I tracked a few rides with my GPS recently and when I downloaded them to tracklogs afterwards, I noticed that the ascent information was always quite different – for example GPS said 645m, software download said 800m!

which one is correct? I have to say that the ride didn’t “feel” like it had 800m climbing in it… but it sounds good 🙂

Tracklogs is correct

GPS is correct

Both are poor with low precision and questionable accuracy. Tracklogs will “correct” the data using “known” elevation points “along” your route. Speaking professionally, its’ a bad practice and you should be aware of the issues when using the service.

Tracklogs elevation data is based on limited actual measurements. Each measurement effectively sets the height for an area around it. If you think of the side of a hill, Tracklogs sees it as a number of steps. These steps could have , say, 5 Metres height difference. Now, suppose you’re traversing this hill, and the path just happens to slightly wander across the “step”, Tracklogs will think you’ve been going up and down by 5 Metres each time.

similarly, if not based on a barometer measurement, the GPS altitude can be out by 10s of metres 🙁

The accuracy of GPS altitude is pretty poor at the best of times, that’s why fancy ones have a barometer built in. This is unless you have a base station that your GPS can speak to help it resolve the atmospheric delay the signal experiences.

Anyone know what height measurements does Tracklogs use for it’s altitude? Is it based on OS contours/spot heights (which are pretty accurately derived from airphoto’s) or some thing like the freely available 90m SRTM DEM.

And even if you *do* use barometric measurements as well, you have to set relative values every time you go out, as the pressure in the UK varies so much.

Someone told me that the whole of the UK is defined as coastal from an barometric/air pressure fluctuation point of view as we’re so small. No idea if it’s true, though.

Best just to forget all about it and have a nice cup of tea instead.

We’ve done quite a bit of testing on this and there’s quite a few factors.

GPS Barometric – These are good in principle but unless manually calibrated you do rely on sat signals to get the initial values. This can cause an error as the GPS calibrates itself during the start of the ride. Wind also poses a problem and on a windy day atop of helvellyn you’ll likely see the GPS jumping about by 50m or so as it gusts. Obviously this also happens if you ride fast.

GPS Satellite – The GPS was originally intended for 2D positioning so the height resolution isn’t great. Think of it like a big flat triangle, the error is most significant in the small height of the triangle. You do avoid all the barometric issues and overall it’s not bad.

TrackLogs – Uses SRTM (that’s NASA satellite data) to model the height of the land. Conor isn’t quite right in that TrackLogs will either use the elevation data recorded by the GPS or substitute internal elevation model data at each point (right click on the profile to choose). The GPS data is primarily for aerial activities like paragliding where the ground heights are irrelevant. This internal model isn’t stepped, it’s interpolated between the known points so you don’t get false maximums by tracking along the edge of a measurement. Take a look at the 3D view with terrain switch off and you’ll see the wire frame of the elevation model.

SRTM is not interpolated between known points – as is was not gathered as a point dataset (such as lidar would be). SRTM is produced by flying a Interferometric radar system on a space shuttle which orbited the earth for a couple of days in 1999 (I think). Two Synthetic Aperture Radar images were acquired at the same time, from slightly different angles (there was a 90m long boom hanging out of the shuttle with a SAR on it and a SAR on the shuttle). The different viewing angle allows the SAR images to be interfered and differences relate to topography. This dataset was gathered at 30meters – so each pixel (it is in effect an image) contains the average height data for a 30m by 30m square on the ground.

US military would not release the 30m dataset as they were scared so they resampled it to 90m. Soon there will be a 30m dataset freely available for most of the world produced from ASTER satellite images.

Lucas – I didn’t say the SRTM was interpolated, TrackLogs performs the interpolation between the grid of SRTM measured points to provide more accurate estimates between the samples and to eliminate the step effect mentioned by druidh.

uh, interesting, no easy answer to that one then! Thanks everyone for having a go though!!

I use a Geko 301, I don’t think that’s barometric (or is it? 😳 must check instruction manual…)

I did wonder whether there was something like the step effect druidh mentioned going on, but it seems that this isn’t likely?

Think I’ll go and have that cup of tea now (barney :wink:) and read your scientific explanations again!

alp_girl – yes the Geko 301 (the grey one) has a barometric altimeter. Next ride try switching it on and getting a lock whilst you get the bike ready. Then reset the trip computer when you’re ready to ride. That will give it the best chance to auto-calibrate itself with the day’s atmospheric pressure and remove any dodgy points hanging on from the previous time is was used. When you get back check the stats and if you’ve got a big difference upload the track and look at the elevation profile. A different start and finish elevation on a circular ride is a dead giveaway of a slow calibration, any others differences should be obvious from the profiles.

The step effect is unlikely, that was an artefact in very early digital mapping. TrackLogs addressed that in v2.something, can’t speak for the others but I’d imagine most platforms interpolate by now.

Here’s a good example of mis-calibration – I probably started moving before the GPS was ready so it didn’t get a proper initial fix on altitude. The ride starts and finishes at the same point, though it goes from being 324 feet below sea level to 204 feet above by the time I return home.

At some point during the ride it manages to calibrate itself, but I’m not sure I could pinpoint exactly where…

This is from a Garmin Edge 305 which seems to use satellites to auto-calibrate and the barometric altimeter thereafter to track changes in height.

Interpolation is fraught with issues. It’s just bad practice mixing data of different precisions to generate a result.

True, consumer satellite navigation units (they are not called “GPS”… that’s the name of a constellation of satellites) have poor Z precision, 10mt at best. But a DEM may be better or worse… it all depends on how far you are from the nearest N measurement points.

But as it’s just used for a bunch of middle class types to feel smug, it hardly matters. I’ll go back to worrying about my survey data for the 150 manholes we did last week.

So surveyors aren’t ‘middle-class types’?

I’m not anyway! I have no class at all 😉

they are not called “GPS”

I think you’ll find they are!!

I’m not sure interpolation is necessarily a bad thing. As you (rather bluntly) elude, recreational digital mapping software and land survey have very different elevation data requirements. When looking down a manhole, and I’m no expect here, I’d imagine any form of extrapolation is bad and it would be advisable to quote what you know as fact complete with error stats. When estimating elevation gains along a footpath or bridleway we have to be a little more pragmatic. Extrapolation is good here as it reduces error trends against empirical measurement. We don’t just make this stuff up you know!

GPS receiver or navigation unit would be acceptable.

But in theory the units can receive signals from other, non-GPS constellations, such as Galileo and Glonass. Our new survey grade receivers pick up Glonass birds as well as real-time corrections over the VRS network. Such and advantage in urban areas!!

Digimap, that’s what I mean, an interpolated DEM made up of 10mt pixels based on points that are 90mts apart is fine if you want to know roughly how much elevation you have climbed or lost etc. But as you say you have to explain what the data is and where it comes from, most users don’t know this though, and don’t care.

My MSc dissertation was modelling past environments based on a DEM I generated (using a Kriging or nearest neighbour algorithm I think.. was 5 years ago!!), so am aware of the power of interpolated datasets… along with the drawbacks and issues.

Anyroadup – your Geko 301 will be more accurate than Tracklogs.

Anyroadup – your Geko 301 will be more accurate than Tracklogs.

A barometric altimeter subject to changing pressure, wind versus software that uses known positions plotted agains known elevations?

I accept what you say about dodging across contour lines, and in any event elevation profiles are only as good as the software used to calculate them (and some can be pretty crap), but I’d much sooner trust an elevation profile plotted from a map.

digimap/stuartie_c – just checked the ride I’ve still got saved on the Geko, and the starting and finishing elevation details are indeed different although I started/finished at the same point… never thought to look at that before 😯

Enjoyed reading the further discussion though, and I guess I’ll mostly rely on my legs to tell me which estimate is the closer of the two (Geko or Tracklogs) 😛

andym – Member

I accept what you say about dodging across contour lines, and in any event elevation profiles are only as good as the software used to calculate them (and some can be pretty crap), but I’d much sooner trust an elevation profile plotted from a map.

Oh – I reckon that manually calculating it off a map would be the most accurate. However, that wasn’t one of the options in the OP.

To see what interpolation can do, have a look at the East beach on Gibraltar in Google Earth 3D. Now imagine yourself walking along that beach. How accurate do you think that interpolation is?

druidh, what do you mean by “manually calculating off a map”?

alp-girl – guess he’s referring to the technique where you used to count contour crossings on a paper map. It’s do-able for a shortish route, a walk perhaps, for a bike ride it’s a bit labour intensive. There’s no advantage over software methods since the accuracy specs of the OS contours are no better than the modern elevation models.

For your Geko 301, check your calibration settings to make sure you’ve set it to auto-calibrate itself. This gives it the best chance to get accurate readings as atmospheric pressures changes. FWIW we still use barometric here as in auto-calibrate mode the results are pretty good and they respond much better to the rapid changes of elevation you get mountain biking down big hills.

I accept what you say about dodging across contour lines

There is no issue with dodging across contour lines in TrackLogs as the interpolation used is a continous model which contains no steps. Rather than dumb interpolation of a finer grid of stepped values the software interpolates from any nearby known datapoints to produce an estimate at precise points along your path, no discontinuities at all!