Exploited CalCOFI larval fish taxa had higher interannual variability than unexploited taxa of similar maturation age in 1951–2002 and in new 2003–2022 data

Hsieh et al. (2006) found that, in the CalCOFI larval fish survey off southern California (1951–2002), exploited taxa have a higher coefficient of variation (CV) of annual abundance than unexploited taxa with a similar age at maturation. We rebuilt the analysis from the public NOAA ERDDAP tables: oblique tows at the 66 core stations, the same 29 taxa, and the published fishing status and ages. Then we tested the result with independent 2003–2022 data. For 1951–2002, the age-adjusted CV excess was +0.44 (95% bootstrap interval 0.05 to 0.89; permutation p = 0.011); the published values give +0.52. In 2003–2022 the excess was +0.35 (0.06 to 0.64; p = 0.027). In 1985–2002, which has the same survey design, it was +0.28. The excess did not change between these two periods. We removed station-sampling noise with a split-half covariance method and divided each series into a slow part (3-year kernel) and a fast part. The fast part was higher for exploited taxa in all three periods (+0.33, +0.28, +0.38; p = 0.013 to 0.015). This result did not change with covariates for assemblage, spatial patchiness or abundance, or when we removed rockfish or the sardine, anchovy and mackerel group. The slow-part excess occurred only in 1951–2002, and sardine, anchovy and mackerel caused it. On a log scale, proportional variability did not differ before 2003 (p = 0.59 and 0.98) but was higher for exploited taxa in 2003–2022 (p = 0.03). Thus the main CalCOFI result replicates out of sample. Station-sampling noise and slow trends do not cause it. Before 2003 it was in arithmetic-scale booms, not in larger proportional changes. Limits: 13 vs 16 taxa; no correction for multiple metrics; fishing status fixed at the 2005 classification; few summer and fall cruises after 2017. This fixed-grid design cannot separate abundance variability from distribution shifts.

CalCOFI · ichthyoplankton · fishing effects · population variability · coefficient of variation · age truncation · replication · out-of-sample test · California Current · sampling noise · split-half covariance