Datums & heights · 4 min read

NGVD 29 and NAVD 88: the vertical datums and the difference between them

Two national vertical datums built on different principles. The difference varies across the country from roughly −0.4 m to +1.5 m, so no single conversion exists.

The United States has used two national vertical datums in the modern era. NGVD 29 held a set of tide gauges fixed at local mean sea level. NAVD 88 abandoned that idea and held a single benchmark. The change was not cosmetic: it altered the elevation of every benchmark in the country, by an amount that varies systematically from place to place.

Anyone working with older drawings, floodplain data or utility records will meet both. The difference is large enough to matter for drainage and flood work and variable enough that a single conversion number is always wrong somewhere.

NGVD 29

The National Geodetic Vertical Datum of 1929 was originally called the Sea Level Datum of 1929 and renamed in 1973, which tells you most of what went wrong with it. It was created by a general adjustment of the first-order levelling network in which 26 tide gauges — 21 in the United States and 5 in Canada — were each held at their locally determined mean sea level.

The flaw is that local mean sea level is not the same equipotential surface at every gauge. Ocean currents, water temperature, salinity and prevailing air pressure make the sea surface depart from the geoid by up to a meter or two, and that departure differs between, say, the Gulf coast and the Pacific northwest. Forcing all 26 gauges to zero built those differences into the datum as a permanent warp, and the levelling network was distorted to absorb them.

A secondary problem accumulated over the following decades: subsidence, uplift and disturbed benchmarks meant that many published NGVD 29 elevations no longer described the marks they named.

The North American Vertical Datum of 1988 took the opposite approach. It is a minimally constrained adjustment of the levelling networks of the United States, Canada and Mexico, holding exactly one point fixed: the primary tide gauge benchmark at Father Point in Rimouski, Quebec, on the St Lawrence.

Holding one point rather than 26 means the internal geometry of the levelling network is preserved instead of being forced. Relative heights between benchmarks are far better than under NGVD 29. The adjustment also used Helmert orthometric heights, which properly account for gravity variation along the plumb line, where NGVD 29 used a simpler treatment.

The cost is that the datum has no particular relationship to the geoid or to mean sea level anywhere except that one point. NAVD 88 is offset from the global geoid by roughly half a meter and carries a tilt of about a meter across the continent, west to east. For measuring height differences that does not matter. For tying to a global vertical reference, it matters a great deal, and it is one of the reasons the datum is being replaced.

The difference between them

NGVD 29 and NAVD 88 compared
PropertyNGVD 29NAVD 88
Constraint26 tide gauges held at local mean sea levelOne benchmark held, at Father Point, Rimouski
Adjustment typeConstrained, network forced to fit the gaugesMinimally constrained, network geometry preserved
Height typeNormal orthometric, simplified gravity treatmentHelmert orthometric heights
CoverageUnited States and CanadaUnited States, Canada and Mexico
Main weaknessDistortion from sea surface topography at the gaugesRoughly half a meter of bias and about a meter of continental tilt
StatusSupersededBeing replaced by the modernized geopotential datum

The numerical difference between the two datums across the conterminous United States runs from roughly −0.4 m to about +1.5 m. It is smallest and most negative in the northwest, and largest and most positive around the Gulf and the southeast. Within any one project the difference is nearly constant; across a state it is not.

Why this bites hardest in flood work

Floodplain mapping, base flood elevations, drainage design and elevation certificates all live at the decimetre level, and older studies were published on NGVD 29 while current work is on NAVD 88. A base flood elevation transcribed from an old study onto a new drawing without conversion is wrong by up to about a foot and a half, in the direction that matters.

The same applies to sanitary and storm systems. Invert elevations recorded decades ago on NGVD 29 and compared against new survey on NAVD 88 will show a phantom slope or a phantom bellied line where none exists.

Practical handling

  1. Label the vertical datum on every drawing, every profile and every elevation certificate. Two decimal places of elevation with no datum name is not information.
  2. When a record predates about 1991, assume NGVD 29 until you can confirm otherwise, then confirm it.
  3. Convert with VERTCON, not with a constant from another county.
  4. State the conversion on the drawing: source datum, target datum, method, and the value applied at the project.
  5. Where possible, level to a recovered benchmark and adopt its published NAVD 88 elevation rather than converting an old value at all.

The last item is the one that most often gets skipped and most often would have saved the project. A conversion is a model. A recovered mark with a published height is an observation. Prefer the observation.

What replaces NAVD 88

NAVD 88 is being replaced by the North American-Pacific Geopotential Datum of 2022, a geopotential datum defined by a gravimetric geoid model rather than by a network of levelled benchmarks. That change is in progress. Its practical effect is that an orthometric height will be obtainable from a GNSS position plus a published geoid model, without needing to run levels from a benchmark that may have moved since it was last observed. Expect published elevations to change again when it arrives, by amounts comparable to the bias and tilt described above.

Questions

How much is the difference between NGVD 29 and NAVD 88?

It varies across the conterminous United States from roughly −0.4 m to about +1.5 m, and the pattern is systematic rather than random. Use VERTCON to get the value at your location; do not apply a constant borrowed from elsewhere.

Which datum is a 1970s drawing likely to be on?

Almost certainly NGVD 29, since NAVD 88 was not available until around 1991. Treat the assumption as a hypothesis to be verified against a recovered benchmark rather than as a fact, because some older drawings use a purely local assumed datum with no national tie at all.

Why does NAVD 88 not match mean sea level at the coast?

Because it was never constrained to sea level except at one point in Quebec. NAVD 88 preserves the geometry of the levelling network rather than forcing it to fit tide gauges, so coastal NAVD 88 elevations differ from local mean sea level by varying amounts.

Can I just use my GNSS receiver to get NAVD 88 elevations?

Yes, with a hybrid geoid model, and that is standard practice. The models are built specifically so that a GNSS ellipsoid height minus the modelled geoid height reproduces published NAVD 88 elevations. Hold local benchmarks to absorb residual model and datum error.

Sources