Susan Ville
Bernd von Mallinckrodt 🦋Dynamical Indicators Provide S...
Dynamical Indicators Provide Six to Ten Years of Additional Early Warning of Fisheries Collapse Beyond Classical Reference Points
This study examines whether structural-dynamical indicators provide earlier detection of fisheries collapse than classical reference-point metrics such as fishing pressure (F/F_MSY) and spawning stock biomass (SSB/B_MSY). While these traditional indicators achieve near-perfect classification performance in datasets with strong separation between collapsed and stable stocks, they encode system state rather than trajectory. Using 46 stock trajectories calibrated to the RAM Legacy Stock Assessment Database and focusing explicitly on the early-phase regime (10–20 years prior to collapse), we evaluate two dynamical indicators: the lag-1 autocorrelation of fishing pressure (X3) and the log-rate of change of relative spawning biomass (X5). Discrimination performance (AUC) is saturated across all models and therefore uninformative. Instead, we assess predictive value through lead-time analysis. We find that X3 and X5 detect impending collapse a mean of 7.6 and 9.0 years earlier, respectively, than classical reference-point indicators (p < 0.001). Operational deployment requires persistence filtering of at least four consecutive years to control false positives, under which X5 achieves a false-positive rate of 0.091 while retaining substantial lead-time advantage. These results indicate that dynamical indicators provide a complementary early-warning layer whose contribution is temporal rather than discriminative, shifting the evaluation of collapse prediction from classification performance to temporal detectability. The framework is not a replacement for classical fisheries metrics but an extension that captures trajectory-level information preceding threshold crossing. Keywords: early-warning signals, fisheries collapse, critical slowing down, lead-time detection, stock assessment, dynamical systems, regime shifts, ecological forecasting
ChuckDiesal
Ilan Kelman
Vladyslav
Ciro Borrelli
💧🌏 Greg Cocks
Academia Cidadã