FB François, thank you for the tip. I believe this is a separate set of standards/classes than used by ASPRS. Very tricky!
Upgrading to the HD dataset will not only improve vertical precision to 10cm in France, but also horizontal accuracy to 0.5m. I’ll work on the plumbing and upgrade the AZs where possible. Next year, I can re-run to make use of the improved coverage.
Anyone else who knows of other good candidate datasets at 1-meter horizontal accuracy or better, do please let me know.
Not necessarily.
This thread has concentrated exclusively on computing the smallest contour surrounding the summit 25m below it. This is very useful because it is a necessary condition for a point to be within the activation area. However it is not sufficient, as there may be depressions within that area which fall more than 25m below the summit.
In other words, one simple contour may not define the activation area if the surface is not convex.
That being said, I don’t know offhand of any clear examples of activation areas which actually have such exclusions of any significant size. It is most likely to happen with summits on a plateau with large activation areas.
Martyn M1MAJ
Insightful, Martyn. Many such “donut holes” exist and you are spot-on that they are usually in large AZs. They are a-dime-a-dozen in the Midwest USA and Northwest France. The examples below are not yet live on sotl.as.
Most common are flattish summits, where noise causes the terrain to “freckle” right around the -25m elevation. These are highly dependent on the underlying LIDAR project’s filter tuning. I don’t consider these meaningful or useful to the activator. Example: W4A/CP-013.
It also happens when terrain is removed by mining. In two cases, I’ve seen entire summits be removed this way! W9/WI-033, below, is sort of beautiful in that they’ve left the summit itself and just enough elevation on the rim to allow some alternate operating locations.
Occasionally, a culvert is to blame. I don’t feel strongly about these, but maybe others do? Example: K0M/SE-003.
Consider that my algorithm only requests tiles that lie on the -25m contour, and it “walks along” the contour requesting DEM tiles until the contour closes on itself. This creates a significant time and space savings over previous methods, which requested rectangular swaths of DEM much larger than necessary. I had assumed that the savings was equally due to tiles skipped outside the AZ as inside; but you gave me cause to audit my cache and find that only 28 summits were large and convex enough to skip interior tiles. Some of those require ~90 tiles to infill, but that’s tractable.
So - I’ll go ahead and augment the AZs with “donut holes”!
-Mike
Thank you this is very good! Many Thanks, Bob W4BTH
Unfortunately, human stupidity knows no bounds - there are already news stories of inexperienced climbers taking on routes above their grade after consulting AI models for route advice and having to be rescued:
(Apologies for the low quality sources, I wouldn’t generally trust anything The Sun says…)
