Dear Garmin... DTM/DEM Improvement for ClimbPro

crispgarmin
6 days ago

I was recently wondering if the DEM in my custom maps was used for things like ClimbPro and I think not, which would be a shame because there are very good local country models with OpenData available at high accuracy (e.g. UK OrdnanceSurvey Terrain50) that could/would give improvements to ClimbPro that always finishes at the wrong place (either before the summit or after a summit is reached and the ground plateaus) 

When my Edge 1040 boots up I noticed a credit to AW3D30 that sounded like a world surface model and indeed it is: 
https://wiki.openstreetmap.org/wiki/AW3D30

Unfortunately, as per the Wiki article "Warning: this surface model, as well as many others, including SRTM, has significant feature: it does not reflect a shape of a ground surface but reflects an envelope surface, "draping" over buildings, tree tops and other natural and man-made objects."

We ride on the ground, obviously, so using a DSM is a terrible idea.  It does explain why riding in forest tracks on climbs gives such poor climb analysis though.



Looking for comparisons in DSMs/DTMs/DEMs I found this research paper confirms that the Garmin-used DSM is the *worst* of the bunch
https://www.tandfonline.com/doi/full/10.1080/17538947.2024.2308734#d1e492

The summary of the paper is that there are better options.
Copernicus (EU funded access to geographical info) provides a very good worldwide DEM/DTM (not surface model, but elevation model) at 1second global coverage - GLO-30

Garmin, please get your legal team to read this and consider replacing the worst DSM with a good DEM/DTM and a whole bunch of improvements will naturally follow:
https://dataspace.copernicus.eu/explore-data/data-collections/copernicus-contributing-missions/collections-description/COP-DEM

You are welcome!

(cross-posted from Beta/X40 Series/Community Discussion)

  • 0 6 days ago

    the most accurate would be if they use the user's barometric datas and build up an own DEM database as the Strava did with barometric measurements (this is why it is accurate now, in the 1st years the Strava had awful DEM database and elevation charts for a climbs/segments). and release this updates as frequent as possible to the users.  (we dont need lidar-grade everywhere, we need it only where you can ride or run or hike, probably there will be always areas where nobody recorded anything before, this user has inaccurate DEM as now, but the 2nd user will have more accurate and better and better as more measurements are done)

  • 0 6 days ago

    so this is (" Strava build up an own DEM database from barometers' data") why the climbpro elevation charts are accurate with preloaded courses... the Garmin wont pay for every countries their lidar geo datas (if it exists, for example , complete Switzerland has it , as i know with 10cm vertical resolution but maybe already there is a much better one.) because nobody (i mean the end-users) will pay this costs . so always the own data is the cheapest, just they need to spend X workhours to write a code , probably this cost is also too much for Garmin, there is no enough profit year by year ... 

  • 0 5 days ago in reply to Luko79

    Whilst I agree big-data sourcing of heights along routes from the gazillion data points that Garmin hold is a good approach, I think that would be a bigger computer science / dev task than implementing a “better” available DEM/DTM to replace/update via the existing map update channels - it appears to cost nothing to licence for the 30m model. As someone who rides a lot where few others (tires) tread I would prefer a global model rather than a linear/routes based approach.

    just like I am sure throwing a billion FIT files at AI or an educational research project would result in a best-fit grade algorithm the issue is, as you say, willingness to address the problem rather than absence of a solution

  • 0 5 days ago

    From my experience, Komoot elevation gain is very accurate and has always been in accordance with my barometric measurement. By contrast, the predictions used in Garmin Connect are sometimes extremely off, by up to 50%. A recently planned route showed 10,000 m elevation gain in Komoot, but Garmin Connect predicted only 6,000 m. The following barometric measurement confirmed Komoot was correct.

    You only have to take a look at the elevation diagram (loaded course) on the Edge 1040. The gradient values fluctuate constantly because the smoothing algorithm is horribly configured. No one builds a road where the gradient fluctuates every 20 meters.