TG Earth model progress.

[bigben]

Hi All

Figured I'd make a post for this as you know how I like to think out loud while I'm working on something   At Oshyan's suggestion I'm building a basic framework for a model of the earth in TG that will ultimately be released for everyone to use.  I did a few quick tests of some of the available data to the point of adding a river/lake layer. https://flic.kr/p/ra5iq5

For now, I'm crunching the elevation data  to get that bit right to start with as this is also proving to be the most time consuming.  Dealing with some of the problems in the data has confirmed for me the need for such a project.

• SRTM data is great, but it only covers latitudes between +/-60° and has issues around the Black/Caspian sea
• GMTED2010 has enough resolution with global coverage. but has a gap between latitudes -56° to -70° and the Himalayas contain some large areas of ugly hole filling
• GEBCO-30 has global coverage with some very nice bathymetry but it's only 30 arc sec resolution and I'm looking at a max resolution of 19 arc sec

No single dataset has what we need, so it's a case of picking the best available data from each. I currently have 500gb of raw data and will hopefully finish crunching the SRTM data down to size by tomorrow so I can start piecing it all together...

Stay tuned

[Roberts]

Perhaps you may find what you need at Jonathan de Ferrantis excellent site:

http://viewfinderpanoramas.org/

He has made available global 3 arc second data, partly just because SRTM has it's flaws with gaps. He also provides 1 arc second data for several mountainranges. A truly amazing feat and resource!

Good luck!

Robert

[Oshyan]

Looks like a promising resource Robert. Might save Ben some work, hopefully so! There's lots more to the proposed model though - color map, specular, subsurface water (as needed), clouds, and some other ideas.

- Oshyan

[bigben]

Thanks Robert.

I've just finished the base terrain, just a bit of bathymetry to go, but I'll still have a look at this data (nice accessible links for harvesting)

At the moment I've stalled a little bit on the ocean model. Getting some odd glitches in the shallow water which vary based on the angle of the sun.  Ocean is a second planet with a water shader, decay distance 1000.  Seems to be a linked to masking the surface. In the last image I removed the mask.  The green surface shows the position of the mask used for the water surface... any ideas for what to look for?

[Oshyan]

Try turning off Cast Shadows in the 2nd planet.

- Oshyan

[bigben]

That option isn't available on a planet(?)  I used a planet-sized sphere and that made a big difference to the water. After a little juggling I also got the opacity working to stop rendering water under the terrain (and flooding low lying land) Possibly still some minor glitches at the boundary of the two objects, but I'll deal with that later.  Image is 1000m decay distance over default base colours.

[bigben]

OK, so this brings us to beta 2.Water settings can be tweaked, this is just a demo.  Decay depth is 5000m... unrealistic, but an interesting visualisation. Ocean floor is just a light sandy colour. Blue marble texture is restricted to inland areas. This creates a few dark artefacts along the water's edge. This is an inescapable problem with the source image along with the loss of several small islands. Both of those issues will get looked at later as there are no quick fixes for them, but I have a couple of ideas.

Rendering without atmosphere at the moment so I can see what's happening, using a blue background instead. Lakes and rivers next..

[Dune]

Your progress is amazing, looks really good. Too bad about the coastal artifacts. The ocean floor can perhaps be made procedurally, it's probably dirt mostly anyway (xcept for the great barrier reef  ).
I also think you should make some more diversity in the waves, but you already had that planned I guess.

[bigben]

Yes, to start with I just want to keep the surface shaders simple as it's important to see where things don't work (and do work) While it's only a small number of components there'll be a fair bit of masking going on. 

[Ariel DK]

When I talk of creating a realistic Earth
model, I did not think there could be
terrain models with such resolution!
you done a great job so far (or I'm a little surprised)
however I'm curious .. how much memory the dataset sacrifice?

[bigben]

Let's just say that when I go home I edit the TGD to change the resolution of the TIFs from 0.01° to 0.02°/px.  (8Gb RAM at home, 32Gb at work..  64Gb on work VM).  I have elevation data down to 0.005°/px in 4 quadrants which kicks of TG at 11Gb RAM usage, although I can load a single quadrant at home without much drama.  The files may be big, but rendering doesn't increase RAM usage that much.

[bobbystahr]

Let's just say that when I go home I edit the TGD to change the resolution of the TIFs from 0.01° to 0.02°/px.  (8Gb RAM at home, 32Gb at work..  64Gb on work VM).  I have elevation data down to 0.005°/px in 4 quadrants which kicks of TG at 11Gb RAM usage, although I can load a single quadrant at home without much drama.  The files may be big, but rendering doesn't increase RAM usage that much.

Good to hear you can load them with 8G RAM as that's all I have...Looking forward to this product more than I can express. I've directed all my change to go to NWDA as that is a very cool store....and reasonable prices as well. Adding this to my wish list. Great work, no; mind blowing at least to me, work...

[bigben]

Yeah, 5 year old computer at home, so I'll have at least one version that won't be too hard on resources

[lat 64]

Yeah, 5 year old computer at home, so I'll have at least one version that won't be too hard on resources

Thank you for that
Puget Sound is very recognizable, great work! will I get to see Alaska too?

Will this be a purchased product?
Russ

[bigben]

Oshyan can clarify this, but AFAIK it will be free from a TG resource site. The source data is all public