Error when using CORS from TxDOT RTN to georeference DJI Phantom 4 RTK and Emlid Reach RS2

Question

I use CORS from TxDOT RTN to georeference DJI Phantom 4 RTK and Emlid Reach RS2. I have been using it in Corpus Christi, TX with no issues. I do Structure from Motion (SfM) in Metashape, verify the check points and the error is within my expected range (2-5 cm).

I recently travel to Weslaco, TX which is 2 hours away from here. I made sure RTN was connected successfully and GCPs where surveyed only when it has a fixed correction. I used a tripod to keep the Reach level and recorded each GCP for about 30 seconds.

For the data collected in Weslaco, reported in Metashape, I am getting an error of 1-3 meters for the check points. We resurveyed the GCPs and the same error was reported twice.

On the Reach we use vertical datum Ellipsoidal height and NAD83(2011)/Texas South as the coordinate system.

What is the source of this error and how can it be fixed? My understanding is that Reach detects automatically the closest mount point and connects to it.

Answer 1

Is it possible to check which CORS station(s) was used? If for some reason it was still using one near Corpus Christi, that would be a sure problem. (You might need to dig into a log file.)

A quick check at https://www.ngs.noaa.gov/CORS_Map/ makes me think you might be farther from a station than you're accustomed to - though I would not think 20km would be that problematic. But if it's possible to re-observe your control points, longer observations might help. Do you have something to view the fix and position in real-time (like a link to a phone app or data collector), or do you have to wait and download from the unit afterwards? If you can view real-time, I would think you should be able to see accuracy of that position before you stop the observation. (i.e. you could continue the observation until you're getting the desired accuracy.)

Answer 2

We performed a third data collection and obtained lower error this time. This time, each GCP was recorded for 1 minute. For some reason the number of satellites that we connect there (16-24) is lower than in Corpus (28-32). I believe recording the GCPs for longer period was the solution. Additionally, the error (m) reported from the images is much lower. Maybe the drone was also able to connect to more satellites this time.