I just tried asia mesh and saw many details added to the terrain. However I am finding there is in general too much variation/unevenness throughout in what should be flat areas, like a rough sea, in areas that should otherwise be mostly flat for eg around omdb. Is the terrain actually that uneven and less perceivable so in real life or this is expected in fsx/p3d with high resolution mesh? I do not have this with default or fsgenesis where with latter there is variation but it is gradual that doesn't take away the feeling of it being flat.

I just tried asia mesh and saw many details added to the terrain. However I am finding there is in general too much variation/unevenness throughout in what should be flat areas, like a rough sea, in areas that should otherwise be mostly flat for eg around omdb. Is the terrain actually that uneven and less perceivable so in real life or this is expected in fsx/p3d with high resolution mesh? I do not have this with default or fsgenesis where with latter there is variation but it is gradual that doesn't take away the feeling of it being flat.

 

There is certainly high frequency noise that is resolved with the higher resolution SRTM data. However, how does one distinguish noise from real variability? That is a question all mesh designers ponder. I could use some spatial filters to remove some of the higher frequency data, but then I will also be removing some of the real variability that the high-resolution data permits. With anything there is a tradeoff, and I felt the SRTM data had the best balance between real variability and spurious noise in comparison to other datasets, such as the ASTER GDEM (too much noise!).

I have been wanting to use the mesh again since it contains additional details compared to other older lower resolution mesh I have been using. However the unevenness in flat terrain continues to somewhat bug me. If there is noise present at that detail level and is indistinguishable from actual elevation then shouldn't that mean the source data is actually meant for a lower resolution where it become possible to separate data from noise and should be decoded and used at that detail? How much of the preserved real variability would be visually actually reliable or discernible in the sim from noise?

On 4/16/2017 at 4:28 AM, him225 said:

How much of the preserved real variability would be visually actually reliable or discernible in the sim from noise?

You could probably cut out a good bit of that noise by going to lower res and using some spatial filter. But you also do lose a a lot of the data density (going to about 1/4 of the data density), so I don't think the tradeoff would be worth it.

It would probably be better at that point just to use Jonathan Ferrenti's lower-res SRTM dataset for the entire world.

For flat areas though, there wouldn't be much of a loss of real variability. But I didn't choose to diagnose the flatness of a tile and then do different levels of LOD. I didn't plan on ever going that deep when I made FreeMeshX.