Removal of linear trends reduced the geometric mean variance ratio for CSV-listed plants from 1.33 to 1.07 in 35 Cedar Creek 16-species plots, 2010–2021

We tested net covariance of planted species biomass in the later harvests of the Cedar Creek Biodiversity II experiment. The planned primary measure used species with a planted-membership flag in the source CSV. We rebuilt species-sorted samples for 154 plots in 2010–2021. For each plot, R is total biomass variance divided by the sum of species variances. R below one means negative net covariance. In 35 plots with 16 planted species, the geometric mean R was 1.329 (95% plot-bootstrap interval 1.171–1.511). After removal of separate species linear trends, it was 1.071 (0.934–1.226). A joint plot-and-year-block interval was 0.712–1.356. A later check included extra plants from the metadata planting lists. It gave R = 1.214 (1.066–1.380) before trend removal and 0.993 (0.855–1.157) after it. Thus, both definitions showed positive net covariance in raw values, while the residual sign was unresolved. Mean standard synchrony was 0.194, above its zero-covariance reference of 0.136. Low synchrony alone did not show net compensation. Linear trends supplied 64.6% of the equal-plot mean raw net covariance. This is a descriptive decomposition, not a causal result. The result concerns planted species at one managed site and is sensitive to the time window. The package preserves raw data, source provenance, code, figures, all planned checks and an independent numerical reconstruction.

biodiversity · grassland · species covariance · variance ratio · synchrony · Cedar Creek · reproducibility