetric reactivity, finite-size threshold bracket, and GPU-scaled system size reached.