Elevation Variation of 5m Despite 100% PPK Fixed Solution in drone data in Emlid Studio

I am conducting a survey over a 67 sq. km area using Emlid Studio and achieving a 100% fixed PPK solution. Due to inaccessibility, I am not using GCPs, but I have marked four validation points to verify the accuracy of my dataset.

For elevation processing, I have applied the EGM2008 geoid model to convert ellipsoidal heights to Mean Sea Level (MSL), aligning with my local vertical datum. However, I am observing an elevation (Z-value) discrepancy ranging from 2m to 5m, which is beyond the expected centimeter-level accuracy.

I would appreciate any suggestions on how to minimize this elevation error

Hi @patelmalhar1996,

Before giving some suggestions I would like to understand better the issue, could you share how you determined the elevation discrepancy of 2m - 5m? Are you comparing the results to known control points, or did you re-survey the same locations and notice the difference?

It would also help to know more about your base station setup:

  • Are you manually entering its known coordinates or averaging them in Single or Fix mode?
  • If using Fix mode, is your base receiving corrections from an NTRIP service? If so, which one?

Also, are all datasets using the same vertical datum? If possible, please send over the files of the project so I can take a closer look. You can share it here or send directly to our support email address: support@emlid.com.

1 Like

Thank you, Victor Ferreira, for your response.

I will outline step by step what you are asking:

  1. We set up our own base station, took the averaged value of that station, and matched it with a known point of that elevation (which corresponds to the top of the dam).
  2. Yes, we are comparing the results to known control points. We first flew the drone, collected raw data in PPK mode, brought the base and rover (drone data) to the office, processed PPK, and obtained a 100% fixed solution.
  3. We first set up the base, performed single-point averaging, and used that reference point as the survey point for the entire area. We also connected our base station to NTRIP, but it showed a 1-meter variation in the elevation of the top of the dam. Due to this discrepancy, we are using our own base setup points, which match the known point, for the survey.
  4. We have processed the data without GCPS to evaluate the reliability of 100% fixed PPK datasets with elevation values.
  5. I am attaching a link to the data, where you will find the following:
  6. Base file
  7. Rover file (Drone Data)
  8. Drone raw images
  9. Known DGPS point (for comparing the fixed PPK dataset DTM with the known elevation)
  • Link to base file, rover file and Known verification point : Emlid - Google Drive
  • Link to Raw data of drone : m2 - Google Drive
    Note : For this particular mission, we achieved a 100% fixed solution. However, we are unable to match our point cloud elevation values with our known DGPS points. Avoid DGPS points which are not following inside of the given raw datasets

One item I’m missing here is determination of the base point. Did you use OPUS or other post processing techniques to determine the true coordinates of the point in reference to the existing control ? You say that you tied to existing control and have an error of 1 meter ?

If this is the case, it sounds like the base point used single averaging and the point was not accurately referenced to the existing control you tied into Am I missing something here in your explanation ?

We have used the averaging method to set up our base station.

We then verified our base station with multiple verification points on the dam, such as the water level and the top of the dam, which matched the available data. Our base setup is correct, as the verification points align with the data.

However, the problem is that I am not getting accurate elevation values from the 100% fixed drone aerial datasets after processing them in AGISOFT. (I have already discussed the processing steps in my previous question.)

Are they random in nature or?
How do you determine the offset in Metashape?

Hi @patelmalhar1996,

Did you use Averaging Single, Float, or Fix for this?