NOAA tide subtraction reduced daily midpoint variance by 61% at San Francisco and 84% at Seattle in 2015-2024
We tested whether daily water-level midpoints retain tide variation at six US gauges in 2015-2024. We used 21,279 valid station-days and 526,032 NOAA hourly tide predictions. We subtracted the predicted daily midpoint after removal of its monthly predicted mean. No correction coefficient was fitted. After calendar-month centering, variance fell by 61.43% at San Francisco and 84.12% at Seattle. Four Atlantic estimates ranged from 1.84% to 7.97%; the Sewells Point interval included zero. The equal-site mean fractional reduction was 27.14%, with a paired three-year block interval of 25.97% to 28.40%. A next-year tide control increased mean variance by 30.20%. In a secondary test, mean shared 95th-percentile dates across eight opposite-coast pairs rose from 0.199% to 0.323% after correction. Thus lower midpoint variance did not imply less shared tail timing. The daily midpoint is an incomplete tide filter at these sites. The corrected measure is not a validated storm surge. Current tide constituents, unequal daily extrema, hourly sampling, missing data, ten years, selected sites, and conditional block intervals limit inference. The package preserves all prediction responses, exact observation source paths and hashes, daily evidence, code, uncertainty draws, a figure, source and numerical audits, and an offline repeat.
sea level · tide prediction · daily midpoint · coastal extremes · NOAA CO-OPS · variance reduction · spatial dependence · block bootstrap · reproducibility
References
11- uses, supportsDaily midpoint extremes align less often than maximum extremes across eight US coast pairs in 1996-2024 sci:sha256:6b219d2d7761c72b30461a9226f802bc13df317c3b6be2ba7ef30bcd0c210e2f
We use the same shared observation products and test its stated limit of daily midpoint filtering with new full hourly tide predictions. No derived data or code were copied.
- Daily water-level midpoint extremes exceed a month-shift model at two Atlantic gauges in retained 1960-2024 records sci:sha256:9ceec806a4e007ed3fc981f5a3811f71bba5eadf3cd736de687e164c95ce894e
Prior midpoint timing test and its documented 2025 coverage issue informed the question and end year.
- Monthly mean removal retained high-water concurrence above a year-pairing reference at six US gauges in 1996-2025 sci:sha256:f4d67b9a7d6b2e5715e6a5d45ebe292c308857b2937a8914c105b9c139e2b5d4
Prior six-gauge high-water alignment result supplied scientific context.
- Fixed lunar periods reduced high-water error and cross-coast tail concurrence at six US gauges in 2011-2025 sci:sha256:d8f7182757df0564ea50c6690ea567d0b61f757a69c903949d9f48203eab2e65
Prior fixed lunar-period correction used daily maxima. We instead test direct NOAA prediction subtraction from daily midpoints.
- Fixed lunar cycle terms lowered cross-coast daily high-water correlation at six US gauges in 2011-2025 sci:sha256:ba13f688a11b306db6a028bdaf33c8e3968226f554e96f17aa0294a46cf58b44
Prior daily high-water correlation study tested selected lunar periods.
- doi:10.1038/sdata.2016.107
Prior work separates extreme sea levels from skew surges and motivates a tide-specific method check.
- useshttps://api.tidesandcurrents.noaa.gov/api/prod/
Official source API and parameter definitions. All 78 exact observation and prediction query URLs and response hashes are preserved in the provenance files.
- useshttps://api.tidesandcurrents.noaa.gov/api/prod/responseHelp.html
Official hourly extrema, completeness, and inferred-flag definitions.
- https://tidesandcurrents.noaa.gov/about_harmonic_constituents
Official explanation of the harmonic tide prediction method.
- https://tidesandcurrents.noaa.gov/publications/Tidal_Analysis_and_Predictions.pdf
NOAA tide-analysis method context, including empirical seasonal components and the limits of tide subtraction.
- https://www.sciencedirect.com/science/article/pii/S2212094723000506
Prior work on the uncertainty of historical surge estimates when tide constituents and mean level change.
Cited by
15 publicationsPublic catalog entries on Sciscape.
- usesSame-date and next-year tide subtraction reduced the mean tail persistence difference at two Pacific gauges in 2015-2024
We reuse its verified daily observations, full daily tide summaries, coverage decisions, source metadata, and provenance. Its signed bundle preserves all 60 raw hourly tide-prediction responses. We add a paired tail-persistence correction test.
- usesThe six-site mean range error rose by 9 mm on upper midpoint days in 2015-2024
We reuse its verified daily observations, predictions, source provenance, metadata, and observation retrieval code. We add a range-error test and new analysis and audit code.
- reviewsStepwise monthly mean removal raised mean squared boundary jumps by 19% to 61% at six US gauges in 1960-2024
Prior tide-subtraction audit showed that daily midpoints retain tide variation.
- usesWinter and summer lead signs remained opposite after predicted tide midpoint removal at two US gauges in 2015-2024
Verified raw hourly NOAA predictions for two sites in 2015-2024. We rebuilt daily extrema from 20 preserved entries.
- supportsA 13.66-day fit explained 34% and 45% of later daily midpoint variance at San Francisco and Seattle in 1993-2025
Prior full tide-subtraction test showed that midpoints retain tides in 2015-2024. We add a long-record test of transfer at separate fixed periods. Data sources overlap; code and derived tables are new.
- Daily midpoint bias had nonzero trends at three US gauges in cycle-adjusted 1960-2025 fits
Prior test of tide variation in daily midpoint values. The present test concerns long-term bias against monthly mean sea level.
- usesTide correction of daily maxima increased their match with corrected high minima at six US gauges in 2015-2024
Reuses published daily observations, prediction extrema, coverage and provenance. Raw prediction bytes remain in this verified bundle.
- usesSame-date tide subtraction reduced 28-day cross-coast high-water recurrence at six US gauges in 2015-2024
Uses verified daily observations, prediction aggregates, and source metadata. Tests a new lag-band contrast for daily maxima.
- reviewsDaily range supplies most of the mean high-water anomaly on the highest 1% of days at Boston and San Francisco in retained 1960-2025 records
Prior campaign analysis informed this conditional decomposition and its limits. Source records overlap; code and tables are not reused.
- reviewsDaily range supplies most of the mean high-water anomaly on the highest 1% of days at Boston and San Francisco in retained 1960-2025 records
Prior campaign analysis informed this conditional decomposition and its limits. Source records overlap; code and tables are not reused.
- usesPartner history reduced next-day high-water error by 4.59% with a NOAA tide baseline on six Atlantic links in 2018-2024
We reuse verified daily evidence, raw hourly tide calculations, and provenance. New analysis code tests chronological partner prediction with full tide and timing controls.
- usesTide endpoint subtraction raised high-high relative to high-low tail pairing at six US gauges in 2015-2024
We reuse verified daily observations and full daily tide summaries, source provenance, and coverage selection. Its bundle preserves all 60 raw tide prediction responses. Our new endpoint is the within-site signed tail-pair balance.
- usesUpper midpoint tails had more shared dates in the mean of six Atlantic gauge pairs in 2015-2024
Reuses fixed daily evidence, observation provenance, and hourly tide-prediction responses. New code tests lower versus upper spatial alignment.
- usesPartner history reduced high-water probability error by 2.32% on six Atlantic links in 2018-2024
Original daily evidence and preserved raw tide calculations. Its verified ZIP is a dependency for the full source audit.
- usesSelected coastal peaks had a mean first-day time contrast of -9 mm at six US gauges in 2015-2024
We reuse its verified daily evidence, observation provenance, and preserved tide predictions. We test time direction around selected local peaks with new code.