NAD83 vs ITRF: Why Your RTK Coordinates Are 1–2 Meters Off

You tie into a known monument with a fixed RTK solution, centimeter-level RMS values on the screen, and the coordinate still lands a meter or two from the published value. Nothing is broken. Your receiver and the monument's record sheet are almost certainly describing the same point in two different reference frames. Understanding NAD83 vs ITRF is the difference between a clean survey and a mystery bust that eats an afternoon.

What a datum actually is

A datum (more precisely, a reference frame) is the set of conventions that turns "a point on Earth" into numbers: where the origin sits, how the axes are oriented, and which ellipsoid approximates the planet's shape. Latitude, longitude, and height mean nothing without one. The two frames that matter most to US surveyors are:

• NAD83(2011), epoch 2010.00 — the North American datum used by most US survey plans, state plane coordinate systems, and published control. It is fixed to the North American tectonic plate.
• ITRF2020 — the International Terrestrial Reference Frame, a global, Earth-centered frame used by satellite orbit products and global correction services. It is fixed to the planet as a whole, not to any one plate.

In the continental US, coordinates for the same physical point differ between these two frames by roughly 1–2 meters. That is not an error in either frame. It is a deliberate difference in definition, and it is exactly the kind of offset that shows up when your RTK corrections and your project basis disagree.

Plate motion and why epochs matter

The North American plate drifts on the order of centimeters per year. In a global frame like ITRF, that means every monument's coordinates change continuously; a point measured today is not where it was in 2010. To make coordinates usable, ITRF positions are tagged with an epoch: the date the coordinate is valid for.

NAD83 sidesteps most of this by rotating along with the North American plate, so points in the stable interior barely move within the frame. That is convenient for record keeping, but it also means NAD83 has slowly rotated away from the Earth-centered ITRF over the decades. The accumulated difference is that 1–2 meter offset, and it keeps growing a little each year.

The practical rule: a coordinate without a datum and an epoch is incomplete. "NAD83(2011) epoch 2010.00" and "ITRF2020 epoch 2026.5" are full descriptions. "NAD83" alone is not.

Your corrections source decides your frame

Here is the part that catches people. When you work RTK, your rover's coordinates inherit the reference frame of the base or network sending corrections, not whatever your data collector's coordinate system label says. If you are new to how corrections work, start with our RTK GNSS surveying primer.

• State CORS networks in the US typically broadcast coordinates consistent with NAD83(2011). Tie into one of these and your points align with published US control. As of mid-2026, 22 US states run free public RTK/CORS networks; see our guide to free NTRIP casters and CORS networks.
• Global correction and PPP services generally deliver positions in ITRF2020 at the current epoch. So do many community bases, and any base you set up by averaging its own autonomous position.

Mix the two on one project and you get internally consistent points that are systematically shifted 1–2 meters from each other. The RMS on screen stays at a centimeter the whole time, because precision and reference frame are separate questions.

Heights: ellipsoid vs orthometric

Datums have a vertical cousin. GNSS natively measures ellipsoid height, the distance above the mathematical ellipsoid. Plans, pipes, and water all care about orthometric height, elevation relative to the geoid, the gravity-defined surface that approximates mean sea level. The two differ by tens of meters in much of the US.

A geoid model converts between them: GEOID18 is the current US model, and EGM2008 is a common global one. If your elevations look wildly wrong, the first suspect is a missing or mismatched geoid model, not the receiver. The second suspect is a bad antenna height, which deserves its own paranoia (we cover that in our stakeout basics guide).

NATRF2022 and the modernized NSRS

All of this is in the middle of a generational change. The National Geodetic Survey's modernized National Spatial Reference System, including NATRF2022 as the NAD83 replacement, is landing in 2026. The new frames are designed to align closely with the ITRF and to handle time-dependent coordinates explicitly, so the epoch bookkeeping described above becomes a first-class part of US practice rather than fine print. Software and workflows that already record datum and epoch on every point will make that transition without drama; workflows that treat coordinates as timeless numbers will not.

Practical rules for the field

• Know your caster's datum. Before you trust a mountpoint, find out what frame its base coordinates are in. State CORS pages usually document this; community bases often do not, so ask or check against known control.
• One frame per project. Pick the basis your deliverable requires and make every corrections source, control tie, and import agree with it.
• Check into known control first. A two-minute observation on a published monument at the start of the day flags a frame mismatch before it contaminates hundreds of shots.
• Stamp datum and epoch on every deliverable. A CSV of coordinates with no frame metadata is a future dispute waiting to happen.
• Record the geoid model too. Elevations without a named geoid are as ambiguous as coordinates without a datum.

This is also a place where tooling should help instead of hoping you remember. Open Survey is being built to stamp datum and epoch on every export (CSV, GeoJSON, KML/KMZ, DXF, GeoPackage), and to log fix type, RMS, correction age, and antenna height with every observation, so the metadata that resolves a 1–2 meter question is always in the file.