Covariance Concentration as an Early-Warning Signal for Modular Structural Collapse

Read full text Summarise with VIRAT
Authors: Bernd von Mallinckrodt
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Classical early-warning signals (EWS), such as rising variance and lag-1 autocorrelation, assume critical slowing down as a universal precursor mechanism. These indicators can fail when collapse is driven by structural reorganization rather than proximity to a bifurcation. This work introduces Φ, defined as one minus the normalized effective rank of a covariance matrix estimated via Oracle Approximating Shrinkage (OAS), as a complementary indicator for modular structural compression (M3). In this regime, variance is redistributed across eigenmodes rather than amplified, rendering amplitude-based indicators uninformative. Detection is formalized using a surrogate-based hypothesis test with IAAFT (phase-randomized) null models, combined with duration and slope constraints and explicit suppression guards (p/n < 0.5, baseline stationarity, and confirmed modular structure via correlation-based modularity Q). Φ is not proposed as a universal EWS and is not expected to perform in bifurcation-driven (M1) or noise-amplified (M2) regimes. Its applicability is restricted to multivariate systems in which covariance concentration constitutes the dominant precursor mechanism. early warning signals, structural compression, covariance concentration, effective rank, modularity, complex systems, critical transitions, OAS shrinkage, surrogate data, mechanism dependence

Fields of Study & Subjects

Ecosystem dynamics and resilience Complex Systems and Time Series Analysis Theoretical and Computational Physics

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
30 March 2026
Language
EN
Document type
Preprint