WGS 84 vs NAD83

NAD83 is a plate-fixed datum. It is designed so that a survey monument in Kansas keeps roughly the same coordinates decade after decade, because the frame moves along with the continent. That is exactly what you want for cadastral boundaries and engineering control: the corner of a parcel should not need re-publishing every year.

WGS 84 is a global datum, kept closely aligned to the international terrestrial reference frame. It does not follow any one plate. A point in Kansas therefore drifts within it, by roughly 1 to 2 cm per year.

Forty years of that adds up. The two frames were considered coincident at the metre level when NAD83 was adopted; across the contiguous United States they are now separated by on the order of 1 to 2 metres, and the direction of the offset varies with location. The gap widens every year.

A common misreading is that the difference comes from WGS 84 using the WGS 84 ellipsoid and NAD83 using GRS 80. Those two ellipsoids have the same semi-major axis of 6378137 m and semi-minor axes that differ by about a tenth of a millimetre. That is nothing.

The difference is the datum: where the origin sits, how the axes are oriented, and which reference frame the whole thing is tied to. Two coordinate systems can share an ellipsoid exactly and still place the same point metres apart, because the ellipsoid has been positioned differently.

This matters when choosing a transformation. Swapping the ellipsoid parameters achieves nothing; you need a datum transformation, and for NAD83 the good ones are grid-based.

A GPS or GNSS receiver computes positions in WGS 84 by default, because the broadcast ephemeris is in it. A consumer receiver, a phone and the GPS in a drone all hand you WGS 84.

Nearly all United States survey control is NAD83. State Plane coordinate systems are defined on NAD83. County parcel layers, DOT centrelines, FEMA flood layers and most federal and state agency data are NAD83. A corrected position from a real-time network in the US is typically delivered in NAD83(2011).

So the mismatch usually appears at the seam between field capture and office data: drone imagery geotagged in WGS 84 is dropped onto a NAD83 parcel layer, and everything is offset by a metre or so in a consistent direction. The telltale is that the offset is uniform across the whole site rather than random.

Note also that NAD83 is not one thing. NAD83(1986), NAD83(HARN), NAD83(CORS96), NAD83(NSRS2007) and NAD83(2011) are successive realisations, and they differ from each other by decimetres or more. A file labelled only 'NAD83' is under-specified. Ask which realisation, and at which epoch.

It does not matter for navigation, for a web map, for showing where a photo was taken, or for any product whose tolerance is larger than a few metres. Treating the two as equal is a long-standing and generally harmless shortcut at that scale.

It matters for anything cadastral, for engineering layout, for volume calculations tied to survey control, for utility locates, and for any deliverable that will be compared against existing NAD83 control. A metre of systematic shift is far larger than the centimetre-level accuracy an RTK drone survey is sold on, so a correctly flown RTK mission reported in the wrong datum throws away the entire benefit of flying it.

It also matters cumulatively. Mixing datums inside one project produces a dataset where some layers agree and others are offset, and the inconsistency is harder to find later than it would have been to prevent.

A seven-parameter Helmert transformation gets you to within decimetres, which is enough for mapping work. For survey-grade results in the United States the National Geodetic Survey publishes grid-based transformations, which model the distortion properly rather than fitting one rigid shift to a whole continent.

Those grid files are large binaries, and they are the reason browser-based conversion has a ceiling. This site's converter runs entirely in your browser with no upload, which means it uses proj4 definitions and cannot apply an NGS grid shift. Expect metre-level agreement from it, not centimetre-level, and use NGS tooling or a full GIS with the grids installed when the tolerance is tighter. That limitation is stated on the about page rather than hidden.

Whatever you use, record the datum and the realisation in the deliverable. The single most expensive datum error is the one nobody wrote down.

The National Geodetic Survey has been working for years on a modernised reference frame to replace NAD83, together with a new geopotential datum to replace NAVD88. The published timetable has moved more than once, so treat any specific date you read with suspicion and check the NGS site directly.

The design intent is worth knowing even while the schedule is uncertain: the replacement is explicitly tied to the global frame with plate motion modelled, rather than fixed to the plate. When it lands, the gap between the North American datum and the global one stops silently widening, and coordinates acquire an epoch as a matter of course.

Until then, the practical advice does not change. Know which datum your data is in, do not assume WGS 84 and NAD83 agree, and write the answer down.