GPS questions

[Jason Stratton]

I look at ride stats quite often when I finish a ride and compare them with others.  The question I have is that my GPS (Garmin Edge 500) is usually much lower in elevation gain than everyone else I ride with...by a substantial amount (10-15%).  Does anyone know why this might be?  Furthermore, when I map my route on Bike Route Toaster or Ride With GPS I get numbers that are equally off...the numbers even differ between the two sites.  Why or how does this happen?

Do the rest of you just your statistics as relative comparisons for yourself?  I find it difficult when I compare a 50 mile ride with 3,000 ft elevation gain from my GPS to a friend's 50 mile ride with 3500 ft gain to decide if they actually climbed 500 more feet.  Any comments from tech guys out there? 

[karlos]

Usually, the Garmin GPS unit's reading  is the best. I have an edge 305 and when I compare with others after a ride we are usually within a couple hundred feet, or about 5%. When a route is laid out on a mapping site, it all depends on the underlying database of elevations (DTED maps) which are used. I have found that Ride With GPS is often significantly higher, maybe 20%, although that is my favorite site for manual route construction. Even given the raw GPS data or the elevations from the database, it can then vary depending on how the ups and downs are smoothed (filtered is the technical term). If you use the program SportTracks for recording your rides, it gives you different filtering options so you can see the effects. It can easily change the answer by 15-20%. You can even load the raw gpx data (as an xml file) into a program like Matlab, and do the filtering manually. You will easily notice 5-15% variation depending on how you choose to filter the discretely sampled data.

I also worked on the GPS satellite system in the 1980's and it is well known that elevation is the noisiest estimate, primarily because you can only surround a point on one side of a hemisphere, which creates a bias, especially when including atmospheric propagation effects. Horizontal position is much more accurate by a factor of about 2 to 4 depending on satellite geometry - see this note. This is why it is so important to filter the data and remove the noise. You can also configure your gps to sample more often, like every second, if you really want to be anal about trying to get the best answer, but then you can run out of memory - not recommended - even if you do, there is even more total noise which requires even a different filtering approach to get close to "truth".

Like the old Chicago song, "Does anybody really know what time it is?", I have stopped worrying about this and just consistently use my Garmin to gauge differences between rides. I know that a 6000' ft ride really is about 50% more climbing than a 4000' ride and try not to get too hung up on absolutes. For sure, your friends using a  different device are not really climbing any more or less than you on the same route, ignoring paperboying  
-Karl

[Jason Stratton]

Thanks for the response.  I will continue to use it for gauging relative differences between rides.

[karlos]

I continually note that RideWithGps seems to be about 20% higher than my Edge 305. The programmers for RideWiithGps have a nice straightforward discussion on this issues page: http://ridewithgps.com/issues/3. The  user comments and responses, including comparisons of various mapping utilities and units, offer insight into what's going on and why you shouldn't get too hung up on absolutes. The RideWithGPS guys stand behind their elevation, so I guess I should feel 20% stronger.

[JohnnyB]

I also have a Garmin 500 and it runs a little higher than others.  I was told that using barometric pressure, as the 500 does, to calculate elevation gain or loss, rather than the standard method of distance from satellite, gives a more accurate reading.  This may help explain the discrepancies as well.

[karlos]

The Garmins, by default, use the barometric reading in a clever combination with the satellite data. The main problem with barometric readings is calibrating by setting against a known elevation. Garmin uses a nice algorithm of using the satellites to assist with the calibration and the barometric reading then is smoother than the raw satellites' solution. The actual calibration point is actually irrelevant for total elev gain (and loss) for a ride as everything is relative to your starting elevation - if that is even off hundreds of feet, this bias subtracts out when differencing to determine gain or loss. Even with the barometric assist, however, it is still the noisiest measurement. The filtering technique, which takes all the discrete differences, and smooths them out, has the greatest effect in comparing measurements between devices and mapping sites.