Elevation Correction

by pkan

Downloads: 64,344 (30) • Reviews: 11

Version:

1.9

Updated:

Nov 29, 2017

Accuracy of the "elevation correction "plugin

Accuracy of the "elevation correction "plugin

Postby florent.sourbier » Mon Jan 23, 2012 7:00 pm

Dear all,

It was well known in the French community of runners and especially ultrarunners (trail running) that the calculation of the elevation (and cumulative elevation) of GPS tracks is a crucial question.
Two methods are commonly used
1] Use a threshold (typically 10 meters) to filter estimation errors of the elevation of GPS that are not equipped with a barometric altimeter or use smoothing techniques.
2] Use the "elevation correction" plugin which seems more efficient since based of the most complete high-resolution digital topographic database of Earth.

In found in the free version of SportTrack 3 that is is possible to "display elevation correction activity detail page".
But with several examples with my Garmin Forerunner 305 (in France), i am really surprised by results!

Note that in France, we have the National Geographic Institute (IGN) which provides us French maps with very precise elevation (few meters) at strategic point.

Especially with the plugin, I can not find the max elevation available in my GPX file corresponding to a well known elevation given by the IGN maps.

I illustrate this on two examples:

1] Cross of Garlaban
Click on picture to enlarge
Image
IGN elevation: 714m
Original elevation (GPX file): 712m
Max elevation in SportTracks with STRM 1": 676m

2] Estats peak
Click on picture to enlarge
Image
IGN elevation: 3143m
Original elevation (GPX file): 3146m
Max elevation in SportTracks with STMR 3" Viewfinder Panoramas: 3099m

On both examples, the elevation given by SportTracks is less than about thirty meters.
I asked other French runners to do this on their own GPS tracks across France, same conclusions...

As mention the "elevation correction" plugin description, this plugin corrects the elevation of GPS tracks using SRTM elevation data, which is automatically downloaded from the official SRTM download server.

The French Wikepedia page about SRTM (but not the english one) says
the data were deliberately altered (resolution of 30 meters behind) before being released for free access. Initially, the government of the United States had even blocked their distribution, referring to the terrorist threats to their interests in the world.
(...)
We must also take into account the precision of + / - 20 m in planimetry (X and Y) and + / - 16 m in Z (elevation) for SRTM3.


This could explain the "thirty meters" of difference.

To conclude: since the STRM accuracy is lower than the GPS (+/- 5m), using this plugin to compute a cumulative elevation is necessarily false?

Regards

florent
florent.sourbier
 
Posts: 3
Joined: Mon Jan 23, 2012 4:13 pm
Location: France

Re: Accuracy of the "elevation correction "plugin

Postby gerhard » Mon Jan 23, 2012 7:33 pm

florent.sourbier wrote:To conclude: since the STRM accuracy is lower than the GPS (+/- 5m), using this plugin to compute a cumulative elevation is necessarily false?


The plugin home page lists some of the limitations of SRTM and has some explanations. The SRTM will average the elevation from the sample points, that are like averaged by them selves, so peaks will be lower. I can also see from run data that the peaks are not exactly where I found them (differs from speed data).
However, overall I feel that the elevation profile is far better with SRTM than GPS elevation. Elevation will go up and down also when it is flat with GPS elevation, GPS elevation overshots at peaks and gives too low just after etc. You need to be very sensitive with smoothing to make sense with GPS elevation, it depends from activity to activity.

Governments always manipulate data, especially mapping. But for SRTM I still trust it more than GPS elevation.

In your case with high peaks, maybe the Panorama set gives a better result?
GPS elevation may actually be more interesting when running up peaks...
gerhard
Contributor!
Contributor!
 
Posts: 1798
Joined: Mon Nov 27, 2006 7:07 pm
Location: Sweden

Re: Accuracy of the "elevation correction "plugin

Postby florent.sourbier » Tue Jan 24, 2012 2:30 am

Thanks gerhard for your answer.

gerhard wrote:I can also see from run data that the peaks are not exactly where I found them (differs from speed data).

Not me. In my first test, Cross of Garlaban, if wou click on the picture to enlarge, we can see on the Google Earth view (right), the shadow of the cross. So i was exactly at the peak...

gerhard wrote:However, overall I feel that the elevation profile is far better with SRTM than GPS elevation.

Did you compare your cumulative elevation (obtain with GPS + SportTracks + elevation correction plugin ) with a barometric altimeter (more reliable) under ideal conditions of weather (low temperature gradient...)?

gerhard wrote:Elevation will go up and down also when it is flat with GPS elevation.

You're right. To overcome this problem, you just have to use a threshold (typically 5 or 10 meters in practice) to filter estimation errors of the elevation of GPS that are not equipped with a barometric altimeter. It is in French, but we can simply use this method on line with VisuGpx: http://www.visugpx.com/
I have done the test on several laps around a staduim (flat) with my Garmin F305 two years ago. The cumulative elevation is 0m with a 10m - threshold and about 12m with the elevation correction plugin in ST...

gerhard wrote:In your case with high peaks, maybe the Panorama set gives a better result?

Yes, SRTM 3" Viewfinder Panoramas is more accurate than SRTM 3" on the Estats peak case for example (see picture) but always far from reality.
florent.sourbier
 
Posts: 3
Joined: Mon Jan 23, 2012 4:13 pm
Location: France

Re: Accuracy of the "elevation correction "plugin

Postby texmurphy » Tue Jan 24, 2012 8:58 am

What smoothing settings are you using in SportTracks for Elevation?
For total gain/loss, the definition of Flat in Climb Zones Category is very important. The default Flat yields too much "flatness". I find -1% to +1% is often ok.
texmurphy
Donated!
Donated!
 
Posts: 2125
Joined: Wed Jul 05, 2006 7:38 pm
Location: Maryland, USA

Re: Accuracy of the "elevation correction "plugin

Postby florent.sourbier » Tue Jan 24, 2012 9:54 am

texmurphy wrote:What smoothing settings are you using in SportTracks for Elevation?

Default parameter: 30
florent.sourbier
 
Posts: 3
Joined: Mon Jan 23, 2012 4:13 pm
Location: France

Re: Accuracy of the "elevation correction "plugin

Postby gerhard » Tue Jan 24, 2012 3:54 pm

florent.sourbier wrote:
gerhard wrote:I can also see from run data that the peaks are not exactly where I found them (differs from speed data).

Not me. In my first test, Cross of Garlaban, if wou click on the picture to enlarge, we can see on the Google Earth view (right), the shadow of the cross. So i was exactly at the peak...

Sure, I meant that with elevation correction you can miss the peaks. With the GPS elevation you see nothing...

florent.sourbier wrote:
gerhard wrote:However, overall I feel that the elevation profile is far better with SRTM than GPS elevation.

Did you compare your cumulative elevation (obtain with GPS + SportTracks + elevation correction plugin ) with a barometric altimeter (more reliable) under ideal conditions of weather (low temperature gradient...)?

I have done a manual calculation on a topographic map and compared to a friends barometric elevation. EC is definetely better for ascending/descending, the profile looks "better". But that is at hills, not mountains.

florent.sourbier wrote:
gerhard wrote:Elevation will go up and down also when it is flat with GPS elevation.

You're right. To overcome this problem, you just have to use a threshold (typically 5 or 10 meters in practice) to filter estimation errors of the elevation of GPS that are not equipped with a barometric altimeter. It is in French, but we can simply use this method on line with VisuGpx: http://www.visugpx.com/
I have done the test on several laps around a staduim (flat) with my Garmin F305 two years ago. The cumulative elevation is 0m with a 10m - threshold and about 12m with the elevation correction plugin in ST...

ST allows a lot of things, but the smoothing and Climb Zones are same for all. Custom algorithms are not possible.
Sure it would be nice.
(I have played with the thought of implementing something like this in Trails plugin, but not come so far. I would assume the formula would need to be configurable. Maybe this can be done in Calculated Fields?)
gerhard
Contributor!
Contributor!
 
Posts: 1798
Joined: Mon Nov 27, 2006 7:07 pm
Location: Sweden


Return to Elevation Correction

Who is online

Users browsing this forum: No registered users and 2 guests