I went through the Motion Based database of GPS data on elevation gain for all of the IM courses in the world. There are a few missing. I took the first five “legit” GPS downloads for each course and averaged them. i was struck by the variation from download to download–frequently 2000 vertical feet. In any event, here are the results of that analysis (total positive vertical elevation change in feet):
I went through the Motion Based database of GPS data on elevation gain for all of the IM courses in the world. There are a few missing. I took the first five “legit” GPS downloads for each course and averaged them. i was struck by the variation from download to download–frequently 2000 vertical feet. In any event, here are the results of that analysis (total positive vertical elevation change in feet):
France 11193
Lanzarote 10282
Lake Placid 7911
Austria 7829
Australia 7659
Louisville 7578
Switzerland 7505
Wisconsin 7353
Canada 6719
Coeur D’Alene 5851
Brazil 5419
Germany 5281
S. Africa 5182
Hawaii 4554
Arizona 3824
W. Australia 2538
Florida 2007
Problem with gps and barometric readings is your sample rate and
speed during the run.
If you samples hit on the downside of a hil or upside but not the top then you will miss some elevation. Now with samples of short intervals 1-10 seconds doubt you will miss a whole hill or climb or at least significant elevation changes you’d count.
I don’t believe there is a standard algorithm or at least rules for calculating elevation gain. BUT if motion based cacluated it from the tracks of each gps upload then most likely they used the same algorithm for each so that helps in the comparison of one course over another.
But if one gps was sampling at 5 seconds ,another 1 second and others 10 seconds even if using the same gps if the sample interval was configurable then that could change… 305 you can select 1 second or smart recording but not enough memory at 1 second.
NOW lets assume all were at same sampling rate…
Well if they all traveled exactly the same speed over the course for every foot then they should (except for gps noise and jitter and all that) be comparable.
BUT heck close enough I guess or as close as your going to get it.
I believe the best method would be to use actual topo data thats all measured the same way and pick a distance interval and elevation window to come up with apples to apples comparison…
Some of the numbers you have are within reasonable range from some readings I have using topofusion and garmin 301,305 and foretrex readings as well as some polar barometrhc readings.
So close enough I’d think. It’s all relative I guess…
Yes–i figured it was related to sampling rate and potentially device as well…Thanks–your post clarifies some of the issues. I’d love to get a more rigorous version/analysis if one exists. given how many IMs are done a year it would seem that we would have this data.
I’m doing IM austria next week and the official elevation change is 5280 which is considerably lower than any of the data on the Motion Based website…
I can tell you that the LP total is off, way off. Four of us with altimeters did a small study last month when up there an d, IIRC, 3067 per lap for a total of 6100 and change.
I hear ya but that is what the GPS data that’s downloaded indicates–check it out–the maps are spot on…that said, would love to have better data if it exists!
I mean, I believe you and I appreciate your doing all that work and sharing the data. It just seems like something must be up with the GPS readings, or … or something.
The other thing to keep in mind is that 2000 vertical feet of climbing at IMFL applied over 56 miles (half climbing and half descending) works out to an average climb grade of 0.6%–which no one can physically gauge without a bike computer…
I can tell you that the LP total is off, way off. Four of us with altimeters did a small study last month when up there an d, IIRC, 3067 per lap for a total of 6100 and change.
Bob
OFF is Subjective… To repeat you rmeasurements depend on your sampling rate… Now lets say you measured altitude at every foot on the road(lot’s of points for sure. depending on how accurate you can measure the altitude if there are enough undulations in the road you would get a larger number than if you measured every 100 feet for example…
NO one can Say this number is right and this one is wrong. NO
matter what method you measure the altidute or how accurate you measure it if you don’t all use the same sampling distance for measurements…
Granted gps is going to be bit jittery and good filtering can rectify some of that BUT YOUR attimeter readings for over all elevation gain is no more accurate than the others since your both using different
sampling rates or even methodologies.
IT’s ALL SUBJECTIVE apples and oranges…
Now if you want to compare one course to another and you use both the same equipment and method for each course and same sample distance then you can say this course has more than that course but you can’t say yours is off or theirs is off in either case.
Take a 1% straight climb for 112 miles think it comes to 5913.6 feet
of climbing. And when you’d finish at top of a 5913 foot mountain.
In this case(never happens) you can sample every 10 feet or every 100 feet or every mile and get the same answer so then you it’s up to the accuracy of your altimeter,gps or whatever as to if you get the same correct answer.
Now take a 1 mile loop. what if in first quarter mile you climb
to 26.4 feet(2 % grade) then go down 26.4 next quarter then up 26.4 .
If you sampled at 1/8 mile interval you would calculate same 5913 feet of climbing if you do this loop 112 times. really small hills but same elevation or climbing.
If you sampled at 1/2 mile interval guess what you would have 0 elevation gain…
Granted most gps units use 4 second and in smart mode I believe they combine colinear tracks and such and at bike speeds your sampling much better than 1/4 mile or 1/8 mile but was just trying to make a point.
Just depends on how undulating the roads are… I guess you could always sample at the highest point before a descent(in a controlled test) then the interval is mute but you then have to decide what is highest level do you measure 1 foot changes ,10 foot,20 foot. Same problem.
sorry all the rambling I’ve been using gps in training and races since 1997 way before the forerunner…(used to use a big gps III with pams bungied to my aero bars(rarely ran with it since battery only lasted 5 to 5 hours…
I mean, I believe you and I appreciate your doing all that work and sharing the data. It just seems like something must be up with the GPS readings, or … or something.
If you sample enough and the road is rough enough and your could detect small enough increments you could get infinite elevation…
whats the ole saying if you measured down to the millimeter or molecular level the coast line of UK you’d approach infinity
So don’t get alarmed with the large numbers instead compare one course
reading with the same method and hardware with another course to get relative elevation of one course to another since elevation gain is a very subjective number thats really not a real measurement…
It’s not like you took a ruler out and measured the distance from point a to point b instead your measuring the amount of ups over a 112 mile course. Heck if you could measure that fine if you measured very time you ran over a half inch rock or 1 inch bump after a while you’d get a very large number which would be useless.
this strikes me as correct and very helpful. the whole concept of elevation gain is simple if you don’t think about it and complex if you do. That said, one course definitely has more gain than another and unfortunately we don’t have a standard test methodology so where are we?
Based on the data I saw I’d say the following:
Huge Climbing Courses: France and Lanzarote
Large climbing courses: Placid, Austria, australia, Louisville, Switzerland, wisconsin, Canada
Modest climbs: Coeur D’Alene, Brazil, Germany, s. Africa
FWIW - Several entries on the motion based site say that the Chicago Marathon has close to 1000 feet of climbing.
Bob
Think the point I was makeing there is no wrong since there is no standard and if your comparing one course to another and use the same method then it’s still useful data just don’t get bogged down by the exact numbers but the relative numbers instead.
I have to laugh. I’m not trying to be an ass here, but there is no way in God’s green earth these measurements are even close to accurate.
Seriously. 8000 feet at Lake Placid? RIGHT. I live at the bottom of a 3000 ft climb. 16 miles to the top. There is absolutely no way that there is as MORE climbing in Lake Placid as there is almost 2.5 trips up Bogus. Not a snowball’s chance in Hades. I’ve done LP 4X and can say that the climb, if you can call it that, from Wilmington to LP is a not that hard. Anyone who has done climbs like Mt. Evans know what 8000 feet of climbing feels like.
Off AT LEAST 2000 vertical feet. My measurement of LP is like 5700.
I went through the Motion Based database of GPS data on elevation gain for all of the IM courses in the world. There are a few missing. I took the first five “legit” GPS downloads for each course and averaged them. i was struck by the variation from download to download–frequently 2000 vertical feet. In any event, here are the results of that analysis (total positive vertical elevation change in feet):
Yeah, GPS calculated elevation gain is pretty nutty.
