*PSA* - Ranked Lists Changing based off LiDAR Analysis

[-wren-]

That matches what I see so far, Wren.

Here's some in the LoJ database that are almost soft-ranked. A few of these could end up being ranked.

I just looked at Mt. Maxwell, and it's actually ranked by 5 feet! The saddle had a spurious contour. I'm submitting this to John.

John's putting together a list of Colorado peaks that aren't in LoJ that have >240' of prominence. There could definitely be some new 13ers hidden here.

Also, of 13ers that have >320' of prominence that haven't been analyzed, there's definitely the possibility of some map errors causing some to be demoted, though this is less likely than the "error range" ones at between 300' and 320' that have already been looked at.

Almost Soft-Ranked.png

Sweet. I put the consensus so far into a custom list and will keep it updated for the sake of progress maps/tracking.

[davebobk47]

Has anyone checked Mt Muir in CA? I feel like thats the type of peak that could potentially have a significant change in prominence.

Thanks for all the work on this stuff! Really cool.

[bdloftin77]

Has anyone checked Mt Muir in CA? I feel like thats the type of peak that could potentially have a significant change in prominence.

Thanks for all the work on this stuff! Really cool.

No problem!

I don't think that area's available yet, via TNM at least. (https://prd-tnm.s3.amazonaws.com/LidarE ... dex.html#/).

Seano also mentioned that Barnard also might become CA's newest 14er (same area - no lidar available via TNM yet).

If anyone finds other CA lidar sources, definitely let us know! Death Valley's lowest point would be interesting, as well as Barnard, Muir, etc.

P.S. TNM has updated their green lidar coverage polygon for Colorado, however this time it seems to have overestimated coverage. El Paso, Las Animas, and Huerfano are correctly shown as available, but some areas north of I-70 that now show as available don't have any tiles pop up when you hold CTRL and drag an AOI. The greenish yellow polygon around Lake Granby should still have downloadable tiles. But there's still a gap to the west before there's more downloadable tiles. (as well as areas well north of I-70 that didn't previously show up).

[Candace66]

Incorrect/missing contours! ...There's definitely a big possibility of finding surprises in peaks that weren't even previously soft-ranked. Haven't yet heard an update from John about the P240 prominence list. Many of these aren't even listed peaks in LoJ, either.

OK, thanks for the info. I'm not shocked to hear of missing contours on USGS topo maps. For one example, it's a problem in the Coxcombs (southern California desert). I'll continue watching all this evolve.

And please do analyze Death Valley when it becomes available! I'll be very interested to see where the low point is, and it's elevation. It will also be interesting to see the LiDAR elevation for Mount Whitney. I'm not familiar with Barnard...if it's a reasonable dayhike, I'll be rooting for it to be promoted.

[seano]

OK, thanks for the info. I'm not shocked to hear of missing contours on USGS topo maps. For one example, it's a problem in the Coxcombs (southern California desert). I'll continue watching all this evolve.

And please do analyze Death Valley when it becomes available! I'll be very interested to see where the low point is, and it's elevation. It will also be interesting to see the LiDAR elevation for Mount Whitney. I'm not familiar with Barnard...if it's a reasonable dayhike, I'll be rooting for it to be promoted.

Barnard's between Whitney and Williamson. It's not the most remote thing in the Sierra, and it's only class 2, but it's still kind of a pain to reach in a day, either via George Creek or (as I did) from Whitney Portal.

I can imagine the Coxcombs giving cartographers nightmares, being a complicated field of huge granite blobs. Fun scrambling, but I wouldn't want to dig through the Lidar data looking for new prominence points.

[-wren-]

I'm not sure what the agreed margin of error will be for this "LiDAR soft rank list" that was talked about a while ago but I feel like they should have to be climbed in order for new 13er finishers to claim their title. With a 5' margin used there would be 591 peaks on the list, and 599 with the 10' margin.

Here are the seven peaks in the 5' margin:

1. "West Eolus" (299') tough one!
2. Emerson Mountain (297')
3. Milwaukee Peak (297')
4. Mount Rhoda (297')
5. Unlisted UN 13299 (296'), west of Mt. Princeton
6. UN 13401 (296')
7. "East Windom" (295')

8 more peaks in the 10' margin:

1. UN 13708 (294')
2. "S 3" (294')
3. "North Crown Mountain" (293')
4. Cinnamon Mountain A (293')
5. UN 13060 C (292')
6. Lomo Liso Mountain (292')
7. Van Wirt Mountain (290')
8. UN 13020 D (290')

[Eli Boardman]

I think the 10 ft. margin is unlikely to be needed except possibly on particular highly forested summits that also are known to have substantial rock outcrops (i.e. 10 ft. boulders indistinguishable from 10 ft. trees). The other case that a >5 ft. margin could be appropriate IMO is on EXTREMELY sharp summits, something like Ancient Art in the Fisher Towers for instance. I think on most normal, above-treeline peaks, 5 ft. is likely more than sufficient for LiDAR data that is of sufficient quality to be used as part of 3DEP.

[bdloftin77]

Has anyone checked Mt Muir in CA? I feel like thats the type of peak that could potentially have a significant change in prominence.

Thanks for all the work on this stuff! Really cool.

More of the California Sierras are in! John's cruising through them. Whitney is finally available, and came out at 14,500' even. Barnard, like Grizzly, is a mere 3' shy of being a 14er. Muir is indeed ranked!

[seano]

Has anyone checked Mt Muir in CA? I feel like thats the type of peak that could potentially have a significant change in prominence.

Thanks for all the work on this stuff! Really cool.

More of the California Sierras are in! John's cruising through them. Whitney is finally available, and came out at 14,500' even. Barnard, like Grizzly, is a mere 3' shy of being a 14er. Muir is indeed ranked!

Wow, Barnard came up short! I've been waiting for that one to get LIDARed for years, and thought it would more likely than not be over 14k' based on handheld GPS. I guess that guy was holding the GPS at chest level...

[bdloftin77]

Wow, Barnard came up short! I've been waiting for that one to get LIDARed for years, and thought it would more likely than not be over 14k' based on handheld GPS. I guess that guy was holding the GPS at chest level...

Yeah I was kinda bummed that CA didn’t get a new 14er! That would have been pretty cool.

[josephnephi]

This is a little random but I was curious if I could train a linear regression model to predict if LiDAR will promote or demote peaks. I trained it over the first 1000 LoJ peaks (by peak id) and got a model that was worse than just guessing!

For those curious about the details:

• I trained it over the following features: LiDAR Elevation, Map Elevation, Lat, Long, LiDAR Rise, Map Rise, LiDAR Saddle, Map Saddle, Isolation, and YDS

• I encoded the "solution" as ether -1, 0, or 1 for demoted, no change, and promoted respectively

• I tried a few higher-order solutions but they were prone to overfitting.

Just taking the average of the solution I got a mean average error (MAE) of: 0.035
This establishes the "baseline" and is as if one were to randomly guess if a peak would change

The model test set MAE (averaged over 200 trials) was: 0.057
This is worse than the average earlier, suggesting the model is worse than just guessing!

Some other fun data:

Weight Vector for 1st Degree Polynomial (Linear Regression):
LiDAR Elevation: 0.2579
Map Elevation: -0.2634
Lat: 0.0074
Long: 0.0067
LiDAR Rise: 0.5333
Map Rise: -0.5333
LiDAR Saddle: -0.4081
Map Saddle: 0.4050
Isolation: 0.0007
YDS: 0.0011

This suggests that the Elevation, Rise, and Saddle are the most important features. I was surprised to see Isolation and YDS being so inconsequential!

I would be interested if anyone has any ideas on how to improve the model!

[Eli Boardman]

I would be interested if anyone has any ideas on how to improve the model!

This seems like a perfect application for Random Forest, both for change/no-change classification and subsequent variable importance:

https://rpubs.com/mdwybron/487157