Elevation fixes for Eastern US (released)

[Alpine Scenery 1,609]

Note that the alpha version should be posted later tonight or tomorrow at the latest. Gets rid of the spires and elevated lakes and ponds, as well as reduces some of the "pop-in" effect from the DEM errors. I'm currently doing manual, but eventually also automated elevation fixes for the Eastern US and then I will compare the results to see which is better performance and otherwise.

This will remove MOST of the spires from Washington DC to Virginia to West Virginia to NC to Tennessee and to Georgia. Will post link once the alpha is ready. 
Right now, I'm about 50% done with the manual, once I get to 75% or so going to do the automated and compare.

Edited December 5, 2020 by SceneryFX

[Farlis 2,280]

So are you telling me that this wasn't fixed in the USA Update?

What kind of Update doesn't fix elevation issues?!

[Alpine Scenery 1,609]

It was made worse in some areas by the update, some remained the same, and a few (very very few) were fixed.
More were created by the update than fixed.

 

[ryanbatc 12,819]

Oh totally...the USA update created a TON of mesh/elevation issues in the US haha.  I don't really see the difference of mesh resolution but its supposedly better in some areas.

[RodgerC 58]

Elevation problems around KCMH (Columbus, Ohio) is pretty ugly at times.....

 

[Gaiiden 98]

the WV errors extend all the way up into PA as well. @SceneryFX here's a Google Map I made over the past few months since release that covers all the errors in the NE US (some are water/ground texture issues and improper buildings but the majority are elevation issues). I visited each airport to make note of improvements needed (I did 100+ freeware scenery airports in FSX to match MegaScenery Earth textures a few years ago & plan to eventually get into MSFS dev) and in the process canvassed the area for other problems as well

Edited December 6, 2020 by Gaiiden

[Alpine Scenery 1,609]

Thanks for the map, yes the worst area is from the border of Pennsylvania through WV and Virginia and then just short of the TN border.

I've got all that area finished, but I will have to go back later, this is going to be an "alpha release" with about 70% coverage and the coverage area will be 95% fixed (give or take).

For the first version, I'm mainly doing TN to NC  to VA to WV to DC to PA border (but probably not have time to do Ohio or Pittsburgh), stopping at PA border for now.

 

[BaronKen 144]

Thanks for your work on this! 🙂

[Alpine Scenery 1,609]

Here is the coverage map for this version, will update once the download link is posted over at Flightsim.TO

Edited December 6, 2020 by SceneryFX

[Jeeper 74]

8 hours ago, Farlis said:

So are you telling me that this wasn't fixed in the USA Update?

What kind of Update doesn't fix elevation issues?!

An Asobo update. 

[marsman2020 1,310]

Super disappointing that Asobo couldn't run some kind of an algorithm on their DEM data to identify and remove these. 

[Alpine Scenery 1,609]

It appears they tried, but the algorithm is failing. The reason I say they tried is because when some lakes are uneven (hence like a dam), they self-correct in elevation but if you are flying fast enough you will see them correct in real-time. 

 

Edited December 6, 2020 by SceneryFX

[marsman2020 1,310]

I'd expect this to be done on the server side before the data gets passed to the sim.

Eg, they would have some algorithm, it might have false positives, it would find however many 100s of of positives over the US, and someone would spend a couple of days reviewing them and either correcting them or flagging them as false positives, and the data would be corrected before it ever gets posted to the servers...

[Alpine Scenery 1,609]

They have errors in the data they used or the way they applied it, but I think part of the problem is they mixed too many different mesh resolutions and varying mesh from different places together. I agree that would be optimal, but it appears they are trying to rig it after the fact, and their post-corrective drawing is also hurting performance. The DEM probably loads in LOD levels (or something equivalent), but whatever it is doing is not correct, the game would probably be 50% faster if they fixed it.

 

Edited December 6, 2020 by SceneryFX

[marsman2020 1,310]

1 minute ago, SceneryFX said:

Well they have an error in the data they used or the way they applied it, but I think part of the problem is they mixed too many different mesh resolutions and varying mesh from different places together. I agree that would be optimal, but it appears they are trying to rig it after the fact, and their post-corrective drawing is also hurting performance. 

 

Yeah, that's just dumb.  With all the fancy "cloud computing" and "AI" they are talking about, and the need to maximize performance in this sim, there is 0 reason for them to ask millions of customers to run the same algorithms over and over on our individual PCs when they could be run once on the server side.  The only legit reason I could see for doing DEM updates on the fly would be for regions around custom scenery.