LoRaWAN Repeater Placement in Complex Terrain: A Field-Tested Strategy Guide

A recent study by Almeida et al. on LoRaWAN deployments in mountainous environments shows a 58.63% decrease in successfully received packets compared to flat terrain in the same experimental setup. In the mountain scenario tested, roughly one third of the sensor nodes failed to reach the ground gateway at all. Any deployment in anything other than an open field needs a strategy for extending coverage.

Repeater, relay, or additional gateway: clearing up the confusion

The terms "repeater" and "relay" get used interchangeably across the industry, but there is an important difference in how these devices work.

Proprietary repeaters are hardware devices that receive a LoRaWAN uplink frame and retransmit it toward a gateway. Products like the ATIM ACW-LW8-EXT can handle up to 8 different LoRaWAN devices, cover ranges up to 15 km in their datasheet conditions, and work transparently, which means the gateway and network server don't know the frame was repeated. They were originally developed for coverage holes in basements, parking lots, and boiler rooms, but the same principle applies to terrain-shielded locations. The key limitation: the standard EXT variant is mains-powered (230 V AC), which complicates deployment in remote terrain. A separate EXT-PWR variant exists with different power options for sites where mains is not available.

LoRa Alliance TS011 relays are a standards-based alternative. The specification was introduced as TS011-1.0.0 in October 2022 and later refined to TS011-1.0.1 with clarifications on MAC commands and the back-off mechanism. A relay is itself a LoRaWAN Class A device that listens on a low-duty-cycle Wake-on-Radio (WOR) channel, collects uplinks from nearby end-devices, wraps them in its own LoRaWAN frame, and forwards them to the network server through the nearest gateway. The critical advantage of relays is that they can be battery-powered with multi-year autonomy, because they only wake when a WOR preamble is detected. The network server needs to support the relay feature, but gateways require no modification. TS011 supports exactly one relay hop in the current specification; multi-hop is not part of the standard.

Adding a gateway is always an option, but it comes with infrastructure constraints. A gateway like the Move Solutions GW-PRO 8 requires 4G LTE, Ethernet, or Wi-Fi backhaul, plus PoE or 12V DC power. In remote terrain, a gateway might require solar panels, cellular, and weatherproof enclosures. Adding a gateway is the most reliable solution for large coverage gaps.

If you have power available and need to cover a confined dead zone (a basement, a tunnel entrance, a shielded valley floor), a proprietary repeater is the simplest fix. If you need to extend coverage to a remote cluster of sensors where running power is impractical, a TS011 relay is your best option since they are battery-powered and standards-compliant. If the coverage gap is structural (you're trying to cover a 5 km valley from one side), you need another gateway with backhaul.

How terrain kills your signal

Fresnel zone obstruction

Radio waves don't travel in a laser-thin line between transmitter and receiver. They propagate through an ellipsoidal volume called the Fresnel zone. For an 868 MHz LoRaWAN link over 5 km, the first Fresnel zone has a radius of roughly 21 meters at the midpoint. Any terrain, vegetation, or structure that intrudes into this zone causes signal loss, even if there's a clear optical line of sight between antenna and gateway.

The engineering rule of thumb is that at least 60% of the first Fresnel zone radius should be clear of obstructions. Below that threshold, diffraction losses become significant and grow rapidly, which is why a sensor that appears to have "line of sight" to the gateway can still drop packets.

Knife-edge diffraction

When a signal grazes a sharp terrain feature, like a ridge or a rooftop edge, it diffracts around the obstacle. The ITU-R P.526 model quantifies this effect and it's not always bad news: diffraction can actually bend signals into shadowed areas, but the losses are unpredictable. ITU-R P.526 typically predicts 15–25 dB of diffraction loss for grazing geometries common in mountainous terrain, so a repeater placed just above the diffracting edge can recover 20+ dB compared to relying on diffraction alone.

The antenna height multiplier

This is the single highest-ROI intervention for terrain-challenged deployments. Research in the 915 MHz ISM band (Experimental Evaluation of Antenna Height Impact on LoRaWAN Performance) shows that elevating a gateway antenna from 1.5 m to 10 m improves mean RSSI by roughly 10–15 dB, consistent with what classical models (Hata, Okumura) predict. Height-gain effects are essentially frequency-independent across the sub-GHz ISM bands, so the same relationship applies at 868 MHz. Path-loss models like Hata predict that raising gateway height from 12 m to 30 m extends range by roughly 40 to 50%, since the height-dependent loss term decreases by 5 to 6 dB.

In practice, before adding repeaters or relays, check if you can mount the gateway higher. A taller mast, a hilltop location, or even moving the gateway from a ground-level enclosure to a rooftop can eliminate the need for additional network elements entirely.

What to measure during a site survey

It is always necessary to do a site survey before deployment. A systematic RF survey takes half a day and prevents weeks of troubleshooting.

What to measure

Walk the deployment area with a test transmitter at each planned sensor location and measure received signal at the gateway position. Your target thresholds:

• RSSI above -110 dBm (above -100 dBm preferred for reliable operation)
• SNR above -10 dB (above -5 dB preferred)
• Packet delivery ratio above 95% (test with at least 100 packets per location)

Any sensor location that fails these thresholds needs a coverage extension, so either a repeater, a relay, or a repositioned gateway. While you're at it, run a quick check at the temperature extremes of your deployment envelope: crystal drift in end-devices at low or high temperatures widens the preamble detection window and can quietly degrade PER at high spreading factors, even when the RF link itself looks fine. Semtech's frequency-tolerance application notes recommend a TCXO over a standard crystal for any deployment expected to operate below −20 °C or above +60 °C, particularly when using SF11/SF12.

Test under worst-case conditions

A lot of seasonal changes can affect signal reception. Deciduous foliage typically adds 3 to 10 dB of seasonal loss between bare-branch winter conditions and full-canopy summer, while snow accumulation on antenna radomes attenuates signal. Note that fully wooded propagation paths through dense canopy can attenuate by 20–50+ dB at 868 MHz — this is a much larger effect than the seasonal variation, and should be designed around rather than absorbed by margin.

Build a 10 to 15 dB margin above minimum thresholds to absorb seasonal and weather-related variation.

UAV-assisted surveys

For large or inaccessible terrain, UAV-based RF surveys are transformative. Research comparing ground-level and aerial LoRaWAN measurements in obstacle-heavy environments found that a ground gateway at 1.3 km achieved an SNR of -16 dB with 73% packet loss, while a UAV relay at altitude achieved +5 dB SNR with only 3% packet loss. A drone carrying a test receiver can map the RF environment across an entire slope or valley in hours, identifying optimal repeater positions that would take days to find on foot.

Document everything

GPS-tag every measurement point. Record antenna heights, orientations, and environmental conditions. This baseline data is invaluable when the network needs expansion or when seasonal changes cause unexpected packet loss six months after deployment. The MyMove platform can track sensor connectivity metrics over time, making it easier to identify degradation trends before they become outages.

Placement strategies by terrain type

Each terrain type presents distinct RF challenges and requires a different approach to repeater placement.

Mountain slopes and landslide-prone areas

Geotechnical monitoring on unstable slopes is one of the most RF-hostile scenarios in structural monitoring. In this type of deployment, wireless tiltmeters track millimetre-scale ground movement. Sensors are scattered across a slope face, often below ridgelines that block direct communication with the gateway.

In this case, place repeaters or relays on natural ridgelines and crests that overlook the sensor field. A single relay on a ridge can serve sensors on both the near and far slopes. Ensure 10 to 20% coverage overlap between relay zones for redundancy — if one relay fails, adjacent relays should still reach the affected sensors.

Power consideration: Mountain deployments are almost always off-grid. Battery-powered TS011 relays or solar-powered repeater stations are the practical options. A small 10W solar panel with a charge controller can sustain a repeater indefinitely at most latitudes.

Duty-cycle consideration: A repeater multiplies transmissions on the same EU868 sub-band. If you stack too many sensors behind a single repeater, the repeater itself can saturate the 1% local duty cycle and start dropping uplinks during burst traffic. Plan repeater fan-out conservatively, especially for sensors with event-triggered transmissions.

Valleys and canyons

Valleys create a natural RF trap. The gateway sits at the top (where backhaul is available), but the sensors sit at the bottom (where the structure is). The valley walls absorb and reflect signals, creating multipath interference and deep fading zones.

In these cases, placing repeaters mid-slope, so roughly half the valley depth, on a protruding rock face or an existing utility pole, can bridge the gap between valley-floor sensors and ridge-top gateway. Avoid placing repeaters directly on the valley floor, because they'll suffer the same terrain shielding as the sensors. In narrow canyons, signals can propagate along the canyon axis through waveguide effects, so orient the repeater antenna to exploit the canyon geometry.

For bridge monitoring in valleys a relay on the bridge deck itself often provides the best geometry, with clear sky view upward toward the gateway position.

Dense urban environments and heritage buildings

Historic masonry, reinforced concrete, and the sheer density of urban RF interference create a different kind of challenge. A 60 cm stone wall can attenuate an 868 MHz signal by 10 to 20 dB, depending on density, moisture content, and any embedded steel. Two walls and a floor slab, and you're looking at 30+ dB of loss.

In urban structural monitoring, the priority is minimising the number of walls between sensor and gateway. Place the gateway on the rooftop of the tallest building in the monitoring zone, with a high-gain antenna well above surrounding rooftops (at least 5 to 7 m above the roof surface). For sensors inside buildings, a repeater at a window or exterior wall position can relay signals outside the building envelope where the gateway can receive them.

When monitoring clusters of adjacent buildings, a single strategically placed exterior repeater can serve multiple structures. The repeater doesn't need to be on the monitored building, since a clear mounting point on a neighbouring structure with better RF geometry is often more effective.

Linear infrastructure: bridges, railways, tunnels

Monitoring linear structures like bridges is a different problem entirely, since the gateway needs to serve sensors distributed along a line instead of a cluster of devices.

For open-air linear structures, place the gateway at one end or at the midpoint, elevated above the structure. LoRaWAN's range is generous along an unobstructed line, so a single GW-PRO 8 gateway can typically cover a bridge up to 2–3 km in length without repeaters under typical conditions, assuming clear line of sight along the deck. Longer spans, strongly curved decks, or bridges with significant obstructing superstructure (truss elements, suspension cables in the near-field) may still need a repeater.

For tunnel monitoring, place a gateway or repeater at each portal (entrance/exit), with overlapping coverage toward the centre. Long tunnels (>500 m) may need an intermediate repeater. Mount antennas on the tunnel wall at a height that avoids the worst multipath from floor reflections, typically 2 to 3 m above track or road level. This strategy uses RF propagation of a tunnel which follows waveguide physics (see e.g., Emslie et al., IEEE Trans. AP, on cutoff frequencies and mode-stripping in tunnels): signals attenuate rapidly in the near-field region (typically the first tens to hundreds of metres, depending on tunnel cross-section, frequency, and wall roughness), then propagate more efficiently as the tunnel acts as a lossy waveguide.

ADR tuning and post-deployment optimisation

Once repeaters are placed and sensors are transmitting, the network still needs tuning.

Adaptive Data Rate (ADR) is LoRaWAN's built-in mechanism for optimising transmission parameters. The network server monitors each device's SNR and adjusts spreading factor (SF), bandwidth, and transmit power to find the best trade-off between range, battery life, and airtime.

Since terrain-challenged links experience high SNR variance, this system is not as straightforward with difficult terrain. A sensor on an unstable slope might report SNR of +3 dB on a clear day and -12 dB after heavy rain changes ground conductivity. Standard ADR sees the good SNR and reduces the spreading factor for efficiency, but then the next transmission fails because conditions have degraded.

Set a conservative ADR margin (at least 10 dB above the minimum required SNR for the target data rate) for sensors in terrain-challenged locations. This sacrifices some battery life and airtime efficiency but dramatically improves reliability. For structural monitoring, where you might transmit once every 15 minutes, the airtime cost of a higher spreading factor is negligible compared to the cost of lost data points.

Keep in mind the EU868 duty cycle constraint: the default sub-bands available to end-devices are typically capped at 1% per ETSI EN 300 220, so at SF12 a single 1.3 s transmission ties up the channel for over two minutes before you can transmit again, which limits how aggressively you can re-send on critical sensors.

Consider disabling ADR entirely for critical sensors in the most volatile RF environments and manually assigning SF12 (maximum range, maximum reliability). A Move Solutions tiltmeter with its 19 Ah LiSOCl₂ battery can sustain years of operation even at SF12 with typical 15-minute reporting intervals.

Monitor frame counters and packet delivery ratios through your network server. When a sensor's delivery suddenly drops from 99% to 85%, it's probably because the RF environment has changed. Seasonal foliage growth, new construction, or even a shifted repeater antenna (wind, ice, wildlife) can degrade a previously reliable link. Catching these trends early through connectivity monitoring prevents data gaps from becoming data losses.

Putting it all together

The decision sequence for any complex-terrain deployment follows a consistent pattern:

1. Start with the sensor placement. That's non-negotiable, driven by the monitoring requirements.
2. Survey the RF environment from each sensor location to the proposed gateway position.
3. Where the link budget fails, evaluate your options in order:
   • First, can you raise the gateway antenna higher or relocate it to better terrain?
   • Second, can a single repeater or relay on a natural vantage point recover the lost coverage?
   • Third, do you need an additional gateway with independent backhaul?

Every repeater you add is a point of failure you need to maintain. The best network is the simplest one that meets your reliability requirements. Sometimes that means investing in a better gateway location rather than papering over a bad one with repeaters.

For monitoring systems protecting critical infrastructure, like retaining walls on active landslides, bridges under load, or buildings adjacent to excavation, network reliability isn't a nice-to-have. A missed tilt reading during a critical movement event isn't recoverable. Design the RF network with the same engineering rigour you'd apply to the structural assessment itself.

Common questions

How far can a LoRaWAN repeater extend coverage?In open terrain, a repeater effectively doubles the link distance. If your gateway reaches 3 km, a well-placed repeater can extend that to 5 or 6 km total. In complex terrain, the gain depends entirely on how much terrain shielding the repeater eliminates. A repeater that clears a ridge obstruction might recover 20 dB or more, transforming an unusable link into a reliable one regardless of absolute distance.

Can I daisy-chain multiple repeaters?Technically possible with proprietary repeaters, but each hop adds latency and reduces reliability. TS011 relays support exactly one hop in the current specification; multi-hop is not part of the standard. For monitoring applications, if you need more than one relay hop the better solution is usually an additional gateway.

Do I need a site survey for every deployment?Yes. Propagation models give you a starting point, but the difference between predicted and measured signal strength in complex terrain can exceed 20 dB. A half-day survey prevents weeks of troubleshooting after installation.

How does weather affect repeater performance?Pure rain attenuation at 868 MHz is typically well below 0.5 dB per kilometre even in heavy rain — orders of magnitude lower than at microwave frequencies. The real weather penalty comes from wet foliage, which can add several dB compared to dry leaves, and from water films or snow and ice on antenna radomes. Design your link budget with a 10 to 15 dB margin to absorb these variations without losing connectivity.

Should I use SF12 for all sensors in difficult terrain?Not necessarily. SF12 gives you maximum range but uses roughly 32 times more airtime than SF7 for the same payload. For sensors reporting every 15 minutes, this is negligible. For sensors reporting every minute, it can cause EU868 duty cycle violations. Start with ADR enabled and a generous margin, then manually override only the sensors that still show unreliable connectivity.