How to measure a climb

There's a little checkbox, "__ Show Elevation" that when checked computes the total ascent which on that ride was 1401 feet.

This user's post has been ignored.

The Mapmyride and BT Route Tracker websites both use an elevation database that is by far the most accurate measure. For my rides it's very accurate because we have USGS benchmarks at the top and bottom of all the climbs to compare against.

This user's post has been ignored.

If you do the "Reset Elevation" after you finish mapping a route, the software is able to get all the elevations in a batch, they are correct, and the numbers are always the same. For some reason when the elevations are looked up as you manually enter the points, there are some errors and that's why the "Reset Elevation" button was put on the page.


When you're measuring elevation gains using contour lines (without interpolation or extrapolation) it's not a fractal. I believe the elevation database uses two or three feet contours. Assuming you're using three footers anytime you cross one you've gained/lost 3 feet which is a discrete value and not effected by changing resolution.

Edited by breckview 2009-07-02 7:08 PM

The Reset Elevation button is also there for routes that have been uploaded from Garmins or GPX files. It forces a query of the USGS database for every lon/lat pair in the file.

Depending on the number of datapoints in a route and one's internet connection speed, this can take a long time. If you get disparate values from the same dataset, it's most likely due to a connection issue.

ETA:

The error breckview mentions will not be present in the next version of the route tracker. That was a design flaw I've since learned how to avoid.

For the record, the USGS dataset was created via radar on a Space Shuttle mission. The reason that moving a datapoint gives you a different value is that the elevation value for the new location is retrieved from the USGS database.

This user's post has been ignored.

Edited by PennState 2009-07-02 7:46 PM

Every measurement on the planet earth has "inherent error" in it and is therefore an estimation. Your example, sampling distance slices (x-axis) of the analog garmin elevation plot is not how elevation gain is measured because of the exact problem you demonstrate.

Elevation gain is computed by sampling elevation in the resolution that's available (the y-axis in your graph). Assume that those elevations listed (100 to 350 by 50) were the elevation contours that were available. The only gains would be between miles 2.6 and 3.3, where you'd have climbed 100 feet because you crossed two complete contours, and in the miles between 4.2 and 4.5 where you climbed 50 feet. So the total climbing would be 150 feet. Assuming those 50 foot contours, no other information is available from that graph unless you make assumptions.

Of course the 150 feet is an estimation (like all measurements) because if you were to survey the route so that one foot contours become available, you'd be crossing them all the time and the gain would therefore probably be much higher. However, the amount of climbing didn't change, the accuracy of the estimation changed due to the increased resolution of the elevation contours.

The point is that the elevation contour information that is available dictates the sampling rate. Now that the USGS contours are down to 3 feet or so, the gain estimation is far more accurate than 20 years ago when they were 40 foot.

Edited by breckview 2009-07-02 9:43 PM

Further, in terms of cycling climbs, there is usually some minimum climb below which the gain is ignored because momentum from the previous descent reduces the actual "climbing" that a rider is doing on the bike.

For example, assume a route had 10 foot rollers every tenth mile over a century. The total climbing would actually be 10,000 feet. But most serious cyclists would not consider those rollers to be actual climbing as opposed to a century in the mountains that contains four, 2,500' climbs which would also be 10,000 feet.

If you were to say that a climb must be 50 feet to count, then the roller century drops to zero feet of climbing.

Edited by breckview 2009-07-02 9:59 PM

This user's post has been ignored.

I totally understand the effects of resolution upon measurements. I've done enormous amounts of research on it as it applies to differing trading timeframes in financial markets. But generally in any analysis we are limited to the resolution of "reliable information". For measuring elevation gain that "reliable information" is the USGS contour lines.

It's true that if we measured elevation gain at the distance resolution of the size of a molecule, the numbers would almost certainly differ greatly, than taking samples on the contour lines. But there's no way to know for sure because "reliable information" isn't available at any smaller resolution than the contour lines.

In the past in mountaineering when we were mapping elevation gains (mostly for planning routes) we'd use linear interpolation between the 40 foot contours (and linear extrapolation beyond them) to try to get more accurate but that assumes that between the contours, elevation is rising (or falling) linearly which is just an estimation. With the 3 foot contours (done by the shuttle and mentioned above) that is no longer necessary.

Somewhere above I think I said that BT and Maymyride were accurate "here" meaning where I live (I hope I said that anyway). One of the main reasons that is true is that since, for example, my 45 mile ride only has 3-4 climbs and zero rollers, I can make sure when I'm mapping it that I put a point at all the low and high points of the route because they are super obvious. On a constant rolling course with lots of small climbs, that would be far more difficult maybe impossible if the climbs are so small they don't show up on the 40 foot contour map ("Terrain" view on the BT Route Tracker).

When you import Garmin location data, the points are based on time samples and they occur so often that the hope is that points end up being taken at the top and bottom of every small climb. But there are a lot of problems with that approach the main one being that with so many points, the processing of them is too time-consuming for a web-based application. Another big problem is that since the GPS locations contain errors, when the elevations are looked-up in the USGS elevation database they could be off the actual road (EG. down a river bed and back up instead of on a bridge). Since errors in an elevation data series don't self-correct but instead accumulate, those location errors can add up into lots of false climbing.


Trust me, we're in total agreement on this issue. I think the Garmin elevation numbers are completely worthless (and I've written about it ad nauseam on this site). But pressure-based altimeters are just as bad. I used them for many years in mountaineering for route-finding before GPS was widely available.

I can remember a specific screaming argument at near 14k feet on a very exposed ridge in a blizzard between me and another climber concerning the elevation error on my pressure altimeter and the location error on his new GPS gadget. If we were off by 20 feet and descended the wrong couloir we'd have been forced to reclimb it (at night and in bad physical condition) once we determined we were in the wrong drainage which is how people die in the mountains. BTW, we went my way and I was correct.


I disagree because the GPS elevation errors appear random and each one accumulates false gain.

The method that I've found to accurately keep track of the climbing that I'm doing is that I have all my rides mapped out in climbing segments on BT's Route Tracker (again easy for me because my climbs are long). Then on any given ride I just add up the gain from the climbing segments. I'm anal about it because on my rides the amount of climbing indicates the difficulty of the ride just as much as the number of miles does. Because this method works so well for me, I'm a huge fan of BT's Route Tracker.


Oh you're more than welcome. I good for a ridiculously detailed discussion on certain goofy topics (bike climbing, weather, etc) all day long...

This user's post has been ignored.

Taken by itself, yes there will be mathematical and logical problems with any estimate of elevation gain or loss.

But here's the thing: the goal of including elevation data in the Route Tracker is to provide a metric by which a person can compare workouts against one another. We're not trying to get Tomahawk missiles to autonomously maintain an altitude of 200 feet above ground level here.  We just want a reasonable idea of how hard a particular route is to run or ride compared to other workouts.

For these purposes the sampling method and dataset is perfectly adequate. I find Garmin elevation values to be not particularly useful because they're not consistent. They can change quite a bit from one workout to the next over the same route (this isn't a knock on Garmin, I'm a huge fan of their engineering).  By using the USGS dataset consistently across workouts we have a pretty good idea of how much up and down there is over a given route. And when we use that same dataset and logic to compare it to another route, we get a good idea of how flat or hilly they are relative to one another.

Is it absolutely accurate? No, but the margin of error is absolutely within the acceptable range for the task.

I must say I enjoyed reading this thread as much as the two of you seemed to enjoy debating it.    Thanks!

I posted a simple climb not too long ago which showed a fairly accurate single climb as compared between a Edge 305 and map my ride.  I went on a little longer more complex ride and then compared the two curves.  This is what I got.

This showed a few errors of almost 100 feet in places.  The difference in my edge and map my ride was about 100 feet.  Since there was almost 2000 feet of climbing involved it didn't make much of a difference to me.

From what I understand the world isn't exactly round.  It's like a squashed basketball.  There are even more irregularities than that.  They have a World Geodetic System that is supposed to compensate for this.  It's supposed to change in 2010.

I got a headache just looking at all the math involved.

When I was using my Edge 305 for climbing data I also noticed that the steepest, most sustained the climbing was, the more accurate the device was. But that's not good because on those types of rides, elevation gain is easy to compute and you don't even need the 305, which is why mine never gets used anymore.

Where you really need the 305 is on rolling terrain which is difficult to compute manually, and that's where I've found that the 305 is weak.

Again, as long as I make sure there's a point at the top and bottom of every climb, I've found the BT Route Tracker to be very accurate (and a ground-breaking tool for me). I've compared it's elevation gain figures to manual computation using USGS topo maps, which is as accurate as one can get in my case because we have USGS point elevation benchmarks at the tops and bottoms of all the climbs.

which usgs dataset are you using? 

I have a headache....

Please correct me if I'm wrong, but while the number of measurements can vary from zero to infinity, the elevation change will not vary from zero to infinity.  You might be taking 1000x more measurements, but each of those measurements is also going to be far, far smaller than your previous measurements.  The overall elevation gain is not going to change drastically, once a certain measuring resolution is reached, no matter how many more measurements you do.

Or am I way off base?

Yes. Imagine a 100 stairs, each one foot tall. If you measure in 100' resolution the gain will be 100'. If you measure in one foot resolution again the gain would be 100'.

Now set a 4x4 across each stair. Measuring at 100' resolution is still 100' in gain. But if you measure at 4 inch resolution, the gain becomes 1033 feet. If you measure at a microscopic resolution, you'll go up and down each molecule on the surface of the 4x4 and the gain is higher still.

The reason why a garmin is more accurate for a long sustained climb is that error is only introduced when the elevation mistakenly drops on a constant climb. Assuming a linear 10% slope for 5 miles. The elevation should be higher each time the garmin takes a sample. As long as it is higher at each sample, the only elevation samples that matter in computing total gain are the first and last.

Compare this with a perfectly flat route at sea level. The only elevation sample that doesn't result in false climbing gain is a perfect 0 feet.

Okay, I think I see now.     My mistake was in thinking that the additional climbs would be offset somehow by the corresponding descents, but they aren't, since we're only talking about the climbs.

I didn't read the whole thread. 

Edited by Bioteknik 2009-07-11 9:01 AM