Elevation discrepancy

I have flown multiple missions with a Mavic Pro and Mavic 2 Enterprise Dual and am generally happy with the results from WebODM.
However I notice that the elevation values are nowhere near correct. I flew a field for DSM and DTM images and the results showed about 30m’s lower than Google Earth so I went back to the field with my phone and it was a lot closer to Google earth elevation than the DSM results.
I did not do GCP’s as I didn’t have my GeoXH with me at the time, and it was more of an experimental flight anyway.
Any reason for the huge discrepancies ?

The GPS values recorded by the drones have an accuracy of +- 15m. The Z axis is even less accurate. The only way you will be able have an accurate Z axis is to use GCPs. Without that GPS data, your maps will not be accurately georectified.

The altitudes you see from your drone telemetry are from a barometer, and are only accurate in reference of the home point at takeoff.


Standard GPS (non-RTK) has normal Z outputs that vary by as much as 150 FEET and still meet the GPS standard.

With dual frequency GPS, and other tricks, you can get vertical accuracies to less than 10 feet. It some cases. under a meter.

I would have thought that my drone receiving 13 sat’s would have better accuracy that 150 feet. And if that’s true, I should have a lot more problems planning flight routes for mapping, especially when trying to calculate pixel size.
I will embarking on doing some flight with GCP’s to see how much better it is. I notice that othophoto overlays on google earth is always out by a few meters, and I’m hoping this will correct that, as I am doing shoreline erosion modelling and that overlay is a great start for clients if it was more accurate.

FWIW, Google itself may be off by a few meters, so take those differences with a grain of salt.

GPS in the Z is pretty awful without corrections. Even with lots of satellites, the accuracy is limited by atmostpheric effects when those aren’t corrected for in any way. Lots of satellite signals being distorted in similar ways aren’t going to give better data. Also a good distribution of satellites for better Z accuracy is typically worse for horizontal accuracy. Ground control and or RTK/PPK are the ways around this.

1 Like

We’ve touched on this in another thread, as well:

As Steven alluded to, Google Earth imagery isn’t truth data, so don’t rely upon it to align features. You’re going to need survey, monument, or GCPs to align your data to.

Unfortunately, Z-Accuracy on GPS isn’t great, which is why most sensor platforms throw it out in favor of data from Barometers or other sources when available. So, absolute accuracy (height above datum) is going to be pretty poor, but it tends to be fairly self-consistent and relative accuracy will be acceptable.