Surveying a Site With No Cell Coverage or Poor GNSS Reception

Plenty of Alberta work is nowhere near a cell tower. Pipeline right of way in the northwest, a gravel pit an hour off the highway, a lease road past Hinton, a cutblock, a remote facility. And plenty of work that has perfect cell service is still difficult for a drone to position in: the floor of a deep pit, a tank farm, under canopy, tight against a building.

Both get reported the same way, as bad GPS out there. They are different problems with different fixes, and confusing them is how a job gets quoted with the wrong plan.

Two problems that share a complaint

Centimetre-grade drone positioning needs two things. It needs the satellite signals themselves, and it needs a correction telling the receiver how much those signals are being distorted by the atmosphere at this location right now.

So:

• No cell coverage removes the delivery route for the correction. The satellites are fine. The drone is receiving them normally. What is missing is the live correction stream that would normally arrive over the cellular network.
• Poor reception is a problem with the satellite measurement itself: not enough sky, or signals arriving by a reflected path. The correction can be perfect and still not save a measurement that is contaminated.

The first is a logistics problem with a standard solution. The second is a physics problem that has to be worked around rather than solved.

No cell service: you lose the convenience, not the accuracy

Network RTK depends on a cellular link to a correction service. No signal, no stream, and a crew expecting to work that way is stuck.

The answer is older than network RTK and is routine on remote jobs. Either carry your own base, or correct after the fact:

• A local base station. A GNSS receiver set up on site broadcasting corrections directly to the aircraft over radio, with no network involved. Real-time positioning, same result, one more piece of equipment and a setup window at the start of the day.
• PPK. Fly first, correct later. The aircraft logs raw satellite observations throughout the flight, the base logs its own, and the two are processed together afterwards to produce the corrected camera positions. Nothing has to be live, so nothing has to be connected.

PPK is the better fit for remote work for reasons beyond coverage. There is no correction link to drop mid-flight and no partial dataset to discover at the end of a long drive. And because the processing happens afterwards, it can be redone with better base data or different settings without reflying. On a site that takes two hours to reach, that robustness is worth more than real-time convenience.

What a buyer should take from this: no cell coverage is not a reason to accept a lower accuracy specification. If a contractor quotes degraded accuracy because the site is remote, the question is whether they are equipped to run a base and process PPK. That is standard practice, not a special service.

The one thing to verify on a remote job

Everything on a base-and-PPK job hangs off a single occupation, so it is worth one question.

The base can be set up two ways. Over a known point, meaning an existing survey monument or a control point with published coordinates, which anchors the whole job directly to the project's datum. Or over an unknown point, where the base logs its own position and is tied to the datum afterwards, either by a long occupation processed against a reference network or by surveying the base position later.

Both are legitimate. What is not legitimate is a base set over an unknown point whose position is then taken from an uncorrected single-point average, because that average can be out by a metre or more. It will not degrade the internal quality of the survey at all. Every point in the model will be beautifully consistent, and the whole thing will be sitting a metre from where it belongs, which is the difference between relative and absolute accuracy showing up in the worst way.

So: what was the base set over, and how was its position established? On a job that will be tied into existing survey or used for design, that answer matters more than the drone model.

Poor reception: an obstructed sky and reflected signal

The harder case is where the sky itself is the problem. Two mechanisms, usually together.

Obstruction is straightforward: fewer satellites visible means a weaker solution. Standing on the floor of a deep pit, a high sidewall cuts off a large part of the sky, and the satellites still visible are clustered overhead in a geometry that is poor for resolving height. Under canopy the signal is attenuated and intermittent. Tight against a tall structure, half the sky is gone.

Multipath is the one that causes more damage and gets less attention. A signal that bounces off a rock face, a steel tank, a building wall or wet ground arrives late, having travelled further than the direct path. The receiver may treat it as a direct measurement, and the resulting position is wrong without anything looking wrong. Multipath is worst in exactly the environments the work is in: pit walls, tank farms, process structures, equipment yards.

What this does in practice:

• Vertical degrades first and worst. Height depends on satellite geometry more than horizontal position does, and an obstructed sky damages geometry specifically. If a site is obstructed, the vertical figure is the one to specify and the one to verify.
• The degradation is not uniform across the site. Open ground in the middle of the pit may position normally while the area tight against the sidewall does not. A single accuracy number for the whole job hides that, and the toe of the sidewall is often the part somebody needs.
• It can come and go. Satellite geometry changes through the day, so a corner that was marginal at 9 am may be fine at 2 pm. On a known-difficult site that is worth planning around rather than discovering.

Ground control is the fallback that always works

This is the useful thing about ground control points: they need no signal of any kind. A surveyed target on the ground is a known coordinate whether or not the sky is visible from the aircraft, because the surveying was done separately, by whatever method actually works there.

So the hierarchy on a difficult site runs in this order. Where reception is good, RTK or PPK carries the job with a few points held back as checks. Where reception is marginal, control comes back in to constrain the model and to establish the vertical, which is the axis that failed first. Where reception is genuinely bad, control becomes the primary georeferencing method and the aircraft positioning is a convenience rather than the basis of the survey.

Two practical notes. The targets themselves have to be surveyed, and in an obstructed environment that may mean a total station traverse in from clear ground rather than a GNSS rover occupation, since the rover faces the same obstruction the drone does. And on an obstructed site placement matters more than count, because the point of the control is to pin down the specific areas the positioning cannot be trusted in. The general principles in how many ground control points a job needs still apply, with the distribution weighted toward the obstructed ground rather than spread evenly.

Where even that is not enough, the honest answer is that the method changes. A total station or a GNSS rover still beats a drone in some confined, obstructed geometry, and the right job is the combination rather than one or the other.

What the buyer actually trades

None of this costs accuracy if it is planned for. It costs time, and time is what shows up on the quote:

• A base occupation. Setting up, levelling and occupying a base is a window at the start of the day, longer if the position has to be established by a long occupation rather than read off a monument.
• More targets, in harder places. Placing and surveying control on a pit floor or under canopy is slower than on a flat pad, and some of it needs a traverse rather than a rover.
• Longer flight planning and more flights. Obstructed geometry means lower lines, more overlap and sometimes oblique passes to see into the areas a nadir pass cannot.
• Mobilization. Remote sites are a drive, and that is a cost driver on its own.
• Occasionally a second visit. If the first dataset does not hold the vertical in a critical area, the fix is more ground truth, which is a return trip if the control was not placed the first time.

Which is the argument for describing the site honestly before the quote rather than on arrival. A contractor told it is a 40 hectare pit with 30 metres of sidewall relief, no cell coverage and a tie-in to existing survey will bring a base, bring targets and price a full day. A contractor told it is a 40 hectare site will bring the normal kit and find out at 8 am.

Questions worth asking before a remote or obstructed job

• Do you carry your own base station, and do you process PPK?
• What will the base be set over, and how will its position be established in our datum?
• Given the sidewalls and the canopy, what vertical accuracy are you committing to, and in which parts of the site?
• How much ground control are you placing, and how will the targets themselves be surveyed in an area with poor reception?
• Will any surveyed points be held out as independent checks, and will the comparison be in the report?
• If the vertical does not hold in the critical area, what is the remedy and who pays for the return visit?

Six questions, and the answers tell you whether the plan accounts for the site. Remote and obstructed work is ordinary in Alberta and it is entirely survivable. What is not survivable is finding out after the deliverable arrives that the correction stream was never there and nothing independent was ever measured on the ground.