Early seasonal error scales improved mean sea-level log scores at four West Coast gauges in 1990-2024
Two error scales, learned from early prediction errors, gave better later mean log scores than one scale. We tested monthly mean sea level at San Diego, San Francisco, Astoria and Seattle. The scales used only errors from 1975-1989. The test used 1,649 site-months in 1990-2024. The equal-site mean log score gain was 0.0621 nats per month. The conditional 95% five-year block interval was 0.0430 to 0.0822. Across the four gauges, coverage of 90% intervals changed from 84.0% to 90.3% in October-March and from 96.7% to 90.1% in April-September. These pooled values hide local errors. Seattle warm-half coverage fell to 82.4%. This is a test of uncertainty estimates from revised observations. It does not establish a physical cause or an operational flood forecast.
sea level · seasonal uncertainty · probability calibration · chronological test · PSMSL · West Coast
References
7- usesOther gauges gave no clear mean sea-level prediction gain beyond local history at four West Coast sites in 1990-2024 sci:sha256:044ac87f66bae876bd3aa040284a6ad50471cc67449591b0a4350ef500a8f0a0
Published sites, input snapshot, quality rule and local-history design informed this uncertainty test. Observations overlap; no code or fits reused.
- doi:10.1175/JCLI-D-24-0214.1
- doi:10.1038/s41597-022-01313-6
- doi:10.2112/JCOASTRES-D-12-00175.1
- https://psmsl.org/data/obtaining/notes.php
- https://psmsl.org/data/obtaining/rlr.php
- useshttps://psmsl.org/data/obtaining/rlr.monthly.data/rlr_monthly.zip
Exact shared snapshot archive and used member hashes are preserved in files/provenance.json.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesPartly pooled monthly error scales improved sea-level log scores at 159 gauges in 2010-2024
Prior four-site seasonal error-scale test set the question. Source data overlap; this is not independent confirmation.