Measuring Ascent - Whom do I believe? Garmin, Memory Map etc

Lard Armstrong

Ascent and Descent varies significantlly between my Garmin Edge, Memory Map, and the various other free ride mapping tools on the web. Recently, I compared one ride captured on the GPS which showed 1800M ascent in Garmin Training Centre, but 1300M in Memory Map (track converted to route) and a different figure again in Bikeley.

Whom should I believe ?

Cheers.

vorsprung

GPS based systems or systems where the route is traced onto a map suffer from a common problem. If the exact position does not match the exact position on the virtual map that the system is using then the virtual maps altitiude is wrong.

To see why this is a problem, imagine a canal towpath. it is flat but the terrain it crosses varies in height. The canal towpath has embankments and cuttings around it. If a GPS user rides along the path reporting their position to within a few metres accuracy SOMETIMES they will be up a cutting ( higher up ) and SOMETIMES they will down an embankment (lower down) these variations in height will record as ascent or descent on what is really a perfectly flat path. Using a online map tool like bikely and drawing on the route suffers from the same problem as the "line" may have a width on the map which includes the nearby steep terrain.

To overcome these problems all of these types of system have various counter measures built in, averaging heights and ignoring small amounts of up and down. Because of the differences in these "fiddle factors" on the same real world routes, different GPS/mapping systsms give different amounts of climbing.

According to an article in Arrivee (the UK long distance cyclist mag) in the UK, bikely and other online mapping systems tend to over report climbing compared to GPS units, which are slightly better but still not perfect.

graeme_s-2

I did a triathlon on Sunday. I did two laps on the bike leg, and my Garmin Edge 705 clocked the main hill as ~130m on the first lap, then on the second lap it was apparently ~95m. It felt pretty similar to me both times and I don't think there had been any significant seismic activity in the intervening 30 minutes!

I thought it was supposed to have some sort of altitude meter in it rather than rely on plotting it onto a GPS map. If it does then it's apparently not all that accurate!

rhext

GPS systems can measure elevation in the same way that they measure position, but the precision of the measurements is much lower. Better quality Garmin's incorporate an altimiter, which works on atmospheric pressure variances, and use the output to refine the GPS measurements, but I don't think that either approach is particularly accurate. You can apply a geographic correction when you import the routes (where your route is compared to elevation measurements from a terrain database), but this is also subject to errors.

The bottom line is that the process of tracing elevation across terrain databases is subject to the routing algorithm, while GPS and pressure-based measurements are badly affected by measurement inaccuracies. What you need is a satellite with a height-finding radar following you along your route!

ColinJ

For accurate positional readings, the GPS needs a good signal from at least 3 satellites. Altitude requires a good signal from at least 4 satellites.

I have the cheapest GPS in the Garmin Etrex range and I've checked its altitude readings against my OS maps. It does a pretty good job most of the time (less than 1% error). I have found that being tight up against steep hillsides affects it, for example in the bottom of deep, steep-sided valleys. Heavy tree cover can also disrupt the satellite signals. Up on open moorland it seems very good.

If I compare the altitude profiles from my tracklogs and my route profiles in Memory Map, there usually isn't too much difference. Vorsprung is right, however, in that certain features will catch out mapping software.

To give you an example - here is a short track I plotted with my mapping software. It goes along the M62 at Scammonden Water. The motorway crosses the valley along the top of the dam but the mapping software altitude profile is showing the shape of the valley!



I've found on long rides over varied terrain, Memory Map gives a total ascent value which is about 20% high.

GeorgeShaw

It's not necessarily to do with satellites, as the more advanced Garmins have barometric altitude calculation. As vorsprung says, there are differences in sampling when reading heights (either sampling rates or misplaced samples on maps) and then differences in rounding algorithms in the reporting software.

For example, I can put the GPX output from my Garmin into 4 different prices of software and get 4 different total ascent values - from the same input data.

So you shouldn't "believe" any of them. The point is more, why do you want to know the total ascent? If it's for bragging rights, then pick the biggest number If it's for following your training, then just use a consistent method. If it's for route planning, then you can get an idea from comparing a few routes using both one of the mapping sites and the output from your Garmin and doing the maths.

ColinJ

GeorgeShaw wrote:

The point is more, why do you want to know the total ascent? If it's for bragging rights, then pick the biggest number If it's for following your training, then just use a consistent method. If it's for route planning, then you can get an idea from comparing a few routes using both one of the mapping sites and the output from your Garmin and doing the maths.

I agree with that.

I work out the figures for the forum rides that I organise so I can give people an idea of what they are letting themselves in for. My rides are always hilly and I don't want someone to come along unprepared and then collapse halfway round.

I take a look at what Memory Map calculates for the route then I look at the route profile and manually count up all the significant climbing. I usually quote a figure about halfway between the two.

avoidingmyphd

The problem(s) people have with measuring ascent / descent are not usually linked to the accuracy of the elevation measurement. Obviously, you need relatively accurate elevation data to calculate ascent/descent but you also need to make a judgement about what to count as ascent/descent once you have that data. But many of the sources "we" use are ultimately getting their elevation data from the same place.

If you cross the road, do you count 10cm descent, then 10cm ascent to account for the kerbs?
If you ride up a long hill with a steady gradient of 0.1%, does this count as ascending?
If you fall in a 5m deep hole and climb out again, should this be excluded just to be sensible?

To calculate ascent and descent from raw elevation data you have to decide how to deal with these kind of situations. Compare the numbers from the same route in mapmyride.com and bikeroutetoaster.com and you'll see the impact this has. The former counts only big hills; the latter would count kerbs if the data had them in. They use the same elevation data and end up with massively different ascent/descent.

So believe the provider whose conception of what elevation changes count towards ascent and descent tallies most closely with your own.

derekwatts

Even units with barometric measurement or compensation can be out if the ride is long enough... For every 1mb pressure change that's approximately 30ft at sea level. A long enough ride can see a significant change in barometric pressure depending on the weather. For example a 7 hour ride I undertook recently measured 8500ft of total climbing and 7900ft of total descending, but I started and finished at the same point! But the pressure had dropped significantly during the ride.

As for the Garmins, I don't own one, but do they produce a 3d nav fix or a 2d fix applied to a topographiocal map on the memory card? I would be interested to know as I want to buy one. As Colin says, a 2d (lat/long) fix needs 3 visible sats, a 3d (with GPS altitiude) needs 4 sats, and for what it's worth, a system that can identify an erronous satellite (Basic RAIM) needs 5 visible sats, and 6 if you want to identify and exclude the faulty one from the navigation solution (Full RAIM). But I doubt any bike computers have (or need) that!