The Tide Gauge Reads Today. Its Datum Averages the Years 1983 to 2001.

A water level sensor at a federal tide station writes down a height at set intervals. That height is not an elevation of anything; it is a distance from the sensor to the water. It becomes an elevation only once it is referenced to something fixed, and the fixed thing is not an object. It is an average.

The average has a name and a length. NOAA calls it the National Tidal Datum Epoch, and the NOAA Tides and Currents datum reference defines it as "The specific 19-year period adopted by the National Ocean Service as the official time segment over which tide observations are taken and reduced to obtain mean values for tidal datums." The period currently in force runs from 1983 to 2001.

That is the whole of it. The observation is from this morning; the frame it is quoted against was assembled from water that moved through a window which closed a quarter of a century ago. Nothing is hidden — NOAA prints the epoch beside the values. But the arrangement reaches flood thresholds, nautical charts and marine boundaries, and is easier to see once the assembly order is laid out.

Figure 1. How a tide station height becomes a published elevation. Stage three carries the date.

The Chain

1. A sensor records. Stations in the National Water Level Observation Network report water levels continuously — measurement in the ordinary sense: an instrument, a reading, a timestamp.

2. The reading is tied to land. A height above a sensor means nothing unless the sensor's own position is known, so CO-OPS maintains a bench mark sheet beside the tidal datums at each location. That is what keeps a datum attached to a place rather than to a piece of equipment.

3. The readings are reduced over a fixed window. Here the date enters. Mean Higher High Water, in NOAA's definition, is "The average of the higher high water height of each tidal day observed over the National Tidal Datum Epoch." Mean Lower Low Water is the mirror image, and both definitions end on the same words. The quantity is an average, and the averaging window is the epoch.

4. A datum exists, and elevations can be quoted. Only now is there something to measure against. NOAA's fact sheet on the epoch lists what hangs off this step: "ship navigation, storm hazard mitigation, coastal engineering, marine boundary determination, and beach renourishment."

5. Everything downstream inherits the window. A chart sounding, a property line at the water's edge and a coastal flood threshold are all expressed against a datum produced in stage three. Rebuild stage three and every number expressed against it is restated, whether or not any water has changed.

6. The window does not move on its own. It is replaced, deliberately, on a schedule. Between replacements the reference is frozen while the thing it describes is not.

The shape is familiar from river discharge, where gage height is measured while the streamflow beside it is read off a curve. Here the derivation runs through a calendar rather than a rating curve.

Why the Window Is Nineteen Years Long

The length is not a round number chosen for convenience. NOAA's NOAA National Tidal Datum Epoch fact sheet explains that the period "represents one full rotation of the Earth's longest [18.6 years] lunar cycle, which influences the world's tides," and that "NOAA rounds to 19 years to account for seasonal and environmental trends (e.g. sea level change), and determines the truest average of tidal conditions."

NOAA Technical Report NOS CO-OPS 068 attributes the cycle to "an 18.6 year repeat cycle caused by declinational changes in the plane of the moon's orbit called the regression of the Moon's nodes," which produces "slowly varying cycles in the mean range of tide." A shorter window would catch part of that cycle and miss the rest, so two stations averaged over different parts would not be comparable.

Comparability is the fact sheet's stated reason for one national window rather than a window per station: "By using a common cycle for the country's tidal datums, NOAA ensures a nationally consistent tidal datum network from Maine to Hawaii." The network covers "23 States & 14 Territories" with "130 Primary Stations," and CO-OPS maintains datums and bench mark sheets "for 2,000+ locations under NTDE."

Not all of those locations have nineteen years of their own data. CO-OPS "relies on control (tide/water level) stations to infer what a 19-year average would look like at non-control stations," a result the fact sheet calls "a 19-year NTDE equivalent datum," so that all datums "incorporate data from the same epoch, even if actual observations come from another time period." At stations without nineteen years of their own record, then, the published datum is not an average of that station's own nineteen years. It is an inference about what such an average would have been.

What Triggers a New Window

The epoch is replaced, not slid forward. The fact sheet gives one condition: "An NTDE update is triggered when the change in Mean Sea Level at the Nation's coastal tide gauges averages above 0.10ft (0.03m)." That is a national average across gauges, not a local test, and a single criterion rather than a choice among several.

How often that has happened is a matter of record rather than rule: "Historically, this has resulted in the NTDE being updated every 25-30 years." The fact sheet adds that "long-term sea level trends have shown accelerating rates of sea level rise in many regions (but not all) of the country," which "indicates that the current NTDE update schedule may need to be accelerated."

The replacement now in preparation is NTDE 2002-2020. The fact sheet states the timing directly: "The current NTDE spans 1983-2001. In 2026 it will be replaced by NTDE 2002-2020." NOAA's NOAA datum updates page for the epoch describes the same transition to data "spanning the years 2002-2020" but puts the new products after 2026. The two pages are not saying the same thing about the date, and anyone relying on a specific arrival should read both.

The manner of arrival is not ambiguous. CO-OPS "will recompute over 2,000 tidal datums and bench mark sheets," and "All new datums and bench mark sheets will be released online at one time (i.e. NTDE 2002-2020 will not be released incrementally by region)." New sheets "will supersede those of NTDE 1983-2001," while "Data for prior Epochs will remain accessible on CO-OPS' website for users." One release date for the whole country keeps the network coherent, and makes the transition abrupt: there is no period in which half the coast is on the new epoch and half on the old.

A Second Procedure, Standing Beside the First

A national window has an obvious weak point: places where the land itself is moving fast. NOAA handles those separately, and the separation matters. The fact sheet points readers to the modified datums procedures with the explicit qualification that they are "distinct from NTDE" — not a regional flavour of the epoch, but a different procedure running alongside it.

The datum reference page gives the scope: "Tidal datums in certain regions with anomalous sea level changes (Alaska, U.S. Gulf Coast) are calculated on a Modified 5-Year Epoch." CO-OPS 068 names the areas as "central Louisiana, the southern Cook Inlet, and the southeastern Alaska coasts."

The substitution is partial, which is the detail most easily lost. Under the modified procedure, "for the mean tide level datums, a shorter, more recent 5 year computational period is used to better reflect the current local elevations relative to the land," while "the 19-year NTDE computational period for the mean ranges of tide was used." The report writes Mean Higher High Water as DTL5 + 1/2 Gt19 — the centre of the tide from five years, the spread of it from nineteen. Such a datum is a composite of two time spans, not a move of the whole calculation to a newer one.

Figure 2. The two numerical tests in NOAA Technical Report NOS CO-OPS 068, and what the report says follows from meeting them.

Two Tests, and What Meeting Them Actually Does

CO-OPS 068 states the primary criterion as a rate: "The Modified Procedure process should only be used for those stations with dominant vertical land movement as evidenced by relative sea level trends of >= 9.0 mm/yr." A second sentence adds a size test beside it: "Stations showing a relative sea level trend >9mm/yr, and stations showing a difference in MSL between the currently accepted datum and the new 5-year preliminary datum computation greater than 0.05m, should be reviewed in detail."

Two things there are worth reading slowly. The sentence names two groups of stations — those above the rate and those above the difference — and sends each of them to review. And the consequence is review: the sentence does not say such a station is moved to the shorter period, or must be. Between a numerical flag and a changed datum sits a judgement, and the report does not collapse the two.

The report is equally explicit that the tests do not carry equal weight, and says why: "More weight is given to the sea level trend threshold of 9mm/yr than the MSL datum computation difference of 0.05m as there can be some natural variation that may cause datum differences greater than the 0.05m in the short term that are not considered a long term trend."

That sentence is the key to the tables below. A large gap between two computations can be noise; a sustained rate is what the procedure is built to catch.

Reading the Two Tables Against Each Other

CO-OPS 068 prints two tables of the same quantity: mean sea level at a station computed over the 1983-2001 epoch and over a five-year period of 2007-2011, the difference between them, and the station's relative sea level trend. One table holds the stations moved to the modified procedure, the other comparison stations that stayed on the national epoch.

Figure 3. Relative sea level trend and mean sea level difference at twelve stations. Source: NOAA Technical Report NOS CO-OPS 068, Table 1 (Modified Procedure update stations) and Table 2 (comparison stations). Difference values are the report's own; each reproduces as the 2007-2011 value minus the 1983-2001 value.

Side by side, the two tables separate cleanly at the stated rate. The six stations in the modified table carry trends of 17.12, 12.92, 11.54, 10.42, 9.45 and 9.24 mm/yr in magnitude — every one at or above 9.0. The largest magnitude anywhere in the comparison table is 6.39 mm/yr at Pier 21, Texas. No overlap, no borderline case.

The difference column does not separate the same way, and that is the point. Grand Isle, Louisiana, which was moved, differs by +0.105 m between the two computations. Valdez, Alaska, which was not, differs by -0.159 m; Rockport, Texas, also not moved, by +0.136 m. Both stations that stayed on the nineteen-year epoch show a larger gap than the one that left it.

Without the weighting sentence that looks like an inconsistency. With it, it is the procedure working as written: Valdez has a trend of -4.92 mm/yr and Rockport 5.16 mm/yr, both well under the rate threshold, so their larger short-term differences are the kind of variation the report says should not by itself move a station. The rate test decides; the size of the gap, alone, does not.

This is a recurring shape in federal measurement: a published figure can move because the quantity moved or because the thing it is expressed against moved, and the number alone does not say which — the same reason a cell sites out of service percentage shifts when the reporting area does. Here the reference is a date range, printed beside the value.

What the Change Is Worth in Feet

NOAA's NOAA Tides and Currents page on new tidal datum updates describes a later application of the procedure, using a five-year period of 2012-2016 rather than the 2007-2011 span documented in CO-OPS 068. It lists six control stations receiving modified datums — Grand Isle, Louisiana, and Juneau, Skagway, Seldovia, Kodiak and Yakutat in Alaska — the same six that appear in the technical report's modified table.

The magnitudes are given as "~0.03m (~0.10 foot) to ~0.16m (0.5 foot)." The same page states that no changes are required to NOAA's tidal prediction products, and describes chart accuracy as unaffected at standard scales, while noting that changes may be required for regions with rapid land movement.

Scale decides whether a few inches matter. For a ship's keel clearance, a tenth of a foot is nothing. For a boundary drawn at the water's edge, or a flood threshold a tide either crosses or does not, it is the difference between two counts. Those thresholds are a separate product with separate custody: NOAA's annual high tide flooding outlook states that "Flooding thresholds are supplied by NOAA's Office for Coastal Management." The datums they are expressed against come from the epoch; the thresholds themselves do not.

How much of an elevation is local shows in two stations from the fact sheet: Mean Lower Low Water sits at -0.82ft/-0.251m relative to Mean Sea Level at Bayou La Batre Bridge, Alabama, and at -3.03ft/-0.925m at North Channel, New York — same datum name, same epoch, more than two feet apart. The epoch fixes when the averaging happens, not what it yields.

Where This Doesn't Apply

The epoch governs tidal datums, and tidal datums only. An elevation quoted against a geodetic vertical datum is a different reference built a different way, and nothing in the epoch schedule moves it.

It does not apply to the water level observations themselves. A recorded height is what it is; the epoch changes what it is reported relative to, not what the sensor saw.

It does not apply to tide predictions in the way one might expect. NOAA's page on the modified datum updates states that no changes are required to its tidal prediction products, so a shifted datum does not automatically mean a shifted prediction table.

It does not apply to inland flood forecasting, where river stages are quoted against local gage datums and flood categories set by forecast offices, with no tidal epoch involved.

The modified five-year procedure does not apply broadly. It is limited, in the report's words, to stations "with documented anomalous relative sea level trends due to high rates of vertical land motion," and the technical report's table of them runs to six. Most of the network is on the nineteen-year epoch, and a station in a region that contains modified stations is not thereby a modified station.

None of this establishes that any published figure is wrong. A datum computed over 1983-2001 is a correct answer to the question the epoch asks. The limit is narrower: a coastal elevation is a statement about a period as much as about a place, and when the period is replaced the statement is restated across the country at once, on a day set by a release schedule rather than by the sea.