Metres climbed ?

[Ljaydee]

This morning I rode an undulating 33miles around Windsor Great Park and Chobham Common.

Back home the Garmin Edge 25 says I climbed 263 metres, uploaded to Garmin Connect and its now 276 metres. Strava running separately on my iPhone says 376 metres. Uploaded to Ride with GPS (which takes the data from Garmin) and I've now got 564 metres all for the same ride.

Would Ride with GPS add its own climbing figure (from the route taken) to the Garmin figure, as its roughly double ?

Which one (if any) is correct ? and how do I get an accurate figure of the metres I have climbed ?

[Cyril Haearn]

564 meters looks like ascent PLUS descent
I would try checking using a large scale map
I would repeat the ride, in the reverse direction and then compare the figures

[Psamathe]

I (no expert) have always thought that a lot depends on the algorithms used. Ride over a speed ramp and you could add an inch or two to your accumulated climb - but at what point do you regard things as part of a surface and at what point consider things part of a climb.

I actually spent a fair time helping a mobile app developer refine their ascent/descent algorithms (as I was the one that raised an issue with their previous releases) and it was fascinating testing different algorithms and comparing different packages. They commented that there is no "right answer" (though accepted their previous releases were definitely not "right") so the algorithm was adjusted to give sensible reading compatible with other packages.

One thing that they did say that was interesting was that they were using the SRTM on a 90m resolution and that when they tested on 30m data it did not make a lot of difference to the figures.

Ian

[hamster]

A lot of the difference comes from the number of data points collected - obviously collecting more is more accurate.
Secondly, some use accelerometer data to interpolate.
Finally, GPS positioning is heavily dependent on satellite locations at the time and can be particularly weak on heights.

[Psamathe]

....
Finally, GPS positioning is heavily dependent on satellite locations at the time and can be particularly weak on heights.

App that I use (and many) has a number of Settings allowing you to consider how altitude is established. I set mine to be based on pre-downloaded map data (actually SRTM data) rather than use the GPS altitude; though I'm unsure what the app's default setting is.

Good illustration is in ViewRanger (at least the versions I used to use - may have changed) where you get separate plots of "Altitude" and "GPS Altitude".

Ian

[rjb]

Plot your route on bikehike and see what that tells you

[andrew_s]

Metres of climbing is a can of worms

In the first place, what constitutes a climb?
Do you add a metre for a hump-backed bridge over a small stream, 10 cm for a sleeping policeman, or 0.5 mm for every bit of chippings on a coarsely surfaced road?
The smaller a "climb" you include, the higher the amount of climbing you'll get for a route.

Also, all methods of measurement of climb have their problems.
I
If you use GPS height, which is measured every second, there are random errors in what you get, with vertical errors typically being about 3 times the horizontal errors, so you could guess maybe 3 to 5 metre errors. The GPS will apply smoothing, but that may not help much - I remember sitting at a cafe table once, attempting to calibrate my barometric altimeter off someone else's GPS, and watching the given height drift up and down repeatedly over a 40 m range.

If you use a barometric altimeter, you've got to calibrate it at a known height on more or less a daily basis, and you'll get drift during the day that could commonly change the climbing by 30-50 m, or occasionally up to 200. On the other hand, moment to moment stability is good, so you don't get much in the way of spurious fluctuations getting added in.

If you use software to derive map heights from a track, you have to be aware that the given heights are interpolated from a grid of point heights. The grid spacing could be 90, 50, or 30 m for free data such as SRTM or OS Opendata, 5 m for paid for OS terrain data, or down to 10 cm if you've managed to get access to LIDAR data. Most websites would use the free data.
Because you are interpolating between grid heights, there will be variations caused by differences between the actual ground and how the (unknown) interpolation method portrays the ground surface.
Furthermore, the grid heights may not actually be real heights at the indicated grid position, but average heights for a grid cell centred at the grid location. SRTM heights will probably be cell averages, OS Opendata is more likely to be real heights.
All told, plenty of scope for errors, that will soon add up if you include smallish height gains.


FWIW, I take a contour count from OS 1:50,000 mapping as giving the "real" amount of climbing. This means that climbs of less than 10 m don't count, which I figure is fair enough.

[foxyrider]

Consistency IME is better from my Garmin than when I was using a barometric computer alone. Generally I get readings within 5m over repeats of the same route and that's close enough for me. The weirdest thing is the 'climb' that I do between stopping at a cafe and setting off again! This seems less of an issue on Garmin connect than on Strava (although it seems to disappear overnight for the most part).

I'm reasonably confident that today's 1621m ride is pretty close to that, in my head I know that there were 4 200m plus climbs and a whole series of smaller ones.

What has been a surprise has been figures riding in Holland, all those dykes do add up, I 'climbed' 250m of them one day! By comparison I did 80km on the lower Trent and gained just 80m, most of which was on the two river crossings!

[Psamathe]

.....
Furthermore, the grid heights may not actually be real heights at the indicated grid position, but average heights for a grid cell centred at the grid location. SRTM heights will probably be cell averages, OS Opendata is more likely to be real heights.
All told, plenty of scope for errors, that will soon add up if you include smallish height gains.
....

For quite some time the SRTM data (outside the US) did not have non-terrestrial surfaces objects removed i.e. the height data was for whatever was highest e.g. tops of trees. So riding along a flat level road into a forrest and suddenly you've jumped up by the tree height!

I'm unsure if the US has released datasets with trees, etc. removed and even so (even in the US coverage) I understood that the "tree removal" was only on the 30m resolution datasets.

But I've not used the data so welcome being corrected if I've misunderstood this.

Ian

[RickH]

A couple of Garmin specific comment.

Garmin Connect defaults to automatically correcting height data for units that don't have barometric altimeters (using whatever height data the mapping they are using has).

Initial height data can take a while, a couple of minutes sometimes, to settle on something resembling reality after you switch on. With a unit with a barometric altimeter, if you start before it is right, this will usually just give an offset, with GPS height measuring it may introduce errors, beyond those introduced by inaccuracies in GPS height, as the the unit catches up (this may add to or subtract from the total).

If you regularly start from one place & your Garmin has the facility to remember locations you can set an accurate height at that point that the unit will use as a reference height when you start an activity near it & means you don't have to wait.

[Bmblbzzz]

They give differing figures due to various algorithms used, as said already. Which is most accurate? I don't know, the more you take, the more accurate your average will be.

564 meters looks like ascent PLUS descent
I would try checking using a large scale map
I would repeat the ride, in the reverse direction and then compare the figures

In this context where the other figures were around half that, clearly so. But 33 miles = 53km, so in general a total ascent of 564m in that distance would be reasonably flat. Depends on the area, obviously -- not to mention what you count as a climb!

[ianrobo]

I would not recommend Strava. I did a test when in the Alps last year and my Garmin and the guy I was riding with came to a total within a couple of meters of each other - 2348 for a days ride. However if you use correct elevation on Strava it added another 450M on it and no way it was .... Speaking to a few others I then found out no one trusts Strava.

[RickH]

I would not recommend Strava. I did a test when in the Alps last year and my Garmin and the guy I was riding with came to a total within a couple of meters of each other - 2348 for a days ride. However if you use correct elevation on Strava it added another 450M on it and no way it was .... Speaking to a few others I then found out no one trusts Strava.

Don't use the correct elevation option then (I'd never seen there was such a thing, I know there is one in Garmin Connect but also that it is off by default if you have a barometric altimeter). The elevation logs in Strava are identical to those in Garmin Connect for me (but then they have come to Strava via Garmin Connect & not recorded separately direct to Strava).

[ChrisF]

Metres of climbing is a can of worms
.....
FWIW, I take a contour count from OS 1:50,000 mapping as giving the "real" amount of climbing. This means that climbs of less than 10 m don't count, which I figure is fair enough.

You can't even use OS 1:50,000 contours safely. A few years ago I had noticed how different people had very different climb values for (the same) audax rides etc. In the end I did a test, and using the MemoryMap software for OS maps very carefully plotted a 'route' along a small local river from near the source to the sea. The stats came out as: total distance 10.2km. Total ascent 31m. Total descent 77m. So water flows uphill.

[andrew_s]

Height by OS contours is going along a map, which may be a paper one, and manually counting those that your route crosses.
Like I said, it ignores all climbs smaller than 10 m, but if I count contours, and someone else counts contours, we get the same amount of climbing (e&oe).

Memory Map doesn't count contours; it uses a DTM ( i.e. a grid of spot heights).