Face it, sometimes your accuracy just hits the
iceberg. We have seen error reports within .011 of a foot and we’ve seen them crest
near 0.080 of a foot. The key to obtaining LiDAR accuracy is understanding the components.
When I speak to surveyors about the acronym RMS (Root
Mean Square) error reports I often feel like a captain on a sinking ship. ESRI’s definition is cumbersome: “A measure
of the difference between locations that are known and locations that have been
interpolated or digitized. RMS error is derived by squaring the differences
between known and unknown points, adding those together, dividing that by the
number of test points, and then taking the square root of that result.” The
question is… Known and unknown to who?
Think of it this way: RMS error is the difference between collected LiDAR
measurements and traditional collected control points. Where your accuracy can
tank is when you have an outlier in the LiDAR data set, not unlike a transposed
coordinate number in the field book. But many other variables can affect the
accuracy of the RMS error report.
In February of 2009, we scanned a highway in
Minnesota using the DOT’s control points. The RMS error was off by inches.
Unacceptable even by traditional standards. Sifting painstakingly through the
data by both the DOT and my Terrametrix technicians, it was determined that we
were seeing frost heave. The control was set in warmer fall conditions and the
scan was performed in the winter. Learning curve and when we proved that point
we were all amazed at being able to see this movement.
In the winter of 2016, Terrametrix mobile LiDAR collection
on a mile-long interstate viaduct in the Northern mid-Atlantic region resulted
in a RMS error of .060 compared to 64 control points ranging in a .090 high and
.11 low on the bridge. Control set on
hard ground at the approaches had an .013 RMS error against 34 control points
ranging .037 high .034 low. The RMS error report comparisons were showing bridge
load. Control was shot during rush hour; our scan was performed at the peak of
heavy traffic load on the bridge. Control
was set on the bridge mid-span and not as requested over the piers. You could
tell because the error at each point was directly related to how far the
control was from a pier with the greatest errors being near mid span.
A 2.9-mile road course at Watkins Glen in the fall of
2014 resulted in a RMS 0.012 range .023 high to .021 low against 15 control
points every quarter mile. Not bad for an end deliverable used for re-pavement on
a historical race track. Another DOT project in the fall of 2015 involved 10.6
miles of interstate. With 101 control points the RMS error was 0.014;
ranging .039 high to .030 low. These are
typical RMS highway reports using mobile LiDAR.
I must warn anyone that boasts you don’t need survey
control with LiDAR is like telling you the ship is unsinkable. What is your
quality check? Control is…. will be… and always has been… minimum standards and
prudent operating procedure for land surveyors. Occupy the point. Also, be
warned if the RMS error is too good. Anyone who comes up with a zero RMS error
has adjusted the LiDAR data to the control point. Plain and simple. Every
component of measuring has error, the GPS has error, the scanners have error,
the rodman may have had a bad night. (Come on, we’ve all been there). If your
RMS error is zero head for the life boats.
