Selecting web codes 1–3 reversed the estimated change in spring rodent catch ratios at two Sevilleta sites in 1999–2025

We tested how web selection changed a comparison of spring rodent catches between two Sevilleta sites in New Mexico. The source was version knb-lter-sev.8.297979. The matched sample had 10 years in 1999–2008 and 15 years in 2010–2025. We omitted 2009 and an absent shrubland first night in 2024. A filter for web codes 1–3 admitted two grassland codes early but one later, and two shrubland codes in both periods. This filter gave a later-to-earlier change of 0.662 in the grassland-to-shrubland catch ratio. Code 5 gave 1.479, and all recorded codes gave 1.345. Dividing the code 1–3 counts by their observed web-code counts changed 0.662 to 1.324, an exact factor of two. Nominal three-year block intervals for the individual changes included one. Thus the supported result is a reversal caused by web selection, not an established ecological rise or fall. The audit was chosen after a coverage check failed and filter-specific means were seen. We preserve source data, a complete web-support ledger, code, results, checks and repeat instructions. Capture support does not prove trap operation or fixed physical web locations. These ratios do not estimate population density or a causal habitat effect.

Sevilleta · rodents · sampling effort · biodiversity monitoring · data audit · grassland · shrubland · reproducibility