UKC

Strava, Komoot, GPS etc accuracy

New Topic
This topic has been archived, and won't accept reply postings.
 elsewhere 01 May 2023

My mate went for a gravel ride, his route on Strava was 68km and 1500m ascent. 

I too went for a gravel ride, my route on Strava was longer at 70km but only 1275m ascent.

The funny thing is, we were riding side by side. His device is wahoo elemnt bolt that he uploaded to Strava and I used Strava on my phone. 

I was surprised by the differences. I'm curious, is that sort of distance (3%) and ascent (15%) discrepancy normal?

 Monk 01 May 2023
In reply to elsewhere:

A couple of years ago, pretty much all the fastest times in Strava around me were on Android devices. Having used both phones and various Garmins, I'm pretty sure the Garmins are more accurate. The ascent discrepancy surprises me a bit - I thought Strava corrected to the mapped route when you upload.

 GrahamD 01 May 2023
In reply to Monk:

I'm not surprised by the elevation discrepancy (or distance if the route was under trees for a part). Elevation correction to maps can only work to the resolution of the map contour resolution.  On a typical ride, you'll often have ups and downs of 3m or so which may be picked up by a barometric sensor but not on the map and also possibility not on GPS, depending on satellite coverage.

 Cusco 01 May 2023
In reply to elsewhere:

I’ve experienced similar discrepancies when riding with my friend who uses a Garmin as against my Wahoo Element Bolt (and I noticed similar discrepancies when I used to run Strava at the same time as a comparison against the Bolt). 

Here’s a thread along similar lines - but this time about discrepancies in the data from my Bolt on exactly same route which I’ve done four times.

https://www.ukclimbing.com/forums/biking/inconsistent_wahoo_elemnt_bolt_ele... 

OP elsewhere 01 May 2023
In reply to Monk, GrahamD and Cusco.

Having included Komoot in the title I forgot to mention it. The GPX export (from Strava on phone) is always 5-10% less ascent in Komoot than Strava. Different data smoothing algorithm and mapping data?

Didn't realise a barometric sensor is used. Maybe the Wahoo file is bigger with more data points picking up more little bumps and my version is lower sampling frequency.

And the route was in trees for most of the way.

I think I will accuse my mate of paying the flattery subscription on Strava

Post edited at 09:44
 nufkin 01 May 2023
In reply to elsewhere:

Have you tried the 'Correct Distance/Elevation' option on Strava? Might adjust things in your favour in this case

OP elsewhere 01 May 2023
In reply to nufkin:

Never heard of that correction option until today. I'll give it a go.

 petemeads 01 May 2023
In reply to elsewhere:

Barometric altimeters should measure small ups and downs better than GPS estimates or mapping corrections - except my Garmin devices are very sensitive to wind gusts. I have data from running the same flattish course on calm and gusty days and the height gain was completely different - you can see a ripple in the elevation data and see height gains when turning into the wind. Garmin reckon to cover the watch with your sleeve to reduce this pitot effect, which negates the solar power benefits. You can't have everything I suppose...

 ChrisJD 01 May 2023
In reply to elsewhere:

Have a look at the actual route traces of each device side by side; you'll see the sampling rate (it is not a curve trace, its series of straight lines between points) and then probably many lateral errors due to GPS searching whilst in the trees (all of the route). 

Standing still in trees can often lead to quite a bit of GPS wandering and 'extra' mileage.

There may also be some big 'errors' in your traces to give that pronounced height difference, so check the elevation view as well looking for spikes.

 ChrisJD 01 May 2023
In reply to elsewhere:

Interestingly, as an example, did a short impromptu MTB ride on Saturday with a mate; we both use the same Garmin watch model, and upload to Strava via Garmin Connect.

Starting/finishing exactly the same place:

His: 13.42 miles with 1,952 foot ascent. 

Mine: 13.41 miles with 1,942 foot ascent.

and

My Garmin Connect app said mine was 13.41 miles with 592m ascent = 1,942 foot.

So Strava upload not changing a lot from the Garmin data

Post edited at 10:25
OP elsewhere 01 May 2023
In reply to ChrisJD:

The horror, on Komoot my route is only 1080m climbing!

OP elsewhere 01 May 2023
In reply to ChrisJD:

That's impressively consistent data.

In reply to elsewhere:

Are start and end points identical?

Do you both have Strava set to snap to ground (using a DEM)?

Are they mounted similarly?

Random errors in position fix all add up. The noisier the signal, the greater the random walk effect. Noise is affected by the mounting; the ability of the antenna to get a good, clear view of the entire sky, horizon to horizon. Covering the antenna, especially with wet fabric, will attenuate the signal, increasing position noise.

Are they both recording at the same rate (consumer GNSS gives 1 position per second, but logs can be set slower than that, to save memory). A lower recording rate will 'cut corners', and reduce the effect of random walk.

Elevation is the least accurate of GNSS measurements, due to the geometry of the constellation and receiver.

 ChrisJD 01 May 2023
In reply to elsewhere:

> That's impressively consistent data.

The ride before (mainly in trees on a steep hillside) with the same mate:

His: 10.18 miles & 2,293 foot. and mine: 10.79 miles & 2,293 foot

(on this ride, I did nip down a short bit of flat road and back to find another mate, so that explains the distance difference).

So your big difference is probably down to different devices, and especially how they behave in trees (which causes a GPS lots of issue)

I went from an Samsung Android phone to a Garmin watch a few years back and I think the Garmin watch is more defo more accurate/consistent.

OP elsewhere 01 May 2023
In reply to captain paranoia:

Mounted very differently - his device fixed flat to handlebars, my phone in my jersey pocket.

I assumed elevation was from lat/long and mapping data for my phone (no rain, fairly dry). I don't think elevation is in my GPX file output and as you say, satellites give relatively noisy elevation.

 ChrisJD 01 May 2023
In reply to elsewhere:

> Mounted very differently - his device fixed flat to handlebars, my phone in my jersey pocket.

That's definitely adding in more variables and room-for-error from your phone there.

 Marek 01 May 2023
In reply to ChrisJD:

> I went from an Samsung Android phone to a Garmin watch a few years back and I think the Garmin watch is more defo more accurate/consistent.

Hmm. I had cause* to look at the GPX files last year and found that my garmin (Edge 200) had typically 5-17s between GPS data points- with no obvious correlation between the interval and anything else - whereas using Locus Maps on Android I could set a fixed 1s interval between data points. For my purposes (at least) the Android solution was much more accurate (at least for lat/long. I used LiDAR DEM data for elevation)

* Chung method for estimating Crr and CdA.

 ChrisJD 01 May 2023
In reply to Marek:

Did you not tell it to do data recording 'Every Second' ?

Edit: just done a quick google check and your Edge 200 unit may not have the 1 second sampling option (best to check, they may have updated that via software).

The Garmin watch I use does have 1s : that was the first thing I made sure it had before buying and I changed it to '1s' straight away after getting it (default is (not so) Smart sampling).

Post edited at 14:48
 Marek 01 May 2023
In reply to ChrisJD:

> Did you not tell it to do data recording 'Every Second' ?

> Edit: just done a quick google check and your Edge 200 unit may not have the 1 second sampling option (best to check, they may have updated that via software).

Quite. the Edge 200 doesn't have that option and Garmin Express says it's up to date.

 ChrisJD 01 May 2023
In reply to Marek:

That's 'somewhat' frustrating for a biking GPS !


New Topic
This topic has been archived, and won't accept reply postings.
Loading Notifications...