embedding gmap4

[retired jerry]

I record a track with GPS

Then I manually remove 90% or more of the points - the little clouds of points at stops, short side trips to viewpoints, obvious error points. Usually I'll have the track going some place, and then the return track to help.

Then I walk the trail again and make sure I didn't remove actual switchbacks

I think that gives a fairly accurate mileage for a trail

[raven]

Jerry, I see 5 possible flaws on the accuracy of the track and distance computed by your method right off the bat.

(1) If the recording session is done as one track, the GPS unit may use the info recorded on the way out to modify the track on the way back for "customer comfort".

(2) Errors from GPS satellites vary over time. One day's track will differ from the next if not corrected to a known location.

(3) Corners on switchbacks will be missed if the GPS is not taking a location at the exactly correct moment.

(4) The averaging method used in the point recording process, as indicated by the straight recorded tracks in the absence of turns and stops in the Warren Lake photos, imply some recorded positions are approximations at best. Little variations in distance and direction and height that are actually read by the GPS seem likely to be missing by design.

(5) Even if the track is measured from points you have carefully cleaned up for horizontal errors, the vertical excursions will remain to be computed as part of the distance between points.

(Maybe the first reflects paranoia, but GPS units do report elevation gain measured, not total elevation gained, apparently for "customer comfort".)

A GPS unit does not better a bicycle wheel for measuring distance -- on trails. (Amended to reflect use on trails, with thanks to the next post.)

[potato]

Well, he did say "fairly accurate". I guess the question is what's "fair". Plus or minus 10% just so you have a rough idea how far you traveled? In that case I don't feel like your 5 flaws will pose a significant problem. If you need to know the length of a 50+ mile trail accurate to 1/10 mile, I agree there are problems. If your trail consists of 1000 switchbacks that are 20' apart I agree that there are problems but that's pretty unrealistic.

Also I believe your statement "A GPS unit does not better a bicycle wheel for measuring distance" needs some qualifiers
How about if I want to measure the distance from Portland to Hawaii...?

[raven]

Let me add the words "on trails" to that sentence.

Now, consider the meaning of the term "accurate" in measurement in contrast to "biased". I have always taken the term accurate to mean equal odds of a higher or lower true, but unknown, value. The descriptor "more accurate" meaning a lower variance. Biased is, well, off the mark. A calibrated bicycle wheel is accurate on a smooth trail. A GPS unit is biased upward.

What do you mean by the term accurate?

[CampinCarl]

Super geeky alert

Accuracy Assessment of Recreational and Mapping Grade GPS Receivers, Journal of the Arkansas Academy of Science, Vol. 63, 2009

See Table 2, Figure 4, Figure 5

There is "mapping" accuracy and then there is "recreational" accuracy for things like hiking, which doesn't have to be a 15 cm bullseye Also note that they averaged 30 points at known locations. That is much different then collecting 1 point every x seconds in your track as you are hiking along.

I'm not even getting into what surveyors use, those guys are intense about accuracy!

Edit: One last note, this study is a few years old and there are slightly better accuracies these days. Sometimes the manufacturer will give you a decent estimate about what accuracy you can expect in the manual or a factsheet.

[Joseph Elfelt]

Interesting read. Thanks for posting.

There are a number of benchmarks near where we live (they're everywhere) and the horizontal location is known with high accuracy. I plan to use those known locations to compare my older Magellan GPS to the GPS in my iPhone 4s. Just have not got to it yet. I'm curious to see my results compared to this earlier study.

[potato]

Let me add the words "on trails" to that sentence.

Now, consider the meaning of the term "accurate" in measurement in contrast to "biased". I have always taken the term accurate to mean equal odds of a higher or lower true, but unknown, value. The descriptor "more accurate" meaning a lower variance. Biased is, well, off the mark. A calibrated bicycle wheel is accurate on a smooth trail. A GPS unit is biased upward.

What do you mean by the term accurate?

I guess if the trail is extremely rocky (parts of the Loowit trail come to mind), a bicycle wheel wouldn't work so well, but I get your point

I think of it like there are 2 big things that make GPS tracks inaccurate. (1) The fact that it takes a measurement every 5 or 10 seconds (?) and then connects it with a straight line will bias the distance to be lower than reality, since there are a lot of small turns and variations in a trail and it's not a series of line segments. Then (2) the positional noise, which biases the distance to be higher than reality. I think (2) usually overpowers (1) but maybe there are some interesting switchback-intense trails where (1) causes so many corners to be cut that the overall distance recorded is lower than reality. Also, if it were true that GPS receivers are always biased upward wouldn't they artificially adjust distances down?
It must be a challenge for Garmin engineers to design one set of rules (for smoothing tracks or whatever) that will be optimal for all kinds of activities. Or maybe they do handle a hike differently than a plane ride.

I think until we have GPS receivers that are accurate to less than a few inches, never lose signal, and take multiple measurements every second, it can't compare to a bicycle tire on a smooth trail.

I was using the word "accurate" more loosely to mean "within some acceptable range of the actual value".
I'd usually describe your "bias" as a DC offset