Most of the road courses in this area are very close to what could be called "pancake flat." I tend to use my wrist-mounted GPS unit as an early-warning system of sorts when measuring.

During a recent (10km) project, on a course with a start/end elevation of 11ft, max of 84ft, and min of 7ft - total elevation gain/loss of 323ft, I was occasionally surprised by the GPS (mile) split chime going off past the intermediate split count. "No worries," said I, as I walked my bike back to the intermediate split, set my mark, and measured my relative landmarks.

Looking at my GPS data later it didn't surprise me much to see my total distance at ~6.18. You see, the (figure-of-eight-shaped)course has tight turns and about 40% tree coverage, on top of the elevation changes, too.

Two weekends after the race, a young lady who ran the course mentioned that her GPS unit showed a total distance of 6.2 miles. I began scratching my head about this and did not know whether this could be caused by the typical technological limitations of consumer/sport GPS units (single point-horizontal error) or if it's also because of vertical error.

I read a brief review on DC Rainmaker's site about elevation issues (https://www.dcrainmaker.com/2010/05/understanding-sport-device-gps.html) ...naturally, most "typical runners" don't have units with barometric altimeters.

For giggles, I pulled a Google Earth measure of a section of road used for at least two of the most-popular races here in town. One end is at an elevation of 69 feet, the other end at 56. "Point-to-point" distance was 2270 feet, but when elevation - all that Pythagorean theorem stuff, which I took the time to do for each intermediate point Google gave me - was taken into account, the "point-down-the-hill-up-the-hill-to-point" distance was an additional 3.4 feet. Yeah, that's an added .15%.

Maybe I'm simply barking at airplanes because this is a first-seen for me. Do folks who actually have elevation changes deal with this sort of thing when measuring?
Original Post

Even a 5% grade up and then back down over that 2270-foot stretch would be less than 3 feet longer than if it was flat. So I'm not sure how you are doing your calculation. Even on a hilly course (constant 5% grade would be a monster) neglecting elevation change in the distance calculation results in a negligible error.

How far did you ride past those intermediate points before stopping and walking your bike back? However far that was, double it, and that's the extra distance you put on your GPS. You also added distance to you GPS if you got off your bike to measure the locations of intermediate points relative to landmarks. And also if you froze your wheel and walked your bike anywhere.

My guess is that's most of the reason your GPS recorded a longer distance than your Jones.
Sorry I was thinking your GPS measurement was long rather than short. If your GPS measurement was exactly 6.18 miles then it was 64 meters short of the 10010 course distance, or 0.64%. That is a good bit higher than what I typically saw in my tests. But I didn't stop the watch, walk around a bit, and start it again several times. It's very possible that has a negative effect on accuracy.

That stretch of road you measured has a pretty serious hill. 1160 feet with a 5.3% grade downhill, a 530 foot fairly flat section, and then 580 feet with an average 7.6% grade uphill. That second part is a rise of 44 feet in less than 200 yards!

While some tough courses will include a couple nasties like that, other hills they include generally won't have grades over 5%. And the vast majority of the course will be relatively flat or with hills of less than 2% grade. Overall, I doubt the total error due to neglecting elevation changes would ever exceed 0.05%. Much smaller than other sources of error in a google earth or GPS measurement of a course.
No worries.
I will say that I did not use any part of the course I measured which inspired this particular discussion.

There were a couple of climbs of ~.1 mile with a grade of 8-9% and a couple (one fairly short, the other about ~.1 mile) with grades of around 20%.

I'm almost tempted to invest in a handlebar-mounted unit which would have the barometric calculation, so I'm not staring at my wrist while riding; but also to see if the elevation swings are as drastic on a hilly route.

Again, the overwhelming majority of courses I measure don't have serious elevation change, definitely not like this one.
My experience with a mounted / handheld barometric altimeter is that the Venturi Effect causes wild swings. Equally bad (even if you wanted to stop every so often) is normal Pbar variation. I have NEVER seen ambient barometric pressure remain constant enough to believe what I see within 10'. Barometric altimeters are a marvel of sensitivity. The changes between your chest and belt are obvious if you move it back and forth a few times. My Y2Kish model fluctuates + - 2' with each read. Maybe there are better algorithms on newer models.

If I have a course where eligibility for records is at question because of the drop (rare but it does happen), I will go between the START and FINISH several times getting an average 10 reads 10 seconds apart at each place. When the difference stops changing, then (and only then) do I have confidence in what a prohibitively expensive (both time and money) surveyor would say. A lot of monkey business? Yes, so thank goodness it is rare.

Hold a barometer out a car window at 40 MPH once. This works best on snowy days

Oscar
I'll have to take your word for it, as the odds of snow here are about the same as finding chicken lips at the local Winn-Dixie.
I agree with Oscar. My GPS has a barometric altimeter, and fluctuates with the weather. I can start a course at Point "A", and finish at the same spot, and have a difference in elevation of 25 feet. I don't put much faith in the precision of elevation readings on the GPS. I will use GoogleEarth to get an elevation.
We used one of these in around 1989, even had one built into a watch. The only way we actually used the altimeter data for course elevation was if the start and finish readings were the same (assuming a coincident or near coincident start/finish). I still have it, if anyone's interested.