GPS elevation query

by MCLK   March 13, 2012  

I tend to spend quite a lot of time using sites like Bikely and MapMyRide to plan out longer century rides and then often compare them to the results on my Garmin afterwards. I've noticed that the Elevation figures seem to vary quite drastically from site to site, to the point where I don't know which one to trust - Anybody else had similar problems?

24 user comments

Oldest firstNewest firstBest rated

Yes.

I get a lot of variation between my Garmin, bikemap.net and bikeroutetoaster. Like you I don't know what to trust.

I suspect it's to do with the sampling frequency used to calculate heights. For example, if the maps are only sampling elevation every 100m then they may miss small undulations that the Garmin picks up, if it's calculating more frequently.

However, that's speculation on my behalf. I'd like to know if anyone has a better insight into what's happening.

posted by thereandbackagain [150 posts]
13th March 2012 - 13:01

like this
Like (2)

I was thinking it might be a sampling precision issue, too, as it is more noticable on longer routes.

Most of the sites I've tried use google for mapping so I wonder if the GPS information calculated independently as exporting a common route between such sites still produces different elevation results?

posted by MCLK [82 posts]
13th March 2012 - 13:52

like this
Like (0)

Three different methods of calculating elevation:

1. GPS derived.
2. Barometric (I believe this is what Garmins use).
3. Map data.

All have their plus and minus points, and which is more accurate depends on the exact situation and the quality of the data.

posted by Mr Will [84 posts]
13th March 2012 - 15:00

like this
Like (0)

I've just had a look at the ride I did on Sunday, 109 km in total. The elevation figures are as follows

Garmin Barometric 2263m
Garmin non Bar 2366m
Bikemap.net 2050m
MapMyRide 1677m

So quite a range for the same route. Normally I would probably just go with what the Garmin says but when entering an organised ride, a AAA audax for example, it's a little hard to get feel for what I'm up against beforehand, when I don't know which figure to go by - the extra 600m variation (above) would change the nature of a ride quite considerably.

posted by MCLK [82 posts]
13th March 2012 - 16:07

like this
Like (1)

I use a Garmin and elevation figures vary even on the same routes but ridden on different days. How ever you measure it you 'only' climb what you climb and not really knowing just adds to the adventure.

posted by malmesburyclarioncc [26 posts]
13th March 2012 - 16:33

like this
Like (0)

malmesburyclarioncc wrote:
I use a Garmin and elevation figures vary even on the same routes but ridden on different days. How ever you measure it you 'only' climb what you climb and not really knowing just adds to the adventure.

I guess you're right. I'll go with the highest figure and be pleasantly suprised if it works out to be less Smile

posted by MCLK [82 posts]
13th March 2012 - 16:39

like this
Like (1)

Id go with the biggest number.

posted by miffed [160 posts]
13th March 2012 - 16:39

like this
Like (1)

You're quite clearly overthinking this; it's the biggest number, obviously. Unless someone can produce NASA data to the contrary.

posted by nick_rearden [858 posts]
13th March 2012 - 23:26

like this
Like (2)

nick_rearden wrote:
You're quite clearly overthinking this; it's the biggest number, obviously. Unless someone can produce NASA data to the contrary.

I'm not sure I am. I just wanted a reliable way to study a route before jumping on the bike.

posted by MCLK [82 posts]
14th March 2012 - 0:42

like this
Like (1)

As an experiment, I've just looked at a ride of 1km from my house on Map My Rides - it gives the final 200m or so as 3 per cent gradient, which is way out - I have no idea how much it actually is, but if it hit over 20 per cent I wouldn't be surprised.

[Edit: Checked on OS map - it's between 14 and 20 per cent. Contours vs distance suggests closer to 20]

Thing is, it's preceded by a shorter but equally sharp downhill section. So as first poster above said, there's clearly some sort of error there - in this case I reckon it probably takes somewhere around the start of that dip and the top of the climb and misses out the whole bit in between.

Over a long ride, things like that are going to seriously add up.

ps I don't ride that road often Devil

pps My rides are always more than 1km - just thought I'd use that distance on a road I know well to see how MMR compares with the actual terrain Wink

Simon_MacMichael's picture

posted by Simon_MacMichael [7471 posts]
14th March 2012 - 1:27

like this
Like (2)

Been playing around with the Strava Route planner and comparing it to MapMyRide tonight for a ride I'm planning next year

One section of it (Pontypridd to Llandovery - 50 miles) came out as 4200ft on Strava and 2550ft on MMR. Identical routes.

How can they be this far apart? Not like the route is over roads that are hardly used either

posted by VeloPeo [182 posts]
9th January 2014 - 0:03

like this
Like (0)

Note that there is no correct elevation gain number. It will depend inherently on the sampling interval, and you can make those ridiculously large, as well as ridiculously small.

It's a bit like the "How long is the coastline of Britain" question...

posted by Paul J [423 posts]
9th January 2014 - 9:46

like this
Like (0)

Paul J wrote:
Note that there is no correct elevation gain number. It will depend inherently on the sampling interval, and you can make those ridiculously large, as well as ridiculously small.

It's a bit like the "How long is the coastline of Britain" question...

I realise that - I only commented as the Strava elevation figure is massively different to MMR. I'm not expecting them to match exactly but I'd expect a tolerance of +/- 5% not 65%!

posted by VeloPeo [182 posts]
9th January 2014 - 9:56

like this
Like (0)

Oops, I probably mean measurement interval, not sampling interval.

posted by Paul J [423 posts]
9th January 2014 - 9:56

like this
Like (0)

the map data is likely to be the most accurate overall. I don't think the accuracy of consumer GPS units is such that the elevation can be trusted, they are mainly aimed at pinpointing your position on a 2d map. I once took my garmin out kayaking and over a 3 hour session it mapped my elevation at +-60m above / below sea level which was not entirely accurate, which while some of the time was spent submerged, it wasn't that deep.

joemmo's picture

posted by joemmo [691 posts]
9th January 2014 - 10:54

like this
Like (0)

I have Google Tracks on an Android Smartphone, which uses GPS and the elevation for the same ride can vary day by day..I assume MapMyRide uses the same system

I got a bottom of the range Garmin as a Xmas pressie and have noticed the elevation figures on the maps you create in Garmin Direct are more consistent ie they all give the same height for points on my ride. But the calculated figures for total Ascent / Descent vary each day

posted by RedfishUK [34 posts]
9th January 2014 - 11:55

like this
Like (0)

IIRC, and I may well not, most mapping sites use the free elevation data produced by NASA. The data they release for free for the "rest of the world" has a measurement interval of just under 100m (its a round number of arc seconds or some such unit which comes out about 100m on the earth's surface). This interval will filter out many undulations that you will have to go over and so make routes look smoother. In the US they release much higher precision data I believe.

Having had a quick google I think this is the stuff: SRTM on Wikipedia

Please note I may have just imagined all that.

posted by cqexbesd [11 posts]
9th January 2014 - 14:42

like this
Like (0)

This is my take on elevation data:

Plan a route on said route planner
Take the route planners elevation date as a guide
get out on bike and enjoy ride
if there is more climbing than expected suck it in and MTFU!!

posted by ScottChegg [9 posts]
9th January 2014 - 14:46

like this
Like (3)

Heh heh - quite agree ScottChegg.

Got no problems with whatever climbing it turns out to be (the whole route I'm looking at doing according to Strava is 198 miles 13,500ft) - just really surprised that two different sets of mapping data can produce such a wide variance of elevation over the same route, they're not even close.

posted by VeloPeo [182 posts]
9th January 2014 - 15:14

like this
Like (0)

Like many of the above, I've compared readings. If you look at the raw reading from a GPS device as a plot it usually looks jagged - like the kayak example above I've used a forerunner on a flat run which shows lots of ups and downs. When I used sporttracks to record the data, I used an elevation correction function in the software, which smooths the data, giving a much lower number. The websites apply a similar function to the recorded data. When I've run the comparisons, I've found the order to be:
Raw
Garmin with barometric
Ride with gps
Sporttracks smoothed
Bikeroutetoaster (sadly defunct)
Map my run/ride

With the sort of variance described above, mapmyrun being consistently 20-30% lower than bikeroutetoaster. I've never used strava, so I don't know where it sits. I guess that this means that mapmyrun has the most aggressive smoothing algorithm.

So for comparing routes, just use the same one, for demonstrating awesomeness use one of the top ones!

posted by Stratman [36 posts]
9th January 2014 - 21:34

like this
Like (0)

I used my new Garmin Touring for a ride the other day and it showed 1082 m elevation. When I loaded it onto Garmin Connect it changed to 819 m. I then noticed that there waa a switch at the bottom of the page with 'elevation enabled'. When I disabled this it went back to 1082. There is an explanation near the button that explains that the Touring that doesnt have an altimeter takes elevation from GPS, but Garmin Connect will compare the map with actual surveyed heights and correct the elevation aa this is supposed to be more accurate.

posted by Sensible [46 posts]
10th January 2014 - 0:29

like this
Like (0)

I believe Strava now has the best, most accurate and comprehensive elevation data on the internet, as they have been effectively crowdsourcing the data from everyone with a barometric garmin who logs their rides with them. This means they end up with a very good average from all the people who ride the roads in different days/pressures etc.

posted by benezeir [50 posts]
10th January 2014 - 10:06

like this
Like (1)

Quite a few years back I worked for an american company making military radios these all singing all dancing devices had gps fitted, and they had to be tested before being shipped.

It's amazing how we managed to stay on our feet whilst working in a building that rose and fell by over 60 meters every single test Laughing

FATBEGGARONABIKE's picture

posted by FATBEGGARONABIKE [481 posts]
11th January 2014 - 19:05

like this
Like (0)

VeloPeo: No, you can't expect that different measurements will be within 5% of each other.

What you *can* expect is that there is a minimum elevation gain - the difference in elevation between start and end by the system that produces the lowest figure. After that, figures will increase and increase as you measure ever more granularly.

That's ignoring the issue of error in GPS, and whether GPS elevation error has any bias (I think GPS lat/long error is relatively unbiased, so over time the errors can be made to cancel out - maybe elevation is the same, but I don't know) . Wink

posted by Paul J [423 posts]
14th January 2014 - 14:14

like this
Like (0)