RTK Under the Canopy: A 0.18-Ft Disagreement

Hold the total-station distance, 3,000.02 ft, and re-observe A before you finalize anything.

Reasoning:

• Boundary work is about relative accuracy, not absolute position. The controlling question is the vector between A and B, not where either sits in NAD83(2011). A closed traverse that closes 14" and 0.06 ft over 4,100 ft (about 1:68,000) is doing a good job on that vector.
• The RTK discrepancy is too large to be normal. For a 3,000-ft baseline with a short, verified base line, RTK relative error should be on the order of 1 cm + 1 ppm, i.e. ~0.02–0.04 ft. A 0.18 ft difference is four to nine times that.
• The repeated occupation is the tell. Two occupations of A 0.11 ft apart in easting means the position is not repeatable at the level you are claiming. That is a solution problem, not a noise problem.
• RMS is not accuracy. A receiver's reported horizontal RMS describes the internal consistency of its own solution. A mis-fixed integer ambiguity or a multipath-biased solution can report a beautifully small RMS and still be 0.1–0.3 ft wrong. The canopy occupation - 40 s, under heavy foliage - is exactly the environment that produces confident, wrong fixes.

Ranked causes:

1. Multipath and a marginal fixed solution at A (most likely). The canopy generates reflected signals; a 40-second occupation is short enough that a bad ambiguity set can survive.
2. Incorrect antenna height or antenna model at the rover - a 0.11 ft error is 0.034 m, within the range of a slip or a phase-center model mismatch.
3. Poor satellite geometry at the time of the observation.
4. A base coordinate problem - note that a base position error translates the network nearly uniformly and so affects absolute position much more than a 3,000-ft relative vector; it is a weak explanation for the A–B discrepancy but a strong one for any datum tie.

Procedure:

• Re-observe A with a 5–10 minute occupation, off the same point, at a different time of day so the satellite geometry changes.
• Reinitialize the rover between occupations and require a fresh fix; do not accept a re-fix without reinitialization.
• Verify rover and base antenna types and heights against the field notes and the receiver log.
• Move the rover a few meters and return, to test repeatability under the same canopy.
• Tie A and B with the total station from the verified control, and check the RTK against that independent vector.
• Only after the RTK position of A is repeatable within your project tolerance do you publish coordinates or set monuments.

Only with caution, and not for the design grades.

Diagnose the bias first. A uniform vertical offset has a small set of usual suspects:

• Antenna height or antenna phase-center model error (verify against the log, not memory).
• Geoid model version, units, or a mismatch between ellipsoid and orthometric heights in the export.
• Base station height or base coordinate error.
• A wrong vertical datum tag in the project.

Can a constant shift be applied? Yes, but only if you have demonstrated that the offset is constant across the site. Two benchmarks are a start; you need checks at the site's vertical extremes, because the geoid slope and any scale issue vary spatially. A correction derived at one end of a site and applied at the other can easily be a third of the error again.

For the drainage design, do not rely on RTK elevations at all. A 0.4 percent grade produces 1.2 ft of fall in 300 ft, so a 0.28 ft error is nearly a quarter of the total fall and would invert the design intent. Carry vertical control for pipe inverts by differential leveling or a total-station traverse from a verified benchmark, and use RTK for topographic mass data only, with the bias documented in the project report.