How One Control Point Nearly Ruined a Survey — A Real Error-Budget Investigation

Expecting three centimeters of error and getting eleven is the kind of gap that should stop any survey team cold. In a hands-on ROAR West workshop, surveyor and geodetic engineer Rami Tamimi, founder of The Survey School, walked through exactly that scenario from a real project — a full survey of a 5,000-square-foot house in Hocking Hills, Ohio — and showed, step by step, how his team found the cause.

The math going in was sound. Adding up every expected source of error — base station precision, RTK correction, pole setup, the ground control point marker itself, and processing — Tamimi's team calculated an expected RMSE of about 3 centimeters for their GNSS control. "That's what you should expect with GNSS," he said. "If you get anything below it, good for you. That's awesome. But go into it expecting three centimeters." (watch at 31:44) Instead, when they processed the data, they got results of 9 centimeters, 4.5 centimeters, and nearly 11 centimeters across different datasets. "It's consistent throughout all of these data sets," Tamimi said. "I've got some error that is impacting my data because this is not good data."

Rather than guessing, the team ran a systematic elimination process: pulling individual control points out of the calculation, one at a time, to see whether removing any single point brought the error back down. "We just started pulling random ones out to see if any of them would reduce the error," Tamimi explained. (watch at 32:09) One point — point number one — turned out to be the culprit. Reviewing video footage from the day, Tamimi could see the problem in hindsight: "This area is not going to give us good results. Like, I knew going into this that this was a bad location, but with the geometry, I really wanted something over here."

The technical cause was multipath error — GNSS signals bouncing off nearby obstructions (in this case, trees and the house itself) before reaching the receiver, which extends the apparent travel time of the signal and produces a false position. "We always try to avoid establishing control in areas where there's lots of trees or high-rise buildings, anywhere where data from the satellites can be deflected through multipath," Tamimi said. (watch at 32:49) It's a known risk that's easy to accept in the moment, for the sake of convenient geometry, and expensive to discover later.

The broader lesson here isn't really about multipath specifically — it's about method. Tamimi's team didn't catch the problem because they got lucky; they caught it because they had enough redundant, independently collected data (watch at 34:15) (the same points measured with multiple tools and methods) to isolate a single bad input and test it in isolation. That kind of built-in redundancy is what separates a dataset where errors eventually surface safely, before a client ever sees them, from one where a single bad point quietly rides along into every downstream deliverable.

Survey & mapping solutions · DJI drones for survey & mapping in Canada