A base on an autonomous position
The surface comes out internally consistent but shifted in horizontal and vertical. Fixed by occupying the base against NTRIP or by static post-processing, never by accepting a standalone fix.
Accuracy method
How we set ground control and verify accuracy
The number people ask about first is accuracy. On a drone LiDAR survey, accuracy is a product of the ground control plan, not of the sensor alone. A capable sensor flown without disciplined control produces a surface that looks clean but is shifted, internally consistent yet wrong in absolute terms. This page describes exactly how we set ground control and verify accuracy on every project, so an engineer or surveyor can judge the method rather than take a number on trust. If you are new to the method overall, start with our guide to what a LiDAR survey is.
Accuracy on a LiDAR survey is built in layers: the base station georeferences the flight, ground control points confirm the surface sits where it should, and independent checkpoints, held back from processing, prove the result. Skip any layer and the reported figure stops meaning what it should. The sections below walk through the equipment, each step, what we report, and the standards we work to.
Ground control points set and checked with an Emlid Reach RS2, then carried through to the processed terrain surface.
Hardware
The equipment
We fly the DJI Matrice 400 carrying the Zenmuse L3 LiDAR payload, an enterprise sensor with long-range detection and high-density point cloud capture. Positioning is handled by an Emlid Reach RS2 multi-band GNSS receiver. The Reach RS2 does two jobs: it serves as the base station set over a point on the site, broadcasting RTK corrections to the aircraft during the flight and logging raw observations for PPK, and it serves as the rover, occupying the ground control points and the independent checkpoints used to verify the delivered surface.
Step one
The base station
We set the Reach RS2 as a base over a point on the site. The base does two things. First, it feeds RTK corrections to the aircraft during the flight, so every laser pulse is positioned in real time against a known coordinate framework rather than a standalone GPS fix. Second, it logs raw observations throughout the flight, which lets the trajectory be post-processed with PPK where the cell signal is poor or the correction stream drops mid-flight. The base is what georeferences the flight. Without a tied-down base, the aircraft is navigating accurately relative to itself but not to the ground.
Step two
NTRIP and the base position
The base position itself has to be established, and this is where shortcuts quietly ruin surveys. Where a cell signal is available, we connect the receiver to an NTRIP correction service and occupy the base point against the state reference network, resolving its position to centimeter accuracy before the flight begins. Where there is no cell coverage, or the site is remote, we log a long static session at the base and post-process it against continuous reference stations after the flight. Either way the base is tied to a known coordinate framework.
A base left on an autonomous position produces a surface that is internally consistent but shifted in both horizontal and vertical, and this is the single most common way a drone survey goes wrong. The data looks plausible and the error only shows up when it is checked against independent control, by which point the flight is long over.
Step three
Ground control points
On a LiDAR survey the base provides the georeferencing, so ground control points are not doing the job they do in photogrammetry, where they carry the block adjustment. Here they are used to confirm the surface sits where it should and, where a small vertical shift is needed, to apply it. We use targets the sensor can resolve cleanly and distribute them across the site, on the perimeter and through the interior, rather than clustering them in one convenient corner. A cluster checks one part of the site and tells you nothing about the rest. The same approach applies whether the deliverable is a topographic survey surface or a construction earthwork surface.
| Site size | Typical control and checkpoints |
|---|---|
| Under 50 acres | 4 to 6 control points, 3 to 5 checkpoints |
| 50 to 200 acres | 6 to 10 control points, 5 to 8 checkpoints |
| 200 to 500 acres | 10 to 15 control points, 8 to 12 checkpoints |
| Over 500 acres | Distributed across the site, with perimeter coverage and a density agreed in the quote |
Terrain, canopy, and the accuracy the deliverable requires all move these numbers. The plan is stated in the quote before we mobilize.
Step four
Independent checkpoints
This is the section that matters most. Checkpoints are surveyed the same way as the control points, with the rover against the same base, but they are deliberately held back from processing. The surface is never fitted to them. After processing we compare the surface to the checkpoints and report the residuals, point by point.
A reported accuracy figure means nothing unless the points it was measured against were excluded from the adjustment. If the same points used to fit the surface are then used to check it, the residuals only describe how well the surface conforms to its own inputs, and the number collapses toward a tautology. This is what separates a verified accuracy from a manufacturer specification. The sensor datasheet tells you what the hardware can do under ideal conditions; the checkpoint residuals tell you what the delivered surface actually did on this site, against positions the adjustment never saw.
The accuracy report
What we report
Every project ships with an accuracy report. It lists the number of control points and checkpoints, their coordinates, the vertical residual at each checkpoint, and the resulting RMSE. We state the delivered figure of 0.6 to 1.0 in (1.5 to 2.5 cm, or 0.05 to 0.08 ft) vertical accuracy with RTK positioning and ground control, and we show the evidence behind it rather than asking you to take the number on trust. If a checkpoint residual is larger than expected, the report says so and we explain why, whether the cause is canopy, a soft target on loose ground, or a control issue we caught and corrected.
Datums
Coordinate systems and datums
We tie to the project control when it exists, so the delivered surface drops into the drawing set your team is already working in. When no project control exists, we default to NAD83 State Plane horizontal with NAVD88 vertical, using the geoid model the project calls for. If your drawings run on a different datum, a local grid, or a site calibration set by another surveyor, we tie our collection to it and deliver in that framework. Converting a delivered surface to a different datum after the fact is avoidable work, so we ask clients to state their coordinate system and vertical datum with the scope.
Reference documents
Standards we work to
ASPRS Positional Accuracy Standards for Digital Geospatial Data, Edition 2, Version 2 (2024)
Defines how positional accuracy is computed and reported for digital geospatial data.
USGS Lidar Base Specification
Defines point density and classification practice for public lidar.
We report accuracy in the terms these documents use, so the RMSE and checkpoint residuals in the report can be checked against them rather than read as a marketing figure.
Intended for planning, design, and quantity verification, not for certification.
Honest limits
What can go wrong, and what we do about it
Control clustered in one part of the site
One corner is checked and the rest is assumed. Fixed by distributing control to the perimeter and the interior so the check covers the whole flown area.
Dense canopy with low ground point confidence
Pulses cannot reach the ground through solid evergreen canopy, and the surface starts interpolating across the gap. We flag those areas in the deliverable rather than letting the surface interpolate across them silently.
Checkpoints used in the adjustment
This makes the reported accuracy meaningless because the surface was fitted to the same points. We do not do this, and we say so in the report.
Common questions
Frequently asked questions
How many ground control points does a drone LiDAR survey need?
It depends on site size, terrain, and the accuracy the deliverable requires. A site under 50 acres typically carries 4 to 6 control points and 3 to 5 checkpoints, while 200 to 500 acres carries 10 to 15 control points and 8 to 12 checkpoints. Control is distributed across the perimeter and interior rather than clustered, and the final count is stated in the quote before we mobilize so the plan is agreed up front.
What is the difference between a control point and a checkpoint?
A control point is used in processing to confirm or fine-tune the surface. A checkpoint is surveyed the same way but deliberately held back from the adjustment, so the finished surface is compared against positions it was never fitted to. The checkpoint residuals are what make a reported accuracy figure meaningful. Without independent checkpoints, the number is a sensor specification rather than a verified result.
Do you need cell service to fly with RTK?
No. The aircraft receives RTK corrections from the base station over radio during the flight, so cell service is not required for the aircraft itself. Cell service is only needed to establish the base position against an NTRIP network. Where there is no cell coverage, we log a long static session at the base and post-process it by PPK, then fly the same way.
What accuracy can you actually deliver?
We deliver 0.6 to 1.0 in (1.5 to 2.5 cm) vertical accuracy with RTK positioning and ground control, verified against independent checkpoints held back from the adjustment. The figure achieved on a given project depends on control quality, site conditions, canopy, and distance from the base station. We report the vertical residual at each checkpoint and the resulting RMSE with the deliverables, so the number is measured, not assumed.
Can you tie to our existing project control?
Yes. When a surveyor has already established control on site, we occupy their points with the rover and tie our collection to their coordinate framework, so the delivered surface drops into the existing drawing set with no transformation on your end. Tell us the coordinate system, vertical datum, and geoid model with the scope. If no project control exists, we default to NAD83 State Plane horizontal with NAVD88 vertical.
What makes a drone survey survey-grade?
Three things, not one. A base station set and verified on the site rather than an autonomous position, ground control points distributed across the site rather than clustered, and independent checkpoints excluded from processing so the reported accuracy can be checked. A manufacturer specification describes the sensor under ideal conditions, while survey grade accuracy reports what was achieved on that site on that day. That is what makes the deliverable engineering grade data an engineer can build on.
Why does drone survey accuracy change from one site to another?
Because accuracy depends on how many laser pulses reach the ground, and that changes with what is above it. Open ground returns the tightest numbers. Closed evergreen canopy returns the loosest, because fewer pulses find a gap. This is a physical limit rather than a processing setting, so we state what a site will support before mobilizing. Where tighter numbers are needed under cover, denser flight lines and additional ground control points are the remedy.
Scope a project
Ground control is what turns a dense point cloud into a surface an engineer can build on.
For current rates, see our pricing. To scope a project, including projects in the Kansas City metro, request a quote.