Reworking global datasets

[bigben]

Hi All

I have a visualisation workshop coming up and I wanted to include some TG stuff especially now there's an education license.  For this (and a couple of other projects) I'll be updating my global DEM data with newer data. One thing that is a bit of a pain is that you can't rotate DEMs with the planet so part of this will be a workaround to create a modified dataset of unsigned 16-bit geotiff that a) includes bathymetry and b) is loaded as an image map so that it will be a bit easier to animate the globe.  The resolution of this will be limited by how much you can squeeze into a single image unless it's possible to do a partial spherical projection(?).  I can get the full resolution data for SRTM15+ into TG as DEMs as a 2x2 tile set.

[bigben]

2 test renders, Earth rotated 40deg on Y-axis in between (and background size increased to include the moon... a few things left to remember )  36K terrain and texture, blue shader below 0m to check altitude.  Since unsigned 16/32-bit geotiffs only handle positive values I added 12000m to the elevation data before exporting the DEM. In TG then, displacement multiplier set to 65536  and offset to -12000 using the image as a 0-1 scalar function.

[Dune]

Great work, and good solution. I'm eager to follow your progress.

[bigben]

Thanks..  The other good thing with this is that by using geotiff you can keep the same model but reset the orientation (0 rotation and apex set to 90,0) and then switch in higher res geotiff... or at worst swap the image nodes for geoimage ones. Caspian and Aral seas (well sea and depression) are buggers in that they require a bit of extra masking as they're below sea level, so I have an ocean mask (excluding Caspian Sea) for terrain masking and a water mask (actually various water masks) for surface masking. Rebuilt my sun-direction vector mask from a screengrab in here and added the city lights.  Luminosity of 0.5 is enough to be visible. 1 is a bit to strong. 

This image has lights off at 90deg to sun and then fully on 3deg further around.  Might shift both more towards the night but luckily both settings are just scalar constants.  Back to a flat ocean with a basic reflection. Have to dig up my water mask with inland lakes off another disk. That makes a big difference in a lot of places.  Gotta read my notes on the moon as well... it's facing the wrong way

[Ariel DK]

welcome back, just one doubt:

The resolution of this will be limited by how much you can squeeze into a single image unless it's possible to do a partial spherical projection(?)

what do you mean with this?

[bigben]

With spherical projection you need a single image with 360x180deg.  With DEMs it doesn't matter as long as they are lat/long projection.  You want at least 16-bit for elevation and it doesn't take too long to get to a 4GB TIFF.  A 36K texture (0.01deg/px)  is 1.2GB 

SRTM15plus data is 0.0046deg/px. To load that I have 4 images covering 180x90deg weighing in at 1.8GB each.  If we could specify size and position as a ratio of 360x180 (assuming the image is also 2:1) we could load these as size = 0.5 and position lower left at 0,0.5  0, 1 0.5,0.5 0.5,1   but I'm not sure how difficult that would be code.

Or support BIGTIFF?

[Ariel DK]

Thx for the explanation.  0.0046deg/px... how many meters per pixels is that?
over there are some websites and softwares with real coordinates calculators that can be useful

and there is some computers in the world that still using 8 gb RAM, so no, please, dont support BIGTIFF

[bigben]

I still have the same PC at home but I'm working on the train with a kick-arse laptop (i3 with 8GB RAM, files stored on a 32GB SD card)    TG handles lots of images pretty well. Takes a while to open the file initially but then it works pretty well. I'll convert the res a bit later (GlobalMapper lets you sompecify res in alternate units on export so it's nice and easy)  ... around 320m

[Oshyan]

Hey Ben, great to see you back and working on this project again! Current/new results are looking great.

- Oshyan