Two US trawl richness slopes required at least 49 and 36 taxon deletions to reverse their signs in fixed 1982-2019 samples

We tested how taxon labels contributed to sea temperature and richness slopes in two US trawl surveys and two periods. We decomposed fixed-effort richness slopes into label contributions, then found the smallest complete label set whose removal reversed each point-estimate sign. In the Eastern Bering Sea in 2001-2019, at least 49 of 468 active labels had to be removed. In Northeast US fall surveys in 1982-2000, at least 36 of 228 had to be removed. These sets supplied 23.8% and 16.9% of mean expected richness. Two other slopes had intervals that included zero and reversed after two and six deletions. A two-haul effort target kept the broad count contrast. The counts depend on the fixed table, design and deletion rule. They are not error rates or confidence limits. Early Northeast positive contributors had a negative sum in the later period. Shared observations, short time series, gear changes, detection and coarse surface temperature limit interpretation. The result does not establish a causal climate effect. The package gives code, sufficient evidence, raw hashes, a raw rebuild recipe, independent checks and a figure.

biodiversity · observation bias · taxon labels · sensitivity analysis · trawl survey · sea surface temperature · rarefaction