In BioTIME 2.0, short-term species turnover was lower in the second half of time series, also in 382 new studies and in series that started after 1990; the decrease is not specific to 1975

Nwankwo and Rossberg (2026) reported that species turnover over 1-5 year lags became slower in BioTIME communities after break years near 1975. We tested this result out of sample and examined its timing. First, we analysed their published community-level output again. At the 1975 break, 70% of 811 communities came from one bottom-trawl survey. When we count each study once, 14 of 23 studies show a decrease (p = 0.40). For break years 1971-2006, 34 of 36 medians are significant if communities are independent, but only 6 of 36 in a study-level sign test. Second, we rebuilt their method (Ochiai similarity, sample rarefaction, slope over lags 1-5) on BioTIME 2.0 for 35,043 local communities in 420 studies. For 90 single-site studies, our turnover estimates agreed with theirs (Spearman 0.90). When we split each series at its midpoint, turnover was lower in the second half in the 382 studies new in BioTIME 2.0: median change -0.0014 per year, 109 of 168 studies negative (p = 0.00014). Over all studies, the change was about 20% of the first-half turnover (189 of 310 studies negative). The decrease was equally large in series that started after 1990, and it stayed when we removed the first 5 survey years of each series. Thus it is not linked to the change in warming rate near 1975. When we shuffled the survey years, the median change was zero, so the method does not cause the decrease. Tests at fixed calendar decade did not show whether the decrease follows calendar time or series age. Main limitations: our cell and sample definitions differ from the original ones, the new studies cover mostly the years after 1990, and study-level tests have low power.

temporal turnover · BioTIME 2.0 · replication · out-of-sample test · species composition · Ochiai similarity · pseudoreplication · time series · biodiversity change