Mapping website inconsistencies

dabber

I've recently been playing around with a couple of the mapping websites recommended here.

Looking at their classification of gradients there seems to be big differences for which I can't work out whether the problem is with my interpretation of what they are telling me... or errors in their calculations. The two sites I've been using are Mapmyride and RidewithGPS.

I've posted the same route mapped on each of thes etwo sites.
In particular there is a climb with a max gradient showing 15.7% on RidewithGPS http://ridewithgps.com/routes/298618
but showing as a Cat 4 climb av gradient of 4.1%. http://www.mapmyride.com/routes/view/28872578

Any thoughts... in my own mind, I know this climb certainly tests my limited ability.

amaferanga

Why can't it have an average gradient of 4.1% and a max gradient of 15.7%? Or are you saying that one or both of these doesn't fit with the reality of the climb?

dabber

I guess that could be right. However, I've remapped the subsection of this route with the climb I'm talking about.

Give or take a minor difference in start/finish points the two sites agree on elevation.

RidewithGPS... start 757feet... finish 1218feet distance 1.5 miles
Mapmy ride .. start 754feet... finish 1209feet distance 1.52 miles

However, RidewithGPS gives a max gradient of 13.4% and an average of 2.5%
Mapmyride doesn't seem to give a max gradient but gives an average of 5.7%

http://ridewithgps.com/routes/301237

http://www.mapmyride.com/routes/view/28902346/

Which is right? I don't really have enough experience to know or be able to compare with accepted know gradients (I've never climbed Ventoux I know this climb is hard work for me....
That's really why I'm interested to understand this better and to see if I'm missing something. I'd like to be able to relate a ride/climb that I've done in the past with a route that I might encounter or plan to ride in the future and understand a little better the challenge.

steve_pushbiker

Based on the info in your post.
Height difference is 140.5 metres and length is 2414.016 (or 138.68 m and 2446.2) so a straight line would have a gradient of 5.82%, or 5.7% in the second example.

I don't know how the two sites calculate these numbers but the second is telling you the gradient between the two points (start and finish). The first site probably calculates several intermediate points and gives you an average of those gradients (and reports the max gradient of these).

dabber

Thanks for doing the calculations Steve. It's interesting that one gives a %age gradient twice that of the other. When climb gradients are referred to it would seem that they may be fairly meaningless unless you know how or with what they were calculated.

Given the categorisation function of MapMyRide it appears to me to be the more useful of the two... of others I don't know.

nicksaiz

I have mapped a couple of rides on various websites and got wildly inconsistent total climb data:
in Map my ride-260m
bike route toaster- 360m
bikehike - 380
mytrails.com - 460m
bikely - was at the higher end i seem to remember, though i can't log in half the time.
Both rides had a similar spread with mytrails at the top and map my ride at the bottom.
Does anyone know the most accurate one?

sharky1029

I have tested a couple of steep hills on bike route toaster and found it to be fairly accurate on gradient.
I only use it though for planning routes and then uploading to gps or memory map app on the IPhone.
If any of you have a gps that can show the data on a pc such as a garmin using connect you could compare the planned route and the gps data.

rolf_f

The overall gradients are almost completely meaningless. Anything else is a bit tricky to calculate - a maximum gradient is all well and good but you could get that carrying the bike up a flight of steps! The hardest climb needs a gradient calculated over a reasonable distance but how do you define that distance?

I'm surprised the planning routes differ so much. They should be based on similar digital terrain maps. I'd tend to assume that the higher figures are more accurate (as the coarser the terrain map, the lower the overall recorded climb - though you wouldn't think by much).

briantrumpet

As I've posted elsewhere, I think the gradient and climbing inconsistencies in all these sites are because of the resolution of elevation data - I'm pretty sure that anything using Googlemaps for calculations will undercalculate climbing, because of a low resolution of data (in order to make the online calculations quick).

From my basic researches, I think Google Earth has much better resolution of elevation data (as do programs like Mapyx Quo), but they are much more memory-hungry. But they are also much more accurate as a result.

Headhuunter

I used Bike Hike to plot a route through London a week or so ago and whilst riding my Edge 500 kept telling I had gone off course when I knew that I was on the right road. There must be discrepencies between actual roads and position on websites...

avoidingmyphd

There are at least three issues muddled together in this thread now.
1. Average gradients. You can work these out for yourself and I'd be astounded if any of the websites mentioned are doing the wrong sum. If they are using different underlying elevation data, then different websites might differ slightly, but the difference in averages in Dabber's example is very surprising - can you post the two links?

2. Max gradient. It depends on the distance used in the sum. If one website uses max gain over 10m and the other uses max gain over 100, there is scope for enormous differences. As far as I know, none of the websites declare their methods.

3. Total climb. the variations in nicksaiz's post are driven by the fact that the different websites use different thresholds for what counts as "climbing" - map my ride requires a proper hill (so gives low numbers); some others seem to count kerbs, so give high numbers. None are right or wrong. Just choose the number you like (willy wavers choose the big one). As long as you stick to the same website to compare rides, you'll be OK.

None of the three are entirely due to using different elevation datasets. I'd be surprised if they aren't all just using the google elevation data. The main issue in all three is the maths they do with the data. Averages should be the same across all sites. Max gradients and total climbs in a ride can legitimately vary a lot because they involve judgment calls.

dabber

avoidingmyphd wrote:

but the difference in averages in Dabber's example is very surprising - can you post the two links?



.

All useful comments and interesting to see this thread resurrected.

The links are in my earlier posts on this thread. Let me know if they aren't working for you.

btw, since originally posting this thread I've done the climb I've given as the example many more times and it has got a lot easier.

dabber

Just a further comment having looked at both mapping sites again.

The Max 18.1% shown in the RH summary on RidewithGPS...... if I scroll along the elevation profile at the bottom of the map I can't find a max of 18.1% anywhere.

Also, on both sites they show some downhill sections.... I'm pretty sure these do not exist and are plain wrong. I'm not aware of any relief from climbing on this route.

avoidingmyphd

Dabber wrote:

avoidingmyphd wrote:

but the difference in averages in Dabber's example is very surprising - can you post the two links?
.

All useful comments and interesting to see this thread resurrected.

The links are in my earlier posts on this thread. Let me know if they aren't working for you.

btw, since originally posting this thread I've done the climb I've given as the example many more times and it has got a lot easier.

Oops - I didn't see that. The answer is that ridewithgps is plain wrong: on their own data, the average grade of your climb is 5.8%, yet it says 2.4%
I can't explain why- maybe steve_pushbiker is right that they have invented their own version of "average grade". Weird thing to so though, seeing as there's nothing simpler than the real calculation and the data is right there on the page. As promised, I am astounded and for me that alone would be reason enough not to us the site again.

EDIT - I might have solved it - have you got any other climbs mapped on ridewithgps?

dabber

Try this one, I just put it in.

http://ridewithgps.com/routes/665361

avoidingmyphd

Cheers - that's still wrong, but so was my hunch about why. Just don't use ridewithgps for average grades!

Not being able to find the max grade by scrolling along the graph at the bottom just suggests that graph uses a different distance in its "current grade" sum than the "max grade" is using in the summary. Inconsistent, but no big deal.

If you don't remember the dips, then I'm sure you're right- climbing a mountain on a bike is a pretty good way of judging what's uphill and what's downhill! The fact that both sites pick them up suggests that they're both using the same elevation data I think. It wouldn't affect the average though - if the top and bottom are roughly right, it doesn't matter what the graph says you do in between.

rolf_f

avoidingmyphd wrote:

3. Total climb. the variations in nicksaiz's post are driven by the fact that the different websites use different thresholds for what counts as "climbing" - map my ride requires a proper hill (so gives low numbers); some others seem to count kerbs, so give high numbers. None are right or wrong.

I'll be jiggered if you can show me a digital terrain map that includes kerbs. They don't work that way and, besides, you'd need an insane amount of kerb hopping to make a notable difference. And it isn't true that 'none are right or wrong' - the higher resolution ones (assuming they exist) will be less wrong than the coarser ones but all will be "wrong" (or rather an approximation).

For the most part though, the biggest errors will occur where the direction of gradient changes (ie in valleys or on peaks) where the peaks and troughs get increasingly cut off as resolution reduces. Hence why they will always underestimate the climb rather than potentially overestimating. The higher calculated climbs would imply a finer grid.

avoidingmyphd

Rolf F wrote:

I'll be jiggered if you can show me a digital terrain map that includes kerbs. They don't work that way and, besides, you'd need an insane amount of kerb hopping to make a notable difference. And it isn't true that 'none are right or wrong' - the higher resolution ones (assuming they exist) will be less wrong than the coarser ones but all will be "wrong" (or rather an approximation).

For the most part though, the biggest errors will occur where the direction of gradient changes (ie in valleys or on peaks) where the peaks and troughs get increasingly cut off as resolution reduces. Hence why they will always underestimate the climb rather than potentially overestimating. The higher calculated climbs would imply a finer grid.

Don't worry Rolf, you're in no danger of being jiggered. I was joking about kerbs. You're right, of course, that a finer elevation dataset will yield better total climb numbers. But these websites all use basically (maybe exactly) the same elevation data and the issue is what they're doing with it. And what they do is make a whole load of assumptions that mean that their own numbers are internally consistent but impossible to compare with other sites that make different assumptions. Judging which is "more accurate" is impossible unless they release their assumptions.
Some of those assumptions are made to compensate for possible inaccuracies in their data (all the sites seem to ignore sudden sharp elevation changes to some degree or another, presumably because bikes don't ride up cliffs, so if the data says they do, the data is wrong). Others involve value judgments about what counts as "climbing" on a bike (mapmyride are open that climbing 100m at 0.5% doesn't figure in their total climb data. it clearly figures in other sites' numbers and lots of people think mapmyride are wrong on this).

briantrumpet

I know that the program Mapyx Quo uses the OS Open Data elevation data at a resolution of 50m grid - I don't know what the vertical accuracy is, but I guess its resolution is 1m. I wouldn't be surprised is Google Earth's resolution is approaching this (there are various online references to it), but I've never found anything in detail about Google Maps' resolution - any references seem to be very vague.