Garmin connect Vs Strava Vs true elevation data

theBlob

Bokeh master
Sep 28, 2011
2,863
1,450
129
...
#1
I have noticed that there is a discrepency between Garmin connect and strava.

In Speed averages and elevation data. Why is that?

I also noticed that the garmin does not produce accurate data for elevation gains. It seems better if you do the corrections on garmin connect.

On my ride this morning (laps of my local circuit I have) Garmin connect has my elevation as steadily dropping every lap. If I enable the correction feature That steady drop is corrected. So it seems more accurate.

The result is an addition of vertical meters. On yesterdays ride without corrections It recorded 729M vertical with the corrections it had me climbing 950m.

I'll take the 950m but is that accurate?

Last weeks Green line ride is more dramatic. 1200m uncorrected, 1835m corrected!! It would seem that garmin is basically admitting that their unit is useless for recording elevation gain? a 30% discrepency seems a little absurd even for garmin.

Strava of course is not corrected and pulls the data straight off the unit.

Any info on the accuracy of these readings?
 

Malte

Maximum Pace
Sep 26, 2011
496
54
48
Tokyo
#2
Depends on your Garmin, if you have a barometric sensor then this is usually more accurate than the autocorrected value (which is way off especially for tunnels, and non US) So sorry but the lower value might be the more accurate one.
 

theBlob

Bokeh master
Sep 28, 2011
2,863
1,450
129
...
#3
ha ha ha! I don't really care just trying to work it out. And definately it isn't more accurate for the curcuit ride I did.
 

Malte

Maximum Pace
Sep 26, 2011
496
54
48
Tokyo
#4
ha ha ha! I don't really care just trying to work it out. And definately it isn't more accurate for the curcuit ride I did.
You can read about "barometric drift" to get some background info. The sensor is also temperature sensitive, so if you do a short ride and the bike comes from your warm house there is some error, especially in the beginning, that is what you saw.
 

joewein

Maximum Pace
Oct 25, 2011
2,429
874
133
Setagaya, Tokyo
joewein.net
#5
It's an interesting software problem. Basically any site that aggregates GPX uploads has access to a large number of data points for identical coordinates that it can compare for elevation. It could then correct errors in either barometric or satellite-based measurements. Without these corrections, measuring errors can significantly falsify the total elevation gain. You don't really want to count data noise as climbing!

By comparing logs, the site should be able to build a comprehensive database of elevation of points along any routes logged, so that even users with non-barometric loggers should be able to benefit from more accurate data logged using hardware with barometers.

If data errors were more or less random then some kind of bell curve should evolve for how many data points fall into a given elevation bracket, clustered around the correct value.

There are a few complicated cases, such as roads crossing each other on bridges, but these can be taken care of too by also checking nearby data points (i.e. preceding and following the bridge).
 

bloaker

Maximum Pace
Nov 14, 2011
1,535
1,207
433
Miura, Japan
#6
I use a Garmin 305 unit. I can adjust the frequency of logging datapoints. The greater the frequency, the more the battery use. The less frequency, the less accurate the readings. I am always amused when I start and finish in the same spot, but my garmin says I am actually 50+ ft elevation difference. I use this unit more for running than cycling, so I am not too concerned with accuracy and more concerned with battery life... so I just accept it.
 

AlanW

Maximum Pace
Jan 30, 2007
1,214
435
103
Tokyo
#7
Useful function

The Garmin 500 has a neat feature where you can set a number of points with a known elevation. So if you regularly start from the same places (home, or a particular station) you can save these places and when you start the timer at these places the unit will auto-calibrate its altimeter to the saved value.
 

Malte

Maximum Pace
Sep 26, 2011
496
54
48
Tokyo
#8
I use a Garmin 305 unit. I can adjust the frequency of logging datapoints. The greater the frequency, the more the battery use. The less frequency, the less accurate the readings. I am always amused when I start and finish in the same spot, but my garmin says I am actually 50+ ft elevation difference. I use this unit more for running than cycling, so I am not too concerned with accuracy and more concerned with battery life... so I just accept it.
The "Edge 305" has a barometric sensor while the "Forerunner 305" hasn't. If you use it for running than I guess you have a Forerunner. The altitude comes from the GPS which is naturally very in-precise. The frequency shouldn't affect the accuracy.
 

joewein

Maximum Pace
Oct 25, 2011
2,429
874
133
Setagaya, Tokyo
joewein.net
#10
The altitude comes from the GPS which is naturally very in-precise.
This is actually due to a trade-off: When more than the minimum number of satellites is visible (I think 4 are required), most receivers pick one as much vertical as possible and three as close to the horizon and spread out as close to 120 degree angles as possible.

That selection enables the most accurate latitude-longitude triangulation (where you are on a 2D map), but unfortunately it's less than optimal for elevation. In fact, it has a higher elevation error rate than a set of satellites all higher above the horizon.

So really the GPS should use a different set of satellites for altitude and lat-lon calculations, but most don't.
 

bloaker

Maximum Pace
Nov 14, 2011
1,535
1,207
433
Miura, Japan
#11
The "Edge 305" has a barometric sensor while the "Forerunner 305" hasn't. If you use it for running than I guess you have a Forerunner. The altitude comes from the GPS which is naturally very in-precise. The frequency shouldn't affect the accuracy.
Yup, I do use the Forerunner 305.
The frequency should not affect the accuracy of the Elevation at any onegiven point, however if you do two laps on a route, it may be affected. I live in a very flat region. If I set the frequency to 1/4 mile, then on the first ride, i may miss the two high rise bridges on my route. simply because the points are at the base of the bridge at both sides. The next time, it may hit at the peak. This is going to give me very different results to compare my elevation on a given route. Not typically an issue for me, because I typically ride the exact same routes and I don't have any reason to really look at elevation. If I was hitting up a mountain ride, different story.
 
May 22, 2007
3,595
1,422
143
Kawasaki
halffastcycling.com
#12
Ride with GPS wrote to me today:

Having issues with elevation calculations? The fix is here!

Some of our users, depending on location and device, have experienced inconsistent or inaccurate elevation calculations. Starting on Thursday (that's 15th May 2014, futurepeople) we will begin using a new elevation calculation method that provides more accurate figures. Additionally, starting Thursday we will begin updating all existing routes and rides to use the new calculations.

Most routes and rides will not significantly change. This change will primarily affect routes which had obvious over-estimates for elevation gain, as well as some rides that came in under what a Garmin may have reported. We strive to provide reliable and consistent elevation figures and this improvement is a significant step in the right direction. Our goal is to have a planned route match within 10% of what a quality GPS with a barometric pressure altimeter would record on the same route. Please note due to the inconsistent nature of elevation data, we cannot always match the numbers other services or your Garmin might display. Some uploaded rides have bad data, and some planned routes might be in areas without quality elevation data coverage.

Again, the changes are rolling out Thursday and it will take several days to update all existing routes and rides using the new calculations. If you have any questions or issues please let us know, we are happy to help.

Thank you for your support!
- The Ride with GPS team​
 

AlanW

Maximum Pace
Jan 30, 2007
1,214
435
103
Tokyo
#13
I'll be interested to see what numbers the revised Ride with GPS generates. They have always been waaaaaaaay out for me for routes in Japan; usually a 40% overestimation for climbing and descending.
By contrast, @Naomi uses Route Labo to generate routes and it consistently comes in with total ascent/descent prediction within 100 m elevation of the figures from my Garmin(s) after the ride.

AW.