Agriculture is one of the most GPS-dependent industries there is. Autosteer, variable-rate application, yield mapping and drone surveys all assume a clean satellite fix. For most of the year, over most of the country, that assumption holds. This article is about the times it does not, and about what a terrain-based navigation source can realistically offer a fixed-wing farm drone. It also spends some time on the honest limits, because flat, uniform paddocks are the hardest ground there is for terrain matching.
Why the farm depends on GPS, and when it fails
A fixed-wing survey aircraft flying a broadacre property spends most of its time on long, straight transects. Everything it produces, from a multispectral map to a plant count, depends on knowing where each image was taken. Spray and spread aircraft go further: position is not just recorded, it decides when the nozzles open and close. An error in position becomes chemical on the wrong side of a fence, a missed strip or an overlapping double dose.
Rural properties are not immune to GPS problems, they just experience different ones. Satellite geometry can be poor at certain times of day over a low horizon. Tree lines and silos block or reflect signals near the boundaries of a run. Solar activity degrades accuracy unpredictably. Some operators near defence ranges or industrial sites have reported interference. And the correction services that lift consumer GPS to survey grade depend on mobile coverage that is patchy across much of regional Australia.
None of these are constant, which is the problem. A system that is fine most of the time and quietly wrong occasionally is harder to manage than one that is never available.
Long broadacre flights and the drift problem
Broadacre missions are long in both time and distance. That is exactly the situation in which relative navigation methods perform worst, because their errors grow with every kilometre flown. An inertial or visual-odometry fallback might hold a straight line well enough for a short while after a GPS loss, but a flight of an hour or more over a large property needs something that does not drift.
A terrain-referenced system offers a source of absolute position that remains stable for the whole flight, provided the ground it can see is distinctive enough to match. On mixed farmland with roads, dams, tree lines, contour banks, creek lines and paddock boundaries, that condition is often met.
Spray boundaries and repeatability
Two things make spraying and spreading unforgiving. The first is boundaries: the aircraft must know where the paddock ends and the neighbour, the waterway or the road begins. The second is repeatability: the same paddock has to be flown the same way next month, and the tracks have to line up with tramlines and previous applications.
Both come down to absolute position. A navigation source that reports the same coordinates for the same fence corner today and next season, without depending on satellite conditions on the day, is directly useful for both problems. Terrain matching against a fixed database has that property in principle, because the reference does not move.
The honest bit: flat, featureless paddocks
Terrain-referenced navigation works by recognising the ground. A freshly sown paddock that is flat, uniform in colour and identical to the paddock next to it offers very little to recognise. The same is true of bare fallow, a large flooded rice bay, or a canola crop in full flower that stretches to the horizon. Over that kind of ground the matcher can lose confidence, and a well-designed system will say so rather than guess.
Some factors help and some hurt:
- Fence lines, tracks, dams, tree belts, contour banks and infrastructure give the system something to anchor to, even in otherwise uniform country.
- Higher altitude brings more of these features into each frame, at the cost of image detail.
- Crop stages that leave visible row structure or texture are easier than a closed, uniform canopy.
- Recent, high-resolution reference imagery of the actual property matters more here than anywhere else, because subtle features are all there is.
- Long stretches with no anchoring feature at all will rely on the inertial bridge, and the position estimate will degrade until the next confident match.
The practical consequence is that terrain-based navigation over broadacre country is a reliable complement rather than a universal replacement for GPS. It is at its best when treated as a way of removing single-point dependence on satellites, not as a promise that position is always available regardless of the ground.
Practical mission design
Operators who want GPS-denied capability on the farm can plan around these realities. Routing transects to cross roads, dams and tree lines periodically gives the system regular anchor points. Flying survey work at an altitude that keeps multiple features in view, rather than the lowest permitted height, usually improves matching. Keeping the reference imagery current, ideally from the operator’s own previous surveys, helps considerably. And configuring the autopilot to treat the terrain source as a monitored secondary input, rather than blindly trusting either it or GPS, gives the pilot a clear signal when one disagrees with the other.
Where TerrainSLAM fits
TerrainSLAM provides an absolute position by matching the aircraft’s view of the ground against pre-loaded terrain data, running onboard without a ground station or network link, which suits long flights over properties with poor mobile coverage. It is built for the forward flight and endurance of fixed-wing aircraft and adapts across altitudes, and it is designed to report its confidence so an operator knows when the paddock below offers little to match against.