We tested if marine heatwaves cause more year-to-year assemblage change in marine taxa other than trawl-surveyed fish. Fredston et al. (2023) found that heatwaves were not a dominant driver of change in demersal fish. We used 93 marine studies from BioTIME 2.0 (fish, invertebrates, seabirds, algae, plankton). We split the studies into 1,236 one-degree units and rarefied each year to equal samples. This gave 12,907 consecutive-year pairs (1983-2023). A heatwave window was 2 or more consecutive months above the monthly 90th percentile of HadISST anomalies (1982-2011 baseline) between the two surveys. Two-way fixed-effects models with equal study weight and study-clustered errors gave no heatwave effect on Jaccard turnover (-0.005, 95% CI -0.021 to 0.011), Bray-Curtis on relative abundance (-0.002, -0.020 to 0.017), log abundance ratio (-0.018, -0.125 to 0.089) or log richness ratio (-0.043, -0.099 to 0.012). Within-unit permutation tests, ERSST v5, severe and persistent heatwave definitions, a mean-anomaly control, standard-protocol studies and a placebo (heatwave after the survey) gave the same result. Sampling effort did not change in heatwave years. The data exclude general effects larger than about 0.15-0.25 within-unit SD. The community temperature index increased slightly after heatwaves in per-study estimates (+0.048 degC, 95% CI 0.017 to 0.079), but not in the primary pooled model, and the increase in 1982-2002 did not occur again in 2003-2023. Limitations: 1-degree monthly surface SST is a coarse proxy for habitat temperature and cannot detect short events; there are only 8 tropical studies and few pairs after 2019.
Filters: anzsrc-for-2020: 310305 ×
Question: do per-taxon sea surface temperature (SST) slopes fitted before 2014 predict the change of fish assemblages during and after the 2014-2016 northeast Pacific marine heatwave (MHW)? Data: CalCOFI winter-spring fish larvae on the southern California core grid (1951-2023, 45 consistently identified taxa) and RREAS May-June midwater trawl juveniles in central California (1990-2025, 68 taxon-stage units), with ERSSTv5 regional SST. Methods: month and station effects were removed with training years only; slopes were fitted on years up to 2013; skill is the cross-taxon Spearman rho between observed and predicted changes (equal taxon weight), with taxon permutations, earlier warm events and placebo block pairs as references. Results: in CalCOFI the change from 2009-2013 to 2014-2016 followed the slopes (rho 0.56, Pearson r 0.64, p < 0.001, calibration slope 0.99), as well as or better than for four earlier warm events. In RREAS it did not (rho -0.12); RREAS slopes also failed for 1997-1998. In 2017-2019 SST decreased by 0.43 C, but taxa did not change back as predicted (rho 0.12 CalCOFI, -0.20 RREAS), unlike after the 1997-1998 El Nino (0.72, 0.32). RREAS was below all 7 placebo cooling pairs; CalCOFI was below 11 of 12 but similar to two earlier weak after-event windows. A model with current and two prior years of SST predicts the 2017-2019 change (rho 0.45, 0.37), consistent with a lagged response to six warm years. Effort-standardized CalCOFI spring richness was 1.4 SD above the SST prediction in 2015 and above it in all sampled years 2017-2022. Limitations: few events; the change form, equal weighting and rho were chosen after first results (all variants are reported); unscaled CalCOFI results are dominated by northern anchovy; new taxa are not tested; SST is the only driver tested.