2026/03/19 by Bernd von Mallinckrodt · 1 voice
Environmental Science · #Marine and fisheries research #Ecosystem dynamics and resilience #Coral and Marine Ecosystems Studies
paper · doi:10.5281/zenodo.19110905
openalex publication_date 2026/03/19 · openalex created_date 2026/03/20 · openalex updated_date 2026/07/01
This study evaluates whether structural and dynamical indicators provide independent predictive value for fisheries collapse beyond classical stock assessment metrics. Using a dataset of 46 marine fish stocks derived from RAM Legacy and ICES assessments (24 collapsed, 22 stable), we compare baseline models based on fishing pressure (F/FMSY) and spawning stock biomass (SSB/BMSY) with extended models incorporating autocorrelation of fishing pressure (AR1), a structural compression proxy, and trajectory-based change rates. The baseline model achieves near-perfect classification performance (AUC = 1.0) across all prediction horizons up to ten years prior to collapse, reflecting a well-known empirical property of fisheries datasets in which collapsed and stable stocks are strongly separated in F–SSB space. Under these conditions, no incremental classification gain from structural variables can be detected (ΔAUC ≈ 0, likelihood ratio tests non-significant). Despite this ceiling effect, structural and dynamical variables demonstrate substantial standalone discriminative ability (AUC 0.82–0.95), indicating that they encode meaningful information about collapse dynamics. However, no lead-time advantage over classical indicators is observed within the present retrospective design. A structural compression index is shown to be algebraically collinear with classical variables (VIF ≈ 39), precluding its interpretation as an independent predictor in its current formulation. These results do not falsify structural-dynamical approaches but highlight a methodological limitation: incremental predictive value cannot be evaluated in regimes where classical indicators already achieve near-perfect separation. The framework is therefore best interpreted as a complementary structural-dynamical modelling layer, potentially valuable in earlier-phase, data-limited, or high-uncertainty regimes where classical indicators are less dominant. Future work should focus on earlier pre-collapse windows, larger datasets, and the construction of structurally independent indicators derived from time-series properties rather than transformations of existing reference-point metrics. 🔑 Keywords fisheries collapse early warning signals critical transitions structural dynamics adaptive capacity AR1 autocorrelation time series analysis stock assessment resilience complex systems tipping points RAM Legacy database ICES fisheries predictive modelling CRTI framework