I'm looking for wisdom on the drastically different elevation changes I see post-ride on the various hardware & apps I'm using. I typically use a Garmin 200 Edge to track my rides live, then Garmin Express to upload the file to Garmin Connect, where I also have it set up to send to Strava and Map My Walk (using just because I have legacy workout files there). I'm finding that the elevation change reported on each of these is wildly inconsistent. I do understand the Garmin's "elevation corrections" feature, which allows you to either use the Garmin's own raw data or enable elevation corrections, which supposedly cause the elevation to be calculated from professional land surveys. Even accounting for that feature, the elevation reports I'm seeing are wildly inconsistent.

Here's an example of elevation reported from a ride I recently did:
Garmin Edge 200 elevation reported, with elevation correction disabled = 1,483 ft
Garmin Edge 200 elevation reported, with elevation correction enabled = 811 ft
Strava elevation = 1,137 ft (which is odd, because the file came directly from the Garmin software)
Map My Walk elevation = 927 ft (again, very odd because the file comes directly from the Garmin)

I even went into the Garmin elevation profile for the ride and counted up each individual hill's elevation change (a mind-numbing exercise) and got over 1,700 ft of climb.

Any thoughts on why these are so different, and which is likely to be the most accurate? Thanks

Someone else had the same questions like you and went into a lot of effort to explain the differences:
Jeffrey Friedl's Blog » The Voodoo of Elevation Gain and Strava (and How I Get Around It)

TL;DR: Strava's iPhone app has access to both iPhone altimeters (barometric and GPS/GLONASS), but in the end it uses neither. Rather, it throws away that elevation data and instead relies on its own database of elevation data

Whichever one says I climbed the most is right.

Elevation gain based on free mapping databases is always flawed because the free databases are usually averages over a region of land and don't consider man-made features like bridges, tunnels, or roads along the side of a cliff. Also, your computer may record your position infrequently so the course it records is different from what you actually rode. A bicycle computer with a built-in barometric altimeter and sampling several times per minute will almost always be much more accurate than a database-based elevation gain. Some Garmins can do this; don't know if yours does or not.

On my iPhone 6s, I've usually found Strava to under-report elevation gain compared to the RideWithGPS app as well as my Garmin GPS unit with baro altimeter. RwGPS and the Garmin are usually in agreement with each other though (the times I've checked it anyway).

My own feeling is that barometric altimeters are going to be the most accurate, but if a front is blowing through and the ambient air pressure is changing through the day there will be errors.

+1. My regular route includes a segment of road on a bridge over a highway in a ravine. Strava always shows that segment as a ~100 foot drop followed by a 100 foot climb.

Same here. I imagine the databases will be fixed at some point, because it's going to annoy enough people to motivate the engineers to make it better.

I'm not sure that's enough motivation compared to the $5.99/month they pull in from many users . . .

More likely if Garmin ever decides to push Connect in a way like Strava.

Maybe Garmin doesn't care or doesn't have enough financial incentive to care, but google definitely does. They are improving their services frequently and constantly.

RideWithGPS seems to have its own database, just like Strava but it uses that only to estimate the elevation when planning a route. For the trip, it uses the altimeter readings from the phone. That would explain why the two don't match.

I take the actual elevation climbing with a grain of salt. Who really cares about those short rollers, 5 feet or less tall? Even gradual assents of say ½% rise can be difficult to notice, and certainly aren't comparable to the effort to do a 10% climb. At the same time, it gives a good estimate that can be compared across rides.

I would like to see a rating of say hills > 10 meters (33 ft) climbing, and 2% slope.

As far as accuracy, Strava should be able to start compiling their own database of elevations of primary bike routes. Take the GPS data of say 100 rides on any single route, and the data should average out pretty well, although it is interesting watching the real-time elevation data on RideWithGPS. I'll frequently be climbing a hill, and look down to see my elevation dropping.

Strava has an issue of figuring out when a person is actually ON a segment. We've got a local overpass with some riders going over the overpass, and some going under it, and someone defined a segment going over the top.

https://www.strava.com/segments/1062935

Note, I can assure you that the bump shown in middle of the segment doesn't actually exist. But, if you go down the leaderboard, most of the top few riders went under the bridge. It is complex as GPS coordinates never quite match up with the roads, so Strava allows for some variation which allows for jumping between segments.

Here's a switchback posted by someone else that Strava can't differentiate between those climbing and descending.

https://www.strava.com/segments/3267169

Still, those close parallel roads with divergent elevations are the minority of routes that (road) cyclists take. So, in general, averaging elevation should do reasonably well. A few roads might have significant banking which could be an issue, but assuming enough rides on the road, one could treat each direction of travel independently.

Oh, one of my "fastest" descents, Strava had me riding through the bushes.
For a false 57.7 MPH

False 57.5.jpg

But, that is also a good indication of why these precise calculations are so difficult.