Drone survey accuracy - why the answer is always "it depends"

Ask an engineer how accurate a drone survey is and the honest answer is: it depends. It depends on the method, the surface, and - critically - on what level of accuracy you actually need for the job in hand. "As always, when we're talking about surveys, there's two things you want to understand," says Bob Foley of Engineers With Drones. "Are you surveying land? Are you mapping land? Or are you surveying some form of asset? Are you inspecting an asset?" For this article we are focused on land mapping, where the accuracy question is most nuanced.

Two methods, two different accuracy profiles

Land mapping by drone comes down to two technologies: photogrammetry and LiDAR. Photogrammetry stitches together a large number of overlapping images with high precision to generate a 3D model. LiDAR fires laser pulses from the drone and measures where they return, building up a point cloud. Both produce accurate spatial data - but their accuracy characteristics differ significantly depending on what the drone is flying over.

Hardscape vs softscape - the distinction that changes everything

"When it comes to accuracy, you want to be thinking about are you shooting at hardscape or softscape?" Bob explains. Hardscape is buildings, roads, footpaths, gravel, compacted ground - anything solid and defined. Softscape is tall grass, vegetation, ditches, forestry - anything where the actual ground surface is hidden beneath organic material.

The distinction matters because photogrammetry maps what it can see. If a field is full of corn, the drone sees the top of the corn - not the ground. "To the eye, the topography looks like it's nearly a foot and a half, two feet above what it actually is," Bob notes. That same field cut bare is an entirely different proposition. For hardscape, photogrammetry is the preferred technology and delivers strong results. "We're probably talking, depending on the technology, down to about two centimeters," Bob says of hardscape photogrammetry accuracy.

It is worth being clear about what two centimeters means in context. A terrestrial surveyor with the right equipment can achieve millimetre-level accuracy. Drones cannot match that - a drone is a flying machine and has inherent limitations. But for the overwhelming majority of land mapping applications, two centimetre accuracy is more than adequate.

When the vegetation is the problem - use LiDAR

For softscape, photogrammetry accuracy becomes unreliable because the sensor cannot see through the vegetation to the ground beneath. "We always want to be careful when we talk about accuracy that we're being very clear," Bob says. LiDAR changes this because it finds gaps in vegetation canopies and returns measurements from the ground surface below, producing what is called a bare earth topography - an accurate picture of the ground regardless of what is growing on it.

Define what you actually need first

One of the most common conversations Engineers With Drones has with new clients is about accuracy requirements that do not match the task. "What we see with a lot of our clients is they'll say, I want to map this forest and I want two centimetres of accuracy. And we say, no, you don't." Demanding sub-centimetre accuracy for mapping a vegetated area is not just unnecessary - it is technically meaningless if the sensor cannot reach the ground surface in the first place.

"Nobody needs that level of accuracy to map vegetation. It's not a thing. Now, if you really want to know what's under the vegetation, yes, you do." The right approach is to define the actual engineering or planning requirement, then select the method that delivers what is genuinely needed.