Time-varying sampling effort shapes detected marine biodiversity trends and misleads global conservation priorities: code and derived data products

v2 (2026-09-28). Added R/, a minimum working example in R (base R + stats only, no CRAN packages): run_reproduction.R, R_functions.R, renv.lock, README_R.md. It reads the published products under data/ and re-derives the panel sizes, the M0/M1/M2 trend coefficients with CR1 cluster-robust standard errors (cell and 10-degree-region clustering), both pooled growth slopes, the algebraic identity b0 = b1 + c*delta, the Wilson score intervals, and the Kendall tau / Spearman / top-500 rank statistics, asserting equality against the stored JSON (47 checks, all passing). The snapshot download and aggregation step remains Python-only and is out of scope for the R package. No changes to scripts/, data/ or figures/.Code and derived data products for the manuscript of the same title. Source: Ocean Biodiversity Information System (OBIS) snapshot of 18 March 2025 (130,049,900 occurrence records; 5,207 datasets). The package contains the full pipeline (snapshot download, event-key aggregation, trend models M0/M1/M2, injection simulation, robustness checks, figures), the derived grid-by-year panel of sampling events and detected richness, robustness and simulation outputs, and publication figures. The OBIS snapshot itself is openly available from the OBIS open-data bucket and is redistributed under OBIS and dataset-level licences.