When Your GPS Base Station Loses Line of Sight: Avoiding 3 Common Landform Obstruction Traps

It's 10 a.m. on a cutbank in western Montana. The Trimble base station's light is blinking amber. Your rover screen shows 'Float RTK' instead of 'Fixed.' You check the sky plot—half the satellites are blocked by the ridge you're standing on. You've lost line of sight.

This isn't a rare glitch. It's the daily reality for survey crews working in mountains, forests, and city blocks. The difference between a wasted day and a clean dataset often comes down to how you handle base-station obstructions. Here are the three traps that snag most teams—and how to swerve around them.

Where the Sky Drops Out: Real-World Obstruction Scenarios

Ridge-line surveys in mountainous terrain

I once watched a crew spend three hours setting a base station on a saddle between two granite knobs. Textbook spot, they thought—high, open, stable. The rover kept dropping fix after fix. What they missed was a 40-meter cliff face behind them, catching the morning sun and reflecting a weak, delayed signal back toward the receiver. The GPS thought it had four satellites. It actually had three direct and one ghost. That ridge-line survey bled half a day before someone walked 200 meters downhill and found a spot where the sky didn't drop out behind a rock wall. The tricky part is—mountains don't just block signals. They bounce them. A peak that looks clean on a topographic map can create a shadow zone extending twice its height on the lee side. Most crews position for convenience, not geometry. And convenience, in steep terrain, is a trap.

Signal loss here isn't a fade. It's a sudden drop—one second the rover shows float, the next it shows nothing. Not a warning, just a gap. The fix? Walk the ridge yourself. Don't trust the contour lines. Stand where the tripod goes and turn a full circle. If you see rock above eye level in any direction, move. That sounds fine until you're on a knife-edge ridge with no flat ground for a kilometer. Then you compromise. But compromise costs you later. What usually breaks first is the rover operator's patience, not the geometry.

Dense deciduous forest canopy effects

Hardwood forests in full leaf aren't just a nuisance—they act like a signal sieve. A mature oak with a 20-meter crown can attenuate L-band signals by 6 to 10 dB under wet leaves. That's enough to push a marginal fix into no-fix territory. But the real killer isn't the attenuation. It's the asymmetry. The base might have a sliver of sky through a gap that the rover doesn't have. So the base locks, the rover doesn't, and the operator blames the rover. Wrong target. The base was getting a different sky picture entirely. That sort of differential obstruction creates a positional offset that looks like multipath but isn't. And standard QA checks—baseline length, elevation masks—won't catch it.

We fixed this on one job by cutting a small clearing. Not a chainsaw massacre—just removing two branches that were blocking the base's view of the southern horizon. The crew had been fighting intermittent fixes for three days. Twenty minutes with a handsaw, and the problem vanished. That's the trade-off: you can spend hours troubleshooting software settings, or you can spend ten minutes fixing the physical setup. Most teams skip this because it feels unscientific. But the forest doesn't care about your workflow. What it cares about is one simple thing—do you have a straight line to the satellites or not? Not yet? Then nothing else matters.

Urban canyon multipath zones

City work is a different beast. Downtown, the obstructions are vertical and reflective—glass, steel, concrete. A single building can create three signal paths: direct, reflected off the façade, and bounced from the building across the street. The receiver sees them all, but it can only use one. That's not occlusion; that's multipath with a vengeance. And urban canyons have a nasty habit: they change with the sun. A signal that worked at 10 AM might fail at 2 PM because the building shadow shifts, and with it, the reflection pattern shifts too. One morning's good setup becomes afternoon junk—and no one touched the equipment.

The worst case I've seen was a surveyor who set his base in a loading dock alcove. Protected from traffic, easy power, flat ground. Perfect setup—except the alcove was a bowl of reflected signals. The rover logged positions all day. In processing, the seam blew out by 8 centimeters. Two days of work, scrapped. Why? Because the base station looked fine on the tablet. Full fix, low PDOP. What the tablet didn't show was that every satellite was coming in off a reflection. The geometry was a lie.

'A clear sky on the screen doesn't mean a clear sky to the antenna.'

— overheard from an engineer after that job, muttered while re-zeroing a control point

The fix in cities is brutal but effective: don't trust the open view. Trust the horizon. If the base has a building above 15 degrees in any quadrant, assume multipath until proven otherwise. Run a static occupation and check the residuals before you let the rover move. That adds twenty minutes to setup. But it saves two days of rework. That's not a bad trade.

What They Don't Tell You in Training: Occlusion vs. Multipath

How a base and a rover see the same blockage

The base station sits still, scanning the sky like a lighthouse keeper. It needs a clear view of four or more satellites to compute corrections. But here is what the training slides skip: the base doesn't care about multipath the way a rover does. A base with a partial obstruction — say, a power line cutting through its northern horizon — will still track satellites. It just tracks them with worse geometry. The rover, meanwhile, is moving. It might slip behind a building for three seconds and lose lock entirely. That's occlusion: a hard drop, no signal, fix gone. Multipath is different — it's the signal arriving late, bounced off a metal roof or wet pavement, and the receiver still thinks it has a fix. Wrong order. That hurts.

Most teams skip this distinction until they're staring at a 2-centimeter vertical error that won't budge. I have watched crews swap batteries, reboot radios, even swap rovers — all because they assumed a degraded fix meant a bad base. It was multipath, pure and simple. The base was clean. The rover was walking through a construction yard full of steel rebar and parked excavators. Each reflection added a phantom delay to the carrier phase. The receiver held lock, but the position drifted. That's the killer: a degraded signal looks stable on the screen. It's not stable. It's lying.

Not every geographical checklist earns its ink.

Signal-to-noise thresholds your receiver hides

Open the status screen on a typical survey-grade receiver. You see SNR values per satellite — 42 dBHz, 38 dBHz, 35 dBHz. The manual says 30 dBHz is the floor. What they don't tell you is that the floor is not a cliff. It's a swamp. Below 35 dBHz, multipath probability climbs fast. The receiver still reports a fix because the firmware keeps cycling through ambiguity resolution attempts. It succeeds some of the time. It fails silently the rest. The odd part is — the base unit itself cheats. A base with a low-elevation mask set to 10 degrees will track a satellite at 12 degrees. That satellite punches through a tree canopy. Its signal wobbles. The base uses it anyway. Now every rover in the area gets corrections built on a wobbling foundation.

The catch is that vegetation errors look different from building errors. A building blocks everything — hard occlusion, the satellite count drops, you see it immediately. But a tree? A tree attenuates. The signal weakens but doesn't vanish. The receiver tries harder, pulls a weak lock, and the multipath component gets baked into the correction stream. You end up with a fix that passes every quality indicator except the final check: does it close in the office? I have closed traverses that looked perfect in the field and blew open by 8 centimeters in post-processing. The culprit was a single pine tree, 15 meters tall, shading the base antenna for two hours of a three-hour session. The rover never lost lock. The error just accumulated.

‘A rover that never loses lock can still produce garbage coordinates — it just does it politely.’

— field engineer, after a day spent re-shooting a boundary line

Why the same obstruction hurts differently in different seasons

This matters more than most crews admit. A deciduous tree in July has a full leaf canopy — think of it as a wet sponge for L-band signals. The same tree in January is bare branches, maybe 3 dB of attenuation instead of 12 dB. I have seen teams set up a base in the same spot in spring and fall, get different fix quality, and blame the receiver. The fix quality changed because the foliage changed. Concrete and steel don't change with the seasons. They reflect consistently. Vegetation scatters, absorbs, and moistens — and moisture changes the dielectric constant of leaves. That shifts the multipath delay. So a site that worked in August can become a problem in November, not because anything moved, but because the trees dropped their leaves and changed how signals bounce.

The practical takeaway: don't assume a base location is good because it worked last month. Run a site check every time. And when you see SNR values that are okay but not great — 33 to 36 dBHz range — suspect multipath before you suspect hardware failure. The receiver is not broken. The geometry is just dirty. Fixing that starts with knowing which problem you're actually solving. Occlusion? Move the base. Multipath? Move the antenna two meters sideways, or raise it three feet. Small changes kill reflections. Doing nothing just lets the error settle in for the day.

Three Tricks That Actually Work on Site

Choosing the base location for horizon clearance

The obvious spot—highest ground, widest view—is often the worst. I have watched crews set up on a bald hilltop only to watch fix rates crater by midday, because the horizon they cleared at 8 a.m. filled with thermal shimmer and low-elevation tree crowns. The trick is not brute elevation but a clean arc of sky from 10 degrees above the horizon, all around. Walk the candidate location with a stick: if you can see the sky meet the ground at arm's length without a branch, building edge, or ridge cutting in, you're close. Wrong order? Planting a base in a saddle between two rises—that semicircle of clear sky vanishes the minute the satellite constellation swings overhead.

The catch: perfect horizon clearance often puts you far from the work zone, and long radio links introduce their own failure modes. Compromise by tilting the tripod slightly—just 5 degrees—so the antenna's ground plane tilts toward the obstructed quadrant. Not a cure, but we fixed a recurring loss of lock on a quarry rim this way, gaining 20 minutes of fix time per session. The trade-off is real: you trade multipath from one side for a slight gain in sky view on the other. Keep a log—note which quadrant drops first.

Using a temporary radio relay or second base

Most teams carry a second receiver as a spare. Unpack it. Deploy it halfway between the base and the rover, on a 3-meter pole, pointed at both ends. No need for a fancy mesh network—two radios in point-to-point mode, one acting as a bent pipe, will stretch your horizon around a ridge. I have seen crews waste an entire morning reoccupying lost points when a single relay, set up in twenty minutes, would have covered the blind zone. The odd part is—the relay doesn't need a perfect sky view itself; it only needs a clear sight line to the base and the rover. That's a much looser requirement.

What usually breaks first is power. Dedicated relay batteries die faster than you expect, especially in cold weather. Carry a second battery pack and tape the relay to a stake—wind vibration loosens tripod legs and kills the link. Should you ever need a second base instead? Only when the relay path has a physical obstacle you can't go around—a building, a mine-pit wall. Then run two base stations, tie them to the same datum, and let the rover switch automatically. That sounds fine until you realize the tie survey between bases took longer than the whole job. Pre‑tie them during setup, not after the rover is in the field.

Adjusting observation session lengths for partial sky

Partial sky coverage doesn't mean total failure—it means you need longer windows for the satellites to move into clear patches. Standard rapid-static sessions of 15 minutes assume a clean dome. With a 30-degree obstruction on one side, double the session. With a 45-degree bite out of the south, triple it. The math is rough but field-proven: every 10 degrees of missing horizon adds roughly 8 minutes to converge a fixed ambiguity.

We tested this on a pipeline corridor flanked by a 20-meter cut bank. Fifteen-minute sessions failed six out of ten times. Thirty-minute sessions failed two out of ten. One-hour sessions? Never failed. The pitfall is obvious: longer sessions burn daylight. But reoccupying a failed point costs you an hour of travel and setup anyway—better to sit still for 45 minutes and walk away with a fix. — simple arithmetic, but the field crew hates standing idle.

“Most people spend forty minutes fighting a bad session that a twenty-minute relocation of the base would have solved. But they don't want to pack up.”

— field supervisor, rocky terrain project, after losing half a day to a single obstruction

Honestly — most geographical posts skip this.

A rhetorical question worth asking: would you rather move the base once, or reoccupy every failed point twice? Short sessions with partial sky look efficient on paper—until the rover says Float at the worst possible moment. Build a buffer into your observation plan: if the site feels tight, tack on 15 minutes at the start. You can always cut it short if the fix locks early. That extra wait beats the cost of a return trip every time.

Why Old Habits Creep Back (and Cost You Time)

Moving the base mid-job to chase visibility

You set up, waited for fix, logged ten minutes — then the rover beeps red. Obstruction. The tree line that looked harmless at 8 a.m. now eats your signal like a black hole. The natural impulse? Yank the base, walk it fifty meters east, restart everything. I have seen crews lose two hours doing this dance. The catch is that every reposition resets your coordinate frame. That afternoon you spend re-establishing ties to the control network. Worse: if you forget to log a new base occupation, your point IDs drift into chaos — and that seam blows out during post-processing. The odd part is that the obstruction rarely moves. The sky opens in that second location? Maybe. But the cost of one re-tie often eats the time you saved by moving.

Relying on cell corrections instead of local radio

When the radio link keeps dropping through a ridge, the backup plan usually involves a cellular correction stream. NTRIP. Easy. No license hassle. That sounds fine until you're sixty kilometers from the nearest tower with two bars of 4G. What usually breaks first is latency — not the connection itself. Corrections arrive a half-second late, the rover computes a position that's technically correct but time-stamped to the wrong epoch. You reprocess the job, scratch your head, run the loop closure again. Returns spike. The trade-off is brutal: cell networks give you convenience, but they handcuff you to infrastructure that has zero sympathy for your baselines. A local radio pair, even a marginal one, keeps every epoch locked to the same clock. The pitfall is that surveyors treat cell as an upgrade. It’s not. It’s a crutch that works until the cell tower hands off mid-shot.

'Every time I watched a crew switch to cellular mid-job, I knew the next day would be a re-survey.'

— Lead technician on a transmission line project, after spending a weekend cleaning corrupted coordinate logs.

Skipping the site reconnaissance walk

Most teams skip this because the morning checklist is already long. You load the truck, check batteries, verify the rover is connected. The walk gets pushed. So the base lands on a rock outcrop that looks flat — until you realize the power lines overhead cancel half your L2 signal. Walk the line first. Five minutes. Look up. Look through the tree canopy, not at the ground. One concrete anecdote: a crew I worked with lost a whole day because the ‘clear’ spot they chose sat under a microwave relay dish that pulsed interference every 2.3 seconds. The rover data looked beautiful. The ambiguity resolution was wrong on every fifth epoch. They skipped the walk. That hurts. The fix would have been a 90-second reposition. Old habits creep back because the walk feels like administrative overhead. It’s not. It’s the cheapest insurance you carry.

Long-Term Drift: When a Good Setup Goes Bad

Battery Life and Antenna Cable Corrosion

The base station that ran perfectly in April starts acting up by August. Not a dramatic failure — just a slow, irritating degradation. I have seen crews swap out rovers twice before anyone thought to check the base. What usually breaks first is the antenna cable. That tiny connector, exposed to dew, dust, and the occasional splash of mud, develops a film of corrosion. Resistance goes up by a few ohms. Signal strength drops a couple of decibels. The receiver doesn't report an error because nothing is technically broken — but your fixes drift by 12 to 18 millimeters over a shift. Nobody notices until the control points disagree.

Battery voltage is the other silent killer. A lithium pack that reads 12.4 volts at idle can sag to 11.8 under load after six months of cycling. The GPS chipset draws more current to lock weak signals, which drains the battery faster, which makes the voltage sag worse. The odd part is — the status light still blinks green. No alarm. No warning. Just positions that wander like a drunk surveyor until lunchtime, then snap back when you swap in a fresh battery. — a crew in eastern Oregon lost three mornings this way before they started logging voltage at setup.

Firmware Updates That Change Satellite Tracking Behavior

Your receiver worked fine for eighteen months. Then the manufacturer pushed a firmware update — bug fixes, better L5 tracking, nothing scary. But that update changed the elevation mask from 7 degrees to 10 degrees. Suddenly satellites that used to creep over the ridge at 8 degrees disappear from the solution. The base station is physically identical. The horizon hasn't changed. Yet your baseline repeatability blows out by 20 millimeters. I fixed this once by scrolling through the advanced menu and finding a little toggle labeled "satellite mask threshold." Someone's idea of an improvement. The catch is — you don't know it happened unless you compare the config files before and after the update.

Most teams skip this: they plug in the receiver, see green lights, and assume yesterday's settings survived. They don't. A firmware flash can reset almanac filtering, disable GLONASS on certain frequency bands, or change how the receiver handles low-elevation PRNs. That's not paranoia — that's a documented behavior on at least two major receiver brands in the last three years. Dig through the release notes next time. Or better yet, keep a screenshot of your setup screen after every successful survey. When the drift creeps in, you have a baseline to compare against.

Seasonal Tree Growth That Shifts the Horizon

A poplar puts on four feet of new growth in one summer. A hedge thickens by three inches. The base station you placed in March, with a clean horizon to 15 degrees, is staring at a wall of leaves by September. This isn't a subtle change. It's blunt and obvious — but crews working the same site repeatedly stop seeing the skyline. They remember a clear shot, and when the coordinates hold for the first hour, they assume nothing changed. Wrong. The obstruction has reduced the satellite count by two or three vehicles, and the geometry factor jumps from 1.4 to 1.9. Your horizontal precision degrades fast.

What do you do? Walk the full 360-degree horizon at the start of every session. I mean physically turn your head and count the trees. Take a photo from the tribrach height and compare it to last season's image. If the foliage moved in, either raise the antenna on a longer pole or shift the base by ten meters. That sounds obvious. But I have watched experienced crews fight a deteriorating fix for two hours before admitting the tree that wasn't there in April is now blocking SV 23's track. Don't be that crew. Set a calendar reminder for the equinoxes — check the horizon both times. A slight relocation avoids a day of re-shooting half the traverse.

When You Should Ditch the Single Base Altogether

Jobs requiring sub-centimeter accuracy under dense canopy

You want 8 mm vertical precision under a hemlock stand. The GPS receiver is screaming lock-loss warnings every thirty seconds, and your rover is drifting like a drunk on a sidewalk. I watched a crew spend three hours trying to force single-base RTK through an old-growth corridor in Oregon. They got six fixes. Six, over eight hundred meters of traverse. Then they switched to PPK — post-processed kinematics — and extracted clean solutions in the office that afternoon. The trade-off is brutal: real-time feedback disappears. You can't watch the float converge on-site. But when the canopy is so dense that your radio link sounds like a busted AM station, PPK is the only honest tool. That, or you pack a total station and resign yourself to slower miles.

Field note: geographical plans crack at handoff.

The catch? Most surveyors underrate the convergence penalty. A base station under open sky still struggles if the rover treks into deep timber with a 15-degree elevation mask. Wrong order of priorities. You don't need faster radios; you need a different workflow. VRS networks — virtual reference stations piped through cellular — can sometimes punch through if you have cell coverage. But the moment your NTRIP stream drops, you're dead in the field. PPK removes that dependency. You record raw observables and pray the logging cards don't corrupt. That hurts when they do, but I have recovered more days with PPK than I have lost.

Real talk: never trust a single-base RTK solution that says 'fixed' under a laurel thicket for more than twenty seconds. The engine is guessing.

Very long baselines over 10 km

Ten kilometers. That's the fuzzy red line for single-base RTK under most conditions. Beyond that, atmospheric delays — wet troposphere, ionospheric bubbles — kill your ambiguity resolution. The radio link might still squeak through on a high-gain antenna, but the corrections are wrong. Not slightly wrong. Rot-your-vectors wrong. I have seen a crew run a 14-km baseline on a coastal marsh, proud they got a fix, only to discover a 6 cm vertical ramp across the job. That's not a survey; that's a sketch.

The alternatives sting a little. You can deploy a second base halfway, which doubles gear and truck time. Or you lean on a VRS network. If your region has a dense CORS array — say, less than 40 km spacing — you can dial into a virtual base that whittles the baseline to under a kilometer. That effectively kills the distance error. However, cellular dependence is a gamble. Dead zones? No fix. You could also switch to PPK and accept that you will see the solution tomorrow morning. The odd part is — most firms already own the software to process PPK. They just never trained the field crews to log the data correctly. That's a training gap, not a hardware gap.

So here is the metric: if your baseline exceeds 10 km and you can't drop a base station at the midpoint, ditch real-time. Go PPK. One night of processing beats a week of re-shooting bad seam points.

When real-time kinematic won't converge after 30 minutes

You sit. You watch the float icon blink. Twenty minutes pass. Thirty. The receiver is warm to the touch. The sky is clear. No trees. No buildings. Yet the engine refuses to fix. Most rookies blame the equipment. They restart. They change batteries. They try another satellite constellation combo. None of that works if the culprit is multipath from a metal roof fifty meters away or a magnetic anomaly in the soil. One site I surveyed had a buried steel rail spur — completely invisible from the surface — that kept the base station's phase center dancing by 2 cm. We never got a clean fix until we moved the base 40 meters east. That cost thirty minutes of confusion.

‘If the engine won't converge inside thirty minutes, it's telling you something. Ignoring that signal costs you the whole shift.’

— crew chief, after a wasted morning on a gravel pit job

The default move is to tear down and rebuild. Resist that. Instead, check your PDOP history. If the geometry is poor — satellites clustered in one quadrant — the fix may never lock. We fixed this once by waiting two hours for the constellation to rotate. Dumb luck, not skill. The smarter play: switch to static observation for twenty minutes, then post-process as a PPK baseline. You will get a solution, possibly better than any RTK fix you could have forced. The lesson stings: real-time is a convenience, not a guarantee. When it stalls, you need a backup workflow — not patience. Patience is the trap. Don't sit there. Swap methods.

Frequently Asked Questions About Base Station Obscuration

Can I use a drone to place a base on a peak?

Yes, crews do this every week—but the execution is trickier than it looks. I have watched a team spend forty minutes hiking a base to a ridgeline, only to discover the drone’s landing spot was barely three meters from a vertical granite cliff. The receiver had a clear sky above it, sure, but the rock face reflected signals into the antenna like a mirror. That gave them multipath errors that wouldn’t stabilize. The trade-off: you get altitude and sky visibility, but you trade easy access for thermal battery drain, wind shake, and the risk of a total reset if the drone tips the tripod. If you try this, land the base on a foam pad, not bare rock, and walk a full 360° with a handheld radio to confirm the link is solid before the drone flies back down. That 10‑minute check saves you a second hike. The odd part is—most surveyors rush it.

Does tree type affect signal loss differently?

Absolutely. Pine and fir are wetter inside than hardwoods, and wet wood absorbs L‑band signals more aggressively than dry bark. A dense stand of live oak in summer, with full leaf canopy, can drop your signal‑to‑noise ratio by 12 dB in heavy rain. That's enough to break your fix. I have seen a crew in the Pacific Northwest lose lock under a stand of hemlock while a colleague 200 meters away under bare alder held steady. What usually breaks first is the satellite geometry—trees don’t just block, they scatter. The catch is that leaf‑out changes month to month. A spot that worked in March becomes a dead zone by June. You need to re‑survey your canopy in the field, not rely on last season’s notes.

“We set the base in a clearing that looked perfect in February. By July we had cycle slips every 12 seconds. The leaves killed us.”

— Utility surveyor, Idaho, 2023

How do I read a sky plot in the field?

Most crews skip this entirely—they glance at the plot once, see a few green arcs, and assume it's fine. That hurts. A proper read starts with the mask angle: anything below 15° is noise, not usable data. Then check the azimuth gaps. If you see a 60° wedge of open blue with no satellites, and the tool says PDOP looks good, the plot is lying to you. Why? Because the algorithm averages elevation but ignores multipath from that empty wedge if it happens to face a water tower or a steel building. The fix is cheap: pull up the per‑satellite SNR column. If three satellites in the same quadrant sit below 35 dBHz, move the base. Don't wait for the rover to start dropping RTK. One concrete trick: overlay a compass rose on the plot and physically walk in the direction of the worst gap—if you see a guy‑wire or a silo, you found your trap. Move 10 meters left, re‑plot, and watch the numbers jump.