sciscape Publications
← All publications

Konza Prairie grasshopper declines reported as 2.1–2.7%/yr for 1996–2017 are log10 slopes read as percentages; the slopes mean 4.7–6.0%/yr, and the same method gives 3.5–6.3%/yr to 2020

Welti et al. (2020, PNAS) reported that grasshoppers in sweep samples at Konza Prairie, Kansas, decreased by 2.1%/yr in ungrazed watersheds (1996–2017) and 2.7%/yr in bison-grazed watersheds (2002–2017). We rebuilt the analysis from the current public release of the same Konza LTER data (CGR02, revision 26, 1,853 samples of 200 sweeps). OLS of log10 annual mean abundance on year gave the published slopes again within 0.003 (ungrazed −0.022, cycle removed −0.021; grazed −0.035 and −0.028 against published −0.033 and −0.027). The slopes are on a log10 scale, but the paper reports 100 × slope as the percentage. For a log10 slope b the annual change is 10^b − 1, so the published slopes mean declines of 4.7–4.9%/yr (ungrazed) and 6.0–7.3%/yr (grazed). The fitted line in the published Fig. 1A falls to about one third in 21 years, not to 0.64 as 2.1%/yr gives. With data to 2020, the published method gives 4.0%/yr for ungrazed (1996–2020, block-bootstrap 95% interval 1.3–6.5) and 6.3%/yr for grazed watersheds (2002–2020, 3.7–8.5); six long-record ungrazed watersheds give 3.5%/yr for 1983–2020 (1.7–5.1). Count models with site effects give 2.7–3.7%/yr for ungrazed watersheds, with intervals that include zero, and 6.2–7.5%/yr for grazed watersheds. Sampling moved later by 0.38 days/yr; adjustment for sampling date and weather removes 2–51% of the slope, depending on the method, and leaves 2.3–7.3%/yr. A rebuilt climate-index model (spring ENSO + summer NAO − winter PDO) fitted on 1996–2017 did not predict 12 held-out years (1983–1991, 2018–2020) better than a constant mean, but the 1996–2017 trend did. Limitations: we could not get the authors' code or supplement; the data revision differs from 2019; the index is a reconstruction; sweep catches measure relative abundance. We also give forecasts for 2021–2026, whose counts were not released.

insect decline · grasshoppers · Acrididae · Konza Prairie · LTER · tallgrass prairie · log10 slope · reproducibility · sampling date drift · climate oscillation index · out-of-sample test · population trend

Admitted 3 citing publications

References

4
  1. challengesdoi:10.1073/pnas.1920012117

    We reproduce its log10 slopes from the same data, but its stated %/yr decline rates are 100 x log10 slope; the slopes mean about 2.3 times larger annual declines. We also test its climate-index model on held-out years.

  2. usesdoi:10.6073/pasta/a7d6e4660ae286242519b788d280ccc7

    Konza Prairie LTER CGR02 grasshopper sweep-sample data, revision 26, used unchanged.

  3. doi:10.1073/pnas.2007803117

    Published correction to the same paper; it concerns feeding-guild labels only, not the rates.

  4. doi:10.1098/rsbl.2021.0510

    Later study of the same grasshopper community that repeats a decline of more than 2% a year.

Cited by

3 publications

Public catalog entries on Sciscape.