google versus rwgps elevation
#1
Full Member
Thread Starter
Join Date: Feb 2015
Location: Pittsburgh, PA
Posts: 400
Bikes: Schwinn, Nishiki, Santana, Trek, Rodriguez
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 189 Post(s)
Likes: 0
Liked 106 Times
in
69 Posts
google versus rwgps elevation
The total amount of climbing over a route is often very different between google and RWGPS. Which is more accurate in your experience?
#2
Bike touring webrarian
I don't have an answer for you, but I can explain the problem. The routes drawn on maps consist of small straight lines. The elevation differences are calculated on the ends of these lines. On curvy (think switchbacks) and/or mountainous areas, these lines will hit different parts of the road/terrain. Depending on how a particular site creates its lines, its elevation numbers will be different. So will their distance totals.
That said, the totals for any particular route (assuming it isn't too long) shouldn't all that different. How much difference have you seen?
That said, the totals for any particular route (assuming it isn't too long) shouldn't all that different. How much difference have you seen?
#3
Full Member
I can tell you that sometimes google is 1000% wrong! Sometimes a route we takes seems relatively normal then at the end the graph shows a 90 degree angle! Perhaps the total elevation is accurate but the graph is not to be trusted.
#4
Full Member
Thread Starter
Join Date: Feb 2015
Location: Pittsburgh, PA
Posts: 400
Bikes: Schwinn, Nishiki, Santana, Trek, Rodriguez
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 189 Post(s)
Likes: 0
Liked 106 Times
in
69 Posts
I am planning a 3 week loop through Iowa. I export various gpx tracks from rwgps to google, so they are the exact same route. For various single days taken from the entire tour, I often see that the total climbing in rwgps are more than 500 feet greater than in google maps. I find this strange since the map data that rwgps uses is from goggle. Apparently goggle's algorithrim for elevation changes does more averaging over the track. Here is one such example, total climbing: rwgps=2200', goggle=1400'.
https://ridewithgps.com/routes/35599491
https://ridewithgps.com/routes/35599491
#5
Senior Member
Join Date: Jul 2007
Location: South shore, L.I., NY
Posts: 6,885
Bikes: Flyxii FR322, Cannondale Topstone, Miyata City Liner, Specialized Chisel, Specialized Epic Evo
Mentioned: 18 Post(s)
Tagged: 0 Thread(s)
Quoted: 3242 Post(s)
Liked 2,089 Times
in
1,182 Posts
Maybe e-mail RWGPS Tech Support and ask ?
#6
Bike touring webrarian
I am planning a 3 week loop through Iowa. I export various gpx tracks from rwgps to google, so they are the exact same route. For various single days taken from the entire tour, I often see that the total climbing in rwgps are more than 500 feet greater than in google maps. I find this strange since the map data that rwgps uses is from goggle. Apparently goggle's algorithrim for elevation changes does more averaging over the track. Here is one such example, total climbing: rwgps=2200', goggle=1400'.
https://ridewithgps.com/routes/35599491
https://ridewithgps.com/routes/35599491
You might check the elevation profiles closely to see if there are some places where it doesn't quite match the lay of the road.
#7
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,278
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4260 Post(s)
Liked 1,364 Times
in
945 Posts
So, the difference may be in how it's being calculated from the base elevation.
It's fairly common that the elevation gain measured by a Garmin device is larger than the RWGPS estimate.
#8
bicycle tourist
I haven't used RWGPS too much for elevation, but every once in a while, I've found Google Maps to have an interesting discontinuity.
For example, can't reproduce it when I try now, but last summer I plotted a route on the Great Plains between Lubbock and Amarillo. The initial routes has two discontinuities where the road appeared to go from ~3700ft elevation to sea level and then back to ~3700ft in a very short distance. Clearly there isn't such a deep chasm in that area. That by itself was ~7500ft of climb that added considerably to the total for a mostly flat route.
For example, can't reproduce it when I try now, but last summer I plotted a route on the Great Plains between Lubbock and Amarillo. The initial routes has two discontinuities where the road appeared to go from ~3700ft elevation to sea level and then back to ~3700ft in a very short distance. Clearly there isn't such a deep chasm in that area. That by itself was ~7500ft of climb that added considerably to the total for a mostly flat route.
#9
Sunshine
Join Date: Aug 2014
Location: Des Moines, IA
Posts: 16,617
Bikes: '18 class built steel roadbike, '19 Fairlight Secan, '88 Schwinn Premis , Black Mountain Cycles Monstercross V4, '89 Novara Trionfo
Mentioned: 123 Post(s)
Tagged: 0 Thread(s)
Quoted: 10969 Post(s)
Liked 7,495 Times
in
4,191 Posts
RWGPS elevation gain is consistently short compared to my GPS computer.
Strava elevation gain is consistently short compared to my GPS computer.
I swear that most of the climbing in Iowa(except for the northeast driftless area) doesnt get calculated accurately because its short punchy hills over and over between creeks.
The GPS unit has a barometer.
So either the barometer is wrong, or the RWGPS and Strava guesstimates are wrong.
Strava elevation gain is consistently short compared to my GPS computer.
I swear that most of the climbing in Iowa(except for the northeast driftless area) doesnt get calculated accurately because its short punchy hills over and over between creeks.
The GPS unit has a barometer.
So either the barometer is wrong, or the RWGPS and Strava guesstimates are wrong.
#10
Full Member
Thread Starter
Join Date: Feb 2015
Location: Pittsburgh, PA
Posts: 400
Bikes: Schwinn, Nishiki, Santana, Trek, Rodriguez
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 189 Post(s)
Likes: 0
Liked 106 Times
in
69 Posts
The resolution of barometers in handheld devices is approximately 70 feet. The cumulative up/down can also have significant error depending on if a given 0-70 foot elevation change registers a single bit change in the digital converter of the device. If you go up 10 feet and it happens to change one bit, then that would be added to the cumulative as 70 feet. Perhaps the averaging calculation is more sophisticated, but still cumulative error.
Yes, agreed that Iowa has lots of rollers. Rollers are a kind of averaging (best kind) in that the energy climbing/descending is not mostly lost (no braking). Perhaps the averaging that google does, giving lower overall climbing values, is closer to actual biking effort.
Elevation noise aside, looking forward to touring Iowa in June.
#11
Senior Member
Join Date: Dec 2010
Location: northern Deep South
Posts: 8,904
Bikes: Fuji Touring, Novara Randonee
Mentioned: 36 Post(s)
Tagged: 0 Thread(s)
Quoted: 2604 Post(s)
Liked 1,933 Times
in
1,213 Posts
The resolution of barometers in handheld devices is approximately 70 feet. The cumulative up/down can also have significant error depending on if a given 0-70 foot elevation change registers a single bit change in the digital converter of the device. If you go up 10 feet and it happens to change one bit, then that would be added to the cumulative as 70 feet. Perhaps the averaging calculation is more sophisticated, but still cumulative error.
You can measure barometric pressure with much better resolution than 70 feet. There's still some diversity between device manufacturers as to how much you'll filter out as noise. I've ridden rollers with a Casio watch (which doesn't accumulate anything less than 60 feet climb) and a Garmin with barometric sensor (which starts accumulating climbing distance around six feet). You can climb and descend an awful lot of 45' rollers, ending up with 350' of climbing on the Casio but over 800' on the Garmin.
The same thing happens with mapping software. Do you only count when a route crosses a contour line, and then count that as the distance between contour lines? Or do you try to interpolate between contours? Get into the mountains where USGS mapped terrain with 100' contours, and the first wears you out for almost no counted climbs; while the latter, especially on converted rail line in a narrow gorge, doubles the actual climb.
Google does have some elevation data, presumably from driving its vehicles over some roads and correlating barometric pressure with surveyed datum points. If that's where you're riding, that's likely the second best source (behind highway department survey data) for climbing. On most smaller roads where they've never driven their fancy vans, a GPS with barometric altimeter from a reputable company would be my preferred source of information.
#12
just another gosling
Join Date: Feb 2007
Location: Everett, WA
Posts: 19,538
Bikes: CoMo Speedster 2003, Trek 5200, CAAD 9, Fred 2004
Mentioned: 115 Post(s)
Tagged: 0 Thread(s)
Quoted: 3890 Post(s)
Liked 1,939 Times
in
1,384 Posts
I find RWGPS elevation gain on a route is very close to what one of my Garmin 800 devices shows. My other 800 shows a much higher gain. After a group ride we sometimes have an elevation question time. Every one has a different number, most of them showing lots more gain than RWGPS. IMO it's like calories on a sports watch: the manufacturers set up the algorithm to make the user happy. More calories and more gain = happier rider. Simple.
I don't trust Google maps with anything. RWGPS is so much more reliable w/r to routing.
I don't trust Google maps with anything. RWGPS is so much more reliable w/r to routing.
__________________
Results matter
Results matter
#13
Full Member
Thread Starter
Join Date: Feb 2015
Location: Pittsburgh, PA
Posts: 400
Bikes: Schwinn, Nishiki, Santana, Trek, Rodriguez
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 189 Post(s)
Likes: 0
Liked 106 Times
in
69 Posts
I realize there is a difference. For example, here is an engineering analysis which discusses the barometric precision of hand held devices.
https://www.amsys-sensor.com/downloa...amsys-509e.pdf
Perhaps Garmin-like devices do a little better, don't know. My interest is in using rwgps, because it is a useful route planner and provides an elevation profile. But the elevation discrepancy between it and goggle is significant even though rwgps gets their map data from google.
It seems people find rwgps aligns best with their actual experiences.
https://www.amsys-sensor.com/downloa...amsys-509e.pdf
Perhaps Garmin-like devices do a little better, don't know. My interest is in using rwgps, because it is a useful route planner and provides an elevation profile. But the elevation discrepancy between it and goggle is significant even though rwgps gets their map data from google.
It seems people find rwgps aligns best with their actual experiences.
#14
Senior Member
what device?
I noticed older phones had terrible GPS accuracy in any sort of difficult terrain (mountains, forest etc.). Sometimes it would place me a few ft next to a path i was on, sometimes more. Sometimes it would miss a piece of switchbacks and just add a straight line over some hundred yards where it should have been zig-zag and curves. Newer phones (I have a 3 or 4 year old flag ship phone) are MUCH better in the very same locations. Probably not 100% perfect, but i don't notice the obvious problems anymore.
I don't pay attention to elevation, but what I observed also must apply to elevation.
Any barometric elevation measurement with a phone probably is useless without scientific instruments. That also needs to be calibrated to weather since that changes pressure as well.
I noticed older phones had terrible GPS accuracy in any sort of difficult terrain (mountains, forest etc.). Sometimes it would place me a few ft next to a path i was on, sometimes more. Sometimes it would miss a piece of switchbacks and just add a straight line over some hundred yards where it should have been zig-zag and curves. Newer phones (I have a 3 or 4 year old flag ship phone) are MUCH better in the very same locations. Probably not 100% perfect, but i don't notice the obvious problems anymore.
I don't pay attention to elevation, but what I observed also must apply to elevation.
Any barometric elevation measurement with a phone probably is useless without scientific instruments. That also needs to be calibrated to weather since that changes pressure as well.
#15
Senior Member
Join Date: Feb 2016
Location: SE Wisconsin
Posts: 1,852
Bikes: Lemond '01 Maillot Jaune, Lemond '02 Victoire, Lemond '03 Poprad, Lemond '03 Wayzata DB conv(Poprad), '79 AcerMex Windsor Carrera Professional(pur new), '88 GT Tequesta(pur new), '01 Bianchi Grizzly, 1993 Trek 970 DB conv, Trek 8900 DB conv
Mentioned: 12 Post(s)
Tagged: 0 Thread(s)
Quoted: 759 Post(s)
Likes: 0
Liked 810 Times
in
471 Posts
First..everything discussed here is an estimate. Estimates are made on the resolution(& accuracy) of the base data and how it's processed. Accuracy costs money. Highly accurate data(on the scale of global terrain, or even just the USA) can be very expensive to generate. High accuracy files (datasets) are huge and are pricey to manage and process. At the end of the day it comes down to what level of accuracy is adequate for a consumer-level (free) mapping app at a modest cost.
Maps we see and use are very much composite images that pull data from many different sources, depending on the intended use of the "map".
While RWGPS may use google maps as a source, it may pull elevation data from a different source(elevation data is more of a core interest to RWGPS than generalized elevation info is to a consumer-level google maps user), or may use (process) a higher resolution elevation dataset than google uses. Given the volume of data google maps needs to process they may compress some source files(or processing) in the interest of increasing throughput/lowering costs. The accuracy of the output they generate is probably perfectly acceptable to most of their users, however an engineer building a road would be using much better datasets than what google uses.
It may well be that the focus of the OP's interest approaches the resolution of the base data..error increases as you approach the limits of the dataset.
We have knowledge where we have data. Actually measuring elevation of a ride using a GPS device during the ride (assuming the device determines elevation with reasonable accuracy) would probably be the best estimate of the actual elevation experienced as the sampling rate is very high (as it records the ride) compared to what google maps is using..which is probably generated by satellite(?).
On the grand scale..knowing elevation is somewhere around 2200 ft or 1400 ft over a 50 mile ride is better than no info at all, or an estimate of 4500 ft.
As mentioned above..ask RWGPS. Some analyst there can give you a very good explanation..if you can get access to him/her.
Maps we see and use are very much composite images that pull data from many different sources, depending on the intended use of the "map".
While RWGPS may use google maps as a source, it may pull elevation data from a different source(elevation data is more of a core interest to RWGPS than generalized elevation info is to a consumer-level google maps user), or may use (process) a higher resolution elevation dataset than google uses. Given the volume of data google maps needs to process they may compress some source files(or processing) in the interest of increasing throughput/lowering costs. The accuracy of the output they generate is probably perfectly acceptable to most of their users, however an engineer building a road would be using much better datasets than what google uses.
It may well be that the focus of the OP's interest approaches the resolution of the base data..error increases as you approach the limits of the dataset.
We have knowledge where we have data. Actually measuring elevation of a ride using a GPS device during the ride (assuming the device determines elevation with reasonable accuracy) would probably be the best estimate of the actual elevation experienced as the sampling rate is very high (as it records the ride) compared to what google maps is using..which is probably generated by satellite(?).
On the grand scale..knowing elevation is somewhere around 2200 ft or 1400 ft over a 50 mile ride is better than no info at all, or an estimate of 4500 ft.
As mentioned above..ask RWGPS. Some analyst there can give you a very good explanation..if you can get access to him/her.
#16
Full Member
Google maps uses SRTM data with a 90-m horizontal resolution. That is a grid of points on the earth where the elevation is known every 90meters apart. All points in between are interpolated between the nodes to calculate an elevation.
Most high end gps use either a better resolution data set or barometric pressure sensor to calculate elevation.
Therefore, the data sets will result in different elevations profiles.
Most high end gps use either a better resolution data set or barometric pressure sensor to calculate elevation.
Therefore, the data sets will result in different elevations profiles.
#17
Senior Member
Join Date: Dec 2010
Location: northern Deep South
Posts: 8,904
Bikes: Fuji Touring, Novara Randonee
Mentioned: 36 Post(s)
Tagged: 0 Thread(s)
Quoted: 2604 Post(s)
Liked 1,933 Times
in
1,213 Posts
I wonder if RWGPS processes the millions of rides they collect, and uses the tracks generated by barometric altimeter-equipped devices to improve the base maps' elevation gain estimates. Seems like it'd be a neat thing to do, and if they're running their own servers, it could be done with in a low priority background job. I'm sure every device would start with a different estimate of elevation, even from a fixed starting point. But take a few hundred rides, process the elevation changes from all the devices, perhaps do a weighted average based on how good a Garmin's altimeter is compared to a Bryton's. I'd think they could substantially improve on most base maps' elevation gain estimates for a given route -- if that route were sufficiently popular.
#18
Full Member
Thread Starter
Join Date: Feb 2015
Location: Pittsburgh, PA
Posts: 400
Bikes: Schwinn, Nishiki, Santana, Trek, Rodriguez
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 189 Post(s)
Likes: 0
Liked 106 Times
in
69 Posts
Apparently, google and rwgps use the same 90m interpolation for route planning.
https://ridewithgps.com/help/grade-and-elevation
https://ridewithgps.com/help/grade-and-elevation
#19
Senior Member
Join Date: Feb 2016
Location: SE Wisconsin
Posts: 1,852
Bikes: Lemond '01 Maillot Jaune, Lemond '02 Victoire, Lemond '03 Poprad, Lemond '03 Wayzata DB conv(Poprad), '79 AcerMex Windsor Carrera Professional(pur new), '88 GT Tequesta(pur new), '01 Bianchi Grizzly, 1993 Trek 970 DB conv, Trek 8900 DB conv
Mentioned: 12 Post(s)
Tagged: 0 Thread(s)
Quoted: 759 Post(s)
Likes: 0
Liked 810 Times
in
471 Posts
Interesting read.
As I mentioned above, estimates, estimates, and more estimates...and lots of data processing. The shuttle flying overhead and gathering data gets the job done over a large area, but it has its limitations. Undoubtedly a gross improvement over our broad-scale understanding of elevation prior to this work. In the end, accurate or fast, pick one.
What we need is the (I assume) google car that drives around and records the "street view" images to also gather accurate elevation data.
As I mentioned above, estimates, estimates, and more estimates...and lots of data processing. The shuttle flying overhead and gathering data gets the job done over a large area, but it has its limitations. Undoubtedly a gross improvement over our broad-scale understanding of elevation prior to this work. In the end, accurate or fast, pick one.
What we need is the (I assume) google car that drives around and records the "street view" images to also gather accurate elevation data.