Source cut-off years and compilation lag cause more than half of the post-1990s fall in global first records of non-native species in FirstRecords v4.0, 1950–2019

Question: The global count of first records of non-native species in FirstRecords v4.0 falls after the late 1990s (7,996 records in 1990–1999, 2,768 in 2010–2019). Is this fall real, or does it come from recording effects? Approach: We used the citation year of each source list to set its cut-off year. We fitted a Poisson model to counts for each source x location unit (1,405 units, 39,958 records from 1950 on) with a common calendar index. This removes the coverage effect of source cut-offs. We measured compilation lag directly in 80 pairs of older and newer lists for the same country (36,573 records). Relative to species first recorded 40 or more years before its cut-off, the older list was 52% (95% CI 32–69%) complete for species first recorded 0–4 years before its cut-off, about 85–90% complete at 5–19 years, and about 100% complete at 20 or more years. We added this lag curve as an offset and did a bootstrap over sources. Results: The raw ratio of 2005–2014 to 1990–1999 was 0.63. The coverage-adjusted ratio was 0.69 (0.53–0.90), and the coverage and lag adjusted ratio was 0.83 (0.58–1.17). For 2010–2019 the ratios were 0.35, 0.45 (0.29–0.67) and 0.63 (0.37–1.06). On the log scale, the corrections remove 56–60% of the raw fall. A vintage backtest at 2005, 2010 and 2015 put the adjusted estimate nearer to estimates from later sources than the raw or coverage-only estimates, but still below them. The unharmonised per-source table, four lag-curve variants, data with only lags of at least 5 or 10 years, and splits by plants, insects, other taxa and by Europe and outside Europe gave the same conclusion. Limits: The method assumes that the lag curve does not change with time, region or taxon, and that the two lists in a pair record species independently. Copying between lists makes the correction too small. The first-record year is the detection year. The estimates after 2015 depend on few sources. Thus the data do not show a clear fall in the per-source rate of first records from 1990–1999 to 2005–2014.

biological invasions · non-native species · first records · reporting lag · nowcasting · FirstRecords · Global Biodiversity Framework Target 6 · capture-recapture · time series bias