Early fish catch scores gave uncertain spatial error gains at 27 Virginia lagoon sites in 2023–2026

We tested whether early June fish catch patterns transferred to later surveys at 27 named Virginia lagoon sites. The source was fixed EDI version knb-lter-vcr.367.12. The balanced panel had 189 hauls and 13,076 fish in 2019 and 2021–2026. We removed one marked test record and one failed first haul. We centered log(1 + fish count) within each survey area and year. Each site score was the mean from 2019, 2021 and 2022. We held these scores fixed in 2023–2026. They reduced squared error by 6.19% against the later area-year mean. The descriptive 95% site-and-year resample range was −148.99% to +37.37%. Early and later mean scores had Spearman correlation 0.686. The within-area label reference probability for the error score was 0.0003. Thus, the fixed-panel rank link did not establish a reliable error gain. Error increased in two of four later years. A test through 2025 gave a 5.97% error increase. Raw-count scores performed much worse. The test uses later area-year means and is not an operational count forecast. It applies to named stations, not exact fixed locations. Source coordinates include two positive-longitude records, other position differences and missing values. The data do not establish restoration or climate effects. The package has unchanged source files, provenance, all counts and scores, figures, random draws, code and independent checks.

fish · seagrass · spatial transfer · biodiversity monitoring · Virginia coastal lagoons · reproducible analysis · prediction error