Recently, I picked up a garmin 500. previously, I used mytracks and uploaded that info to ridewithgps.com. i decided i wanted something with a bit better data tracking, and to get HR, cadence, etc, so i picked this up.

one of the other things i was looking forward to, was accurate elevation data...oh, was i a fool for expecting this.

I frequently do group rides and start/stop at the exact same point, yet i get different elevation data. sometimes its not A LOT, but sometimes, i've looked and just said "what the hell?". I've also compared with others who i do the rides with, who also have garmin units, and they get different info from me...heres one - starting elevation, 835', ending (same spot), 916' - https://connect.garmin.com/activity/198310306

some examples -
tonights ride https://connect.garmin.com/activity/203160886

garmin connect elevation correction disabled - 1025' at start, 1034' at end. Elevation Gain: 738 ft, Min Elevation: 1,025 ft Max Elevation: 1,154 ft
garmin connect elevation correction enabled - 1025' at start, 1034' at end. Elevation Gain: 898 ft Min Elevation: 862 ft Max Elevation: 1,127 ft

ridewithgps - https://ridewithgps.com/trips/803219 1025' at start, 1041' at end, Elevation: + 714 / - 705 ft ...and if i go to "edit this route", it gives me an option to refetch elevation, and says "Using the elevation data from your GPS, your trip has +714 / -705 feet gain/loss. Our elevation data shows this route has +1378 / 1379 feet"

strava - https://app.strava.com/rides/14611377 1025' at start, 1034' at end, 704' elevation gain

NOW...before you say "the majority of it, you're looking at a 40' difference total, thats not what im concerned with.

when i check with the USGS website, which, i think i trust them more than anyone, when i check certain spots, theres a huge difference. on tonights ride, about 3/4 mile or so in when we turn left (north) onto cubbage road, right there getting onto the road, USGS has that spot listed at 965.24', ridewithgps has it listed at 1073', strava 1073', garmin connect 1075'. thats a 100+ foot difference. im sure strava/GC/RWGPS are pulling data from a similar source, but where are they getting their elevation data, and should i trust the USGS or the source for these sites?

when i use ridewithgps' "our elevation shows XXX feet", i check that spot, its 965' USGS vs 969' RWGPS corrected. a second spot, USGS 1012' vs 1013' RWGPS corrected. a third, 910' USGS vs. 905' RWGPS corrected.

so it looks like ridewithgps' elevation data most closely matches what the USGS has.

and i did see on my garmin that i could set controlled elevation points, i went out on a portion of this route, stopped at about a dozen points along a few points in this route where i had found elevations on USGS site, high spots at hilltops and low spots where they bottom out, as well as my starting spot which is from the same parking spot weekly, but i guess my GPS didn't care that i set an elevation point there

so who do i use to get the most accurate information, or who do you use/trust?

The Edge 500 has a barometric sensor, so you are best off turning OFF the correction.

You will still get some drift, just like your barometer on the wall changes over time. But it is small as your data shows.

You can set the altitude of specific points if you like; do that if you often start/stop rides at the same location (i.e. home, the starting location for club rides, etc).

The USGS data is generally pretty good AT THE SPECIFIC LOCATIONS/BENCHMARKS SURVEYED but those are relatively few and far between by bike riding standards; I'm not so sure that the roadway elevation data is all that good. And all that data needs to be "geo-referenced" into a common reference system... again more errors.

Don't make a big deal out of 30 or 50 feet out of 1000... it just isn't that much.

My $0.02

Thats the thing though, is that its not 30-50' when i base it off the corrected ridewithgps info...pre correction, it was 700ish-feet. When i used the RWGPS elevation data that more closely matched what i found on USGS, it had me at 1300+' of elevation gain. Thats like double...hmm

FWIW: RWGPS, MMR, USGS, etc, have rather coarse granularity for elevation. You can go up and down a lot on a ride without crossing a contour line every time. A Garmin with a barometric sensor will add this in, mapping websites may not. As a result the Garmin will always show more elevation gain than the websites.

Thats the thing though...the garmin is showing about the same compared to the sites data, but with the correction from RWGPS shows at the 3 points i compared to the USGS tool within 3-5'. I can use the usgs tool going up a hill i ride and it might show 969' and go up the.hill 30' and it show 971. I dont think the USGS site is going off a contour line, or at the least, has an accurate way to measure elevation exactly at the marker

Ill pull up the tool i used once i get home, see if i can look up elevation benchmarks and then see what the website marks them at.

And FWIW, heres some info on the dam we ride from...the parking lot on the west side of the dam, we ride from there, across the dam. The lot is pretty close to the same level (maybe a few feet if that) as the bridge over the dam. This article says top of the dam is 911' in elevation, so i can definitely see 25-30' drop from the bridge/path down to the water level/top of the dam

https://en.wikipedia.org/wiki/Hoover_Dam_(Ohio)

I use the Garmin Connect "Correction Enabled" data. Not necessarily because it's the most accurate, but because I simply want consistency, and it seems to be pretty consistent when comparing identical rides/routes (like commutes).

For instance, the elevation for my commute to work can be anywhere from 90' to 165' elevation gain. The Garmin barometer usually gives me about 150-155'. Garmin corrected data gives me about 125-130'. I just count it at 120'.

Just for fun, I picked a few landmarks on an 1,100' climb near my house, and got the elevation data from Google Earth. When I passed these spots, I looked down at the Garmin (Edge 500), and the elevation that it showed me was spot on...but that doesn't necessarily mean that the gain/loss will be perfectly accurate.