How to Set Up a GPS Base Station Without a Known Point

No benchmark, no monument, no CORS — four ways to get trustworthy base station coordinates on unknown ground, from averaging and OPUS to PPP self-convergence, with the accuracy each method actually delivers.

The situation every field crew eventually meets

The site has no survey monument, the nearest CORS is a province away, and the client still expects centimetres. This is the single most-asked question in RTK fieldwork — how do I get base station coordinates on unknown ground? — and the answer matters more than any hardware choice, because every rover, drone geotag, and derived surface inherits whatever the base believes about itself (why inheritance is total).

Method 1: average a single-point position

Every receiver offers it — Emlid Flow’s average-in, DJI’s single-point mode, the “here” button on survey controllers. The receiver averages its autonomous position for a set period and adopts the result. Honest numbers: expect the averaged position to sit anywhere from a few decimetres to a couple of metres from truth (DJI specifies 1.5 m horizontal for an uncalibrated D-RTK single point), regardless of averaging duration — autonomous GNSS errors are correlated, so ten minutes barely beats two. What you get is excellent relative accuracy within today’s survey and a dataset that will not overlay anything else. Legitimate for: one-off, self-contained jobs that will never be compared, imported, or revisited.

Method 2: transfer coordinates with a rover

If network RTK coverage exists anywhere reachable, occupy a temporary point with a rover on the network, mark it, and set the base over it — a do-it-yourself known point. Accuracy inherits the network solution (centimetre inside coverage) plus your marking discipline. Costs: a subscription, coverage at the transfer location, and the transfer errand itself. This is also exactly what “calibrating a D-RTK 3 against network RTK” does — the method wears many brand names.

Method 3: log statics and post-process (OPUS / AUSPOS / CSRS-PPP)

Log raw observations for two to six hours, submit the RINEX to a free national service — OPUS (US), AUSPOS (Australia), NRCan CSRS-PPP (global) — and receive authoritative centimetre coordinates by email. Strengths: free, defensible, jurisdiction-blessed, and it doubles as an audit trail. Costs: hours of occupation before work can be referenced absolutely, internet to submit, and a same-day answer only if you started very early. The classic pattern is retroactive: fly today against a logged base, apply the OPUS coordinates in post — workable, but it delays every real-time consumer (machine control cannot wait for tomorrow’s email). Details of the logging setup: RINEX and PPK.

Method 4: PPP self-convergence — the method that removes the problem

A self-converging base computes its own absolute coordinates from L-Band satellite corrections: power on, ~3 minutes under open sky, ~1.5 cm horizontal / 3 cm vertical in ITRF2020 — no monument, no subscription, no transfer errand, no waiting for email (the full mechanics). The coordinates repeat visit after visit because they are anchored to the global frame, not to a peg; the base can sit anywhere with sky, and moving it changes nothing downstream. For unknown ground this is not one option among four so much as the category answer the other three approximate.

Side-by-side: what each method really delivers

Method	Absolute accuracy	Time to work	Needs	Repeatable across visits
Averaged single point	0.3–2 m	2–15 min	Nothing	No — new offset each visit
Rover transfer from network	~cm (inside coverage)	Errand + setup	Subscription, coverage	Yes, if the mark survives
Static + OPUS/AUSPOS	~cm, authoritative	2–6 h + processing	Internet to submit	Yes
PPP self-convergence	~1.5 cm H / 3 cm V	~3 min	Open sky	Yes — by construction

A decision script for the field

Will this dataset ever be overlaid, revisited, or imported into a design or machine model? If genuinely no — average and move on, deliberately. If yes: does trustworthy network coverage reach the site or a short errand away? Transfer, if the subscription already exists. Neither, and the deliverable is legal-grade? Statics plus OPUS/AUSPOS for the authoritative trail — ideally logged alongside method 4 so today’s work is not hostage to tomorrow’s email. Everything else — which is most drone mapping, agriculture, and construction on unknown ground — PPP self-convergence is the method whose costs (three minutes, one service) are smallest and whose output (absolute, repeatable, ITRF2020) is the one the other methods are trying to reach.

Whatever you choose: write it down

Method, frame, epoch, and antenna height — four fields in the project notes turn any of these approaches into defensible work, and their absence turns even good coordinates into folklore. Averaged bases especially: record the offset caveat explicitly, so the future colleague overlaying your data curses the geometry, not your memory (the metadata that matters). Coordinates are only as durable as their documentation.

Combining methods: the belt-and-braces pattern

The methods stack elegantly for high-stakes work. The professional pattern on unknown ground: converge by PPP and start working within minutes (method 4), while the base simultaneously logs statics; that evening, submit the log to OPUS or AUSPOS and file the report beside the PPP coordinates (method 3). Agreement — typically within a couple of centimetres — gives you two independent, differently-derived pedigrees for the same position, which is more than most monumented setups can claim. Total marginal cost: one file upload. For regulated deliverables, cadastral support, or any client who asks 'how do you know', this two-method receipt ends the conversation.

One-line takeaway

Averaging is fast but relative, transfers and statics are absolute but cost errands or hours, and PPP self-convergence delivers absolute coordinates in the time it takes to unpack the drone — choose deliberately, document the method, and let the base's answer be one you can defend.

What surveyors think of self-established coordinates

Professional acceptance has moved quickly because the pedigree is checkable: a PPP-converged position agrees with an OPUS or AUSPOS solution from the same occupation to within a couple of centimetres, and any surveyor can reproduce that check from the archived RINEX. What licensed practice still rightly demands is documentation — method, frame, epoch, service — and delivery on the legal datum via named transformations. Meet those and self-established coordinates read as modern practice rather than shortcut; skip them and even monument-perfect work reads as folklore. The standard was never really about the monument.

Further reading

The convergence mechanics behind method 4's three minutes are in convergence time explained.

Frequently asked questions

Can I just average longer for better base coordinates?

Not meaningfully — autonomous GNSS errors are correlated over hours, so a 15-minute average is barely better than a 3-minute one. Averaging buys convenience, not absolute accuracy.

How accurate is OPUS or AUSPOS for base coordinates?

Centimetre-class with a solid multi-hour static log — authoritative and free. The cost is occupation time and next-day-class latency.

What does a PPP base need to self-position?

Open sky and an active L-Band correction service — no internet, no monument, no CORS. Convergence to ~1.5 cm typically takes about 3 minutes.

Does the base location matter if coordinates are self-computed?

Only for sky view and radio reach. The coordinates are absolute wherever it sits, so placement becomes a link-planning decision, not a control decision.

Is an averaged base ever acceptable professionally?

Yes — for genuinely self-contained work, chosen deliberately and documented. It becomes malpractice only when its data is later treated as absolute.