From the various dcrainmaker garmin reviews over time, I got the impression that he would rank elevation accuracy as:
Worst: unit with no barometer
Middle: using online elevation correction in garmin connect (uses topographical map data)
Best: unit with barometer (705, 910XT, etc)
My question is on those last two. Why isn’t the topographical map data ideal? Would that ranking vary if you take the location into account? In other words, what topographical data source does garmin use, and are some regions mapped more accurately than others? Maybe the topo data is spot-on for major metro areas, but less detailed for rural areas?
I wonder how they collect data, too… I’d assume it was all gathered from periodic airplane-based measurements… If they just drove around with barometer well… In any event, I wonder if there is some potential in garmin grabbing all the data from barometric devices for any given coordinate and coming up with some crowdsourced topography data. Perhaps the average reading for a given spot in NYC with tens-hundreds of thousands of measurements would be pretty damned accurate.
The reason I’m asking: topo/elevation correction says the half-mary route I did yesterday was 1350’ climbing, but my 910XT says it was 1722’. Huge difference…
Ha… think I figured out what was wrong with that run… the 910 took about a minute and a half to ‘warm up’ the barometer, it seems. Suddenly, after running about 90 seconds, the elevation jumped 400 feet in .05 miles. Without that i think they are pretty close. It does bring up a separate issue though - with this warming-up delay…
I have a hard time believing calculating elevation from a map is going to be correct. Topo maps do a good job of measuring the height of features like mountains, but measuring the heights of all the short climbs that exist in my hilly neighborhood? No way.
There is an answer to this - I’m sure. I didn’t do much research yet, but a quick google shows one technology used - I just wonder if the garmin source uses this data and, regardless, if all areas have info from the same sources or varying accuracy.
Airborne altimetric LiDAR for topographic data collection: Issues and applications
http://www.gisdevelopment.net/application/urban/products/urbanp0003.htm
Excerpt: Accuracy: An accuracy of order of 10 - 15 cm in the vertical and 50 - 100 cm in the horizontal is claimed by manufactures and has been demonstrated by many field studies.
I think I found the answers.
Short story: the elevation correction in garmin connect is accurate to at least 10 meters (it knows the elevation of a coordinate every 10 meters) over all of the US. It is accurate to 3 meters for certain areas - usually LIDAR mapped. This isn’t 10/3 meters of up/down. Rather, it means that every 10/3 meters you run, you hit a spot where it has a very accurate gauge of the elevation. So if you took a smooth rise/fall between the two points, it is spot on. If you crested some magical 3-meter wide 100-foot tall spire, you are out of luck. All of the above are updated very frequently (monthly or better). If you approach a canyon on a bridge, you should run down the canyon wall and up the other side to make sure your data is good. If you take the bridge you are on your own.
The last release notes for the data set including a map of resolution coverage is here: http://ned.usgs.gov/downloads/documents/NED_Release_Notes_Dec11.pdf
Sounds very accurate, and it looks like my city is LIDAR-mapped… just like ALL of west virginia… for some reason.
I would THINK this would be better than the barometer.
Outside the US there are other sources with varying accuracy… see below for more detail.
My data came from this thread:
https://forums.garmin.com/showthread.php?t=16046&highlight=topography
LARRY_F_S 
Junior Member
Join Date: Jan 2011
Location: Fort Worth, TX
Posts: 3

Thanks for the info Peter. Here’s what I also learned after some correspondence with Garmin Support:
- The elevation corrections feature of Garmin Connect is set to Disabled for the 705 and 800 since they have their own altimeter. With Disabled, Garmin Connect uses the elevation data from the 705/800 altimeter.
- The 800 stores track points based on changes in speed and direction. Changes in elevation don’t generate track points.
- Changing elevation corrections to Enabled causes Garmin Connect to ignore the elevation data from the 705/800 and use data from professional surveys. More specifically, the following is what a Garmin Product Support rep told me: “If you enable elevation corrections the website ignores all elevation data from the device and uses a combination of datasets to produce the best elevation corrections possible. Specifically, US coverage comes from the National Elevation Dataset from the USGS, Canada coverage comes from the Centre d’information topographique de Sherbrooke (CIT-S) dataset, and the rest of world data relies on data from the Shuttle Radar Topography Mission (SRTM) dataset from NASA. Elevation data outside of the US and Canadamay not be as accurate."
For my last ride of 26 miles here in north Texas, my 800 displayed an elevation gain value of 1273 feet. After uploading my ride data to Garmin Connect, the elevation gain value was 1273 feet (corrections Disabled) as it should have been. I selected corrections Enabled and the elevation gain value changed to 837 feet. That’s a big difference and I’m not sure which one is more accurate. If I had the 705’s option to record track points 1/sec, I think I would tend to believe the corrections enabled value for it.
The Garmin Product Support rep also commented on my which is more accurate question as follows: “Which is more accurate? That is a good question. Regardless of the device or method used it is important to realize these devices are consumer grade and can introduce a level of error that is not acceptable for survey level data. The Edge 800 is more likely to have more accurate elevation information than the Edge 705 so elevation corrections should be minor. Are the “elevation corrections” more accurate than the Edge 800? Using the auto-calibration feature on the device, the accuracy is +/- 50-125ft (same as GPS elevation). With WAAS and good reception you can obtain an accuracy of +/- 25-50ft. When manually calibrating to a known elevation it is approximately +/- 10ft for the first 15 minutes. Accuracy can change due to natural pressure changes (changes in weather). Potentially the information on our website may be more accurate as many things can effect the altimeter of your GPS. There is a limit however to this potential accuracy, as indicated on our website, “Elevation Corrections has known limitations for certain geographic features such as bridges and cross country borders where different elevation datasets prevail.” So in summary, in certain circumstances the Edge 800 will be more accurate than Elevation Corrections (and usually more accurate than the Edge 705), but overall the Elevation Corrections may provide the additional accuracy you need. Regardless of your choice, please be aware these devices are consumer grade and only survey grade devices should be used if sub-meter accuracy is desired."
Larry