ITRF, Epochs and Plate Motion: Why Modern Coordinates Drift
7 min read
ITRF2020, ETRS89, GDA2020, epoch 2010.0 — modern geodesy attaches dates to coordinates because the ground itself moves. Here is what reference frames and epochs actually mean, and when you must care.
Classical datums assumed the ground stands still. It does not: continental plates move 2–10 cm every year, and modern GNSS measures positions precisely enough to watch it happen. That is why current systems are reference *frames* with *epochs* — coordinates stamped with the date they were valid.
Frames: ITRF and its national offspring
The International Terrestrial Reference Frame (ITRF) is the scientific standard — a global, earth-centred frame computed from VLBI, SLR, GNSS and DORIS observations, updated every few years (ITRF2008, ITRF2014, ITRF2020). WGS 84 as broadcast by GPS is aligned to ITRF at the few-centimetre level.
In a pure ITRF frame, *your house has a velocity*. A surveyed corner in Europe moves ~2.5 cm/year northeast; in Australia ~7 cm/year. Storing raw ITRF coordinates means your cadastre slowly slides out from under its own data.
National frames solve this by fixing the frame to the plate:
- ETRS89 (Europe): coincided with ITRF at epoch 1989.0, then moves with the Eurasian plate. European coordinates stay put; the offset to ITRF/WGS 84 grows ~2.5 cm/year (about 0.9 m by now).
- GDA2020 (Australia): fixed at epoch 2020.0. Its predecessor GDA94 had drifted 1.8 m from the global frame — the reason Australia "moved" in 2017 headlines.
- NAD83 (North America): plate-fixed since 1986; now offset from ITRF by up to ~2 m depending on location.
- PNG94, SIRGAS, KSA-GRF17, QazTRF-23 — the same pattern worldwide: ITRF realisation, fixed epoch, plate-fixed evolution.
Epochs and time-dependent transformations
Converting between a plate-fixed frame and ITRF exactly requires 14 parameters: the usual 7 Helmert terms plus their *rates of change*, evaluated at your coordinate epoch. The EPSG dataset publishes these time-dependent transformations, and the transformation pages on this site show the rate terms and reference epoch wherever they exist.
A practical reading of a time-dependent operation: the further your data's epoch is from the transformation's reference epoch, the more the rate terms contribute. Ignore them and you inherit centimetres-to-decimetres of error per decade.
When you must care — and when not
You can ignore all of this when target accuracy is worse than ~1 m: web mapping, most GIS analysis, navigation. Treat ETRS89 = WGS 84 = ITRF and move on.
You must care below ~0.5 m, or when:
- mixing precise GNSS (PPP/RTK, which delivers current-epoch ITRF) with national control networks,
- comparing surveys taken years apart,
- working in fast-moving or deforming zones (Australia, PNG, Japan, plate boundaries),
- writing deliverable specs — always name frame *and* epoch ("ITRF2020 @ 2024.5", "ETRS89/ETRF2000").
The cheapest insurance is metadata discipline: record frame, realisation and epoch for every dataset, and let PROJ-based tooling apply the published time-dependent operations.
Related coordinate reference systems
Frequently asked questions
- What does an epoch like 2010.0 mean?
- A decimal year stating when the coordinates were valid: 2010.0 is 1 January 2010. In moving frames the same physical point has different coordinates at different epochs, so the date is part of the coordinate.
- Is WGS 84 plate-fixed like ETRS89?
- No. WGS 84 tracks the global ITRF, so points on every continent drift within it. ETRS89, NAD83, GDA2020 and similar frames are pinned to their plates so national coordinates stay stable.
- What is a 14-parameter transformation?
- A Helmert transformation whose 7 parameters (3 translations, 3 rotations, scale) each have a rate of change per year, plus a reference epoch. It converts exactly between frames that move relative to each other.