In 140 BioTIME 2.0 count studies, the faster decline of initially common species is matched by a rise of finally common species, and a truncation correction leaves most regression-to-the-mean bias

Question: Two syntheses report that species which are common at the start of a time series decline faster than rare species (van Klink et al. 2024; Xu et al. 2026, BioTIME 2.0). Both put species into classes by first-year abundance, and both correct for regression to the mean (RtM) with a truncation factor that they estimate on series of 20 years or more. Approach: We processed 4,917 BioTIME 2.0 count assemblages (142 studies, 10 or more years, rarefied) in the same way as Xu et al. We fitted Poisson and OLS trends for each species, and compared four contrasts: forward (first-year classes), the truncation correction, backward (last-year classes), and a time-reversal contrast S = (forward + backward)/2. S is zero in expectation for stationary, time-reversible dynamics with any noise or autocorrelation. We tested all estimators on simulated assemblages with AR(1) fluctuation and Poisson sampling. Results: The forward class 5 minus class 1 trend was -14.3 %/yr [95% study bootstrap -17.8, -11.1]. The truncation correction changed it only to -12.1. The backward contrast was a mirror image (+12.6). The symmetric part was -0.85 [-2.7, +0.8] %/yr, which is 6% of the forward value. Mean-abundance classes gave -1.7 [-4.5, +0.7]. The forward effect decreased as series length increased (-21 to -8 %/yr per log10 unit), as RtM predicts. The symmetric slope was near zero in all realms and in all taxa with 9 or more studies, with a small residual in terrestrial (-1.6 [-3.2, -0.2]) and freshwater assemblages. In simulations, the truncation correction removed only 10-28% of the spurious contrast, because the bias is larger in short series and autocorrelation carries it beyond year 1. The symmetric contrast was unbiased within 0.5 %/yr and detected true differential trends. Conclusion: In this BioTIME subset, the faster decline of initially common species agrees mostly with RtM and stationary fluctuation. Any directional erosion of dominance is at most about 2.7 %/yr. Limitations: This is not an exact replication of the hierarchical INLA model of Xu et al. The test assumes time-reversible fluctuation and constant sampling methods.

regression to the mean · BioTIME · population trends · abundance classes · dominant species · biodiversity change · time reversal · assemblage time series · common species decline · reanalysis