Kelp loss and kelp recovery at 279 California reef sites (1983–2025) had urchin thresholds 1.3 times apart (95% CI 0.7–11) when each site had its own threshold

Question: on the same reef sites, do kelp losses happen at higher sea urchin densities than kelp recoveries? A global synthesis gives a 9.4-fold hysteresis (668 vs 71 g/m2 urchins). Data: 3,288 site-years at 339 California sites from three public monitoring programs: NPS Kelp Forest Monitoring (Channel Islands, 1983-2007), PISCO (1999-2024) and SBC LTER (2000-2025). 2,711 consecutive-year pairs at 279 sites gave 149 kelp losses and 152 recoveries (kelp limit 0.1 plants/m2). Method: a model of annual state change with a memory term, an urchin threshold for loss T_c, an urchin threshold for recovery T_r, and H = T_c/T_r. We fitted it with and without a site-level random threshold, used profile-likelihood intervals and a SIMEX correction for counting error, and tested the estimator on data simulated along the observed urchin series. Results: without site effects, the common H was 25 (95% CI 1.6 to >100; H = 1 rejected, p = 0.003), and SIMEX changed this only to 20-24. When each site had its own threshold, the common H was 1.3 (0.71-11; p = 0.46 for H = 1). Site thresholds differed by a factor of 1.9-4.5 (SD) in the three larger programs. In simulations with true H = 9.4 the estimator gave a median of 10.3 (4.9-31), and none of 100 data sets gave an estimate as low as 1.3. The profile likelihood alone was weaker against H = 9.4 (p = 0.13). Only KFM showed within-site hysteresis (H = 5.6, 1.1-16). Across five state and urchin-year specifications, the common within-site H was 0.63-5.0, with upper 95% limits of 1.6-14. Observed losses occurred at a median of 13.9 urchins/m2 (665 g/m2) and recoveries at 6.0/m2 (288 g/m2). But simulations show that this ratio of medians changes little between H = 1 and H = 9.4, so it is not a test of hysteresis. Conclusion: kelp states persist, but most apparent hysteresis in pooled transitions comes from differences in threshold between sites. A large (about 10-fold) within-site hysteresis was not found, but it cannot be excluded in all specifications. Limitations: annual surveys, urchin behaviour changes count detectability, urchin dynamics are taken as given, and public KFM data end in 2007.

kelp forest · sea urchin barren · hysteresis · alternative stable states · regime shift · Macrocystis pyrifera · Strongylocentrotus purpuratus · long-term monitoring · random effects · California · PISCO · SBC LTER · Channel Islands