Antao et al. (2020) used 21,500 temperate BioTIME assemblage time series. They reported that in the sea, warming increased richness and decreased abundance more at warm baseline temperatures, and that there was no such link on land. We tested these marine results with three methods. (1) With the authors' published series table, a REML version of their model gave the published warming x baseline terms (richness 0.078, abundance -0.177), and an added intercept changed nothing. When the warming effects could also vary among the 75 marine studies, the richness term was 0.063 (95% CI -0.05 to 0.18) and the abundance term changed sign to +0.24 (-0.13 to 0.62). Study-clustered SEs were 3.6-6.3 times the model SEs. Removal of the largest study (a pelagic longline observer programme, 20% of marine series) changed the richness sign to -0.073. Removal of 3 or 5 large surveys changed the abundance sign. (2) Series rebuilt from raw BioTIME 2.0 records for the same studies (r = 0.73 with the original trends) gave a richness term of -0.15 with the original model and +0.05 (-0.14 to 0.23) with study-level effects. (3) In 367 marine series from 29 studies first added in BioTIME 2.0, the terms were +0.074 (SE 0.072) for richness and -0.12 (SE 0.14) for abundance. These are the published signs, and the signs were stable in leave-one-study-out fits, but both intervals include zero. The original-series model predicted new richness trends with calibration slope 0.94 (SE 0.52) and did not predict abundance trends (-0.05, SE 0.06). Land (72 new studies) and freshwater (88 studies) gave no clear interaction. Conclusion: a few large surveys supply most of the evidence for the reported marine coupling, and it is not robust when the effect can vary among studies. The new studies are consistent with it but too few to confirm it. Limitations: frequentist approximation of the brms model, square instead of hexagonal cells, coarse mapped baselines, and a small new marine sample from mainly the North Atlantic.
Filters: anzsrc-for-2020: 410401 ×