Evidence for a linear change in grasshopper genus shares depended on survey date terms at six Konza watersheds in 1997–2020

We tested a common linear time component in grasshopper genus shares at six fixed, ungrazed Konza Prairie watersheds in 1997–2020. The balanced panel has 576 surveys, 144 watershed-years and 35,785 insects in 22 genera. The primary model used square roots of genus shares, separate watershed intercepts and a common time coefficient for each genus. Year explained 5.82% of the reduced-model variation. Moving four-year residual blocks together across all watersheds and genera gave a reference probability of 0.139 over all 720 orders. With mean survey day and its square in both models, partial R² was 0.0584 and the reference probability was 0.046. Mean survey day moved from 206.83 in 1997–2004 to 217.16 in 2013–2020. The primary test did not establish a common linear change. Evidence was stronger in the date model, but survey timing can reflect observation conditions and real ecological timing. This model dependence does not show stable communities or identify a cause. We report all planned checks, share slopes, simultaneous intervals and descriptive era means. The package has unchanged source data, provenance, code, all results and figures. A separate parser checked all 3,168 count cells. Direct fits checked all 1,440 primary and date-model reference statistics. A clean offline repeat matched all generated files. This is a test of recorded catches at one managed prairie, not of true population composition.

grasshopper · community composition · Konza Prairie · survey timing · Hellinger transformation · block permutation · reproducibility