[WIP] Atlas Maps of Valmere | From Real-World Height Data to Vector Map

[Skalimoi]

Thank you for the tutorial! I use a similar approach to make realistic terrain (mostly downloading SRTM data and then merging it in QGIS to use it in Photoshop), but I don't understand why do you need to use GDAL to convert the TIF to a readable file?

What I do is to simply right-click the merged layer, Export -> Save As -> and then change the Output mode from "raw data" to "Rendered image". This exports the full heightmap and it's perfectly readable by Photoshop. Maybe your method takes into account things that I haven't, so let me know if I've missed something

[morne]

Thank you for the tutorial! I use a similar approach to make realistic terrain (mostly downloading SRTM data and then merging it in QGIS to use it in Photoshop), but I don't understand why do you need to use GDAL to convert the TIF to a readable file?

What I do is to simply right-click the merged layer, Export -> Save As -> and then change the Output mode from "raw data" to "Rendered image". This exports the full heightmap and it's perfectly readable by Photoshop. Maybe your method takes into account things that I haven't, so let me know if I've missed something

Oh wow, just tried it. That's brilliant! Much easier and better than my method of manual re-scaling, since it actually retains the correct grayscale values. I'm going to update the tutorial to use this method instead.

EDIT: Actually, I realize now a fundamental difference between the two methods. The rendered image is only 8-bit. My method ensures that the final tif is 16-bit, which is much higher fidelity (65536 values of gray vs. 256 values). This becomes especially important when looking at detail at lower elevations.

[Skalimoi]

Oh wow, just tried it. That's brilliant! Much easier and better than my method of manual re-scaling, since it actually retains the correct grayscale values. I'm going to update the tutorial to use this method instead.

Glad I helped!

[morne]

Glad I helped!

Sorry, see my edit. The "rendered image" output is only 8-bit, which is unfortunate.

[Skalimoi]

Sorry, see my edit. The "rendered image" output is only 8-bit, which is unfortunate.

Ohhhh, I see, no worries. Maybe I should try working with 16-bit images then, I've had some problems with lower elevations that maybe get fixed by doing that.

[morne]

Ohhhh, I see, no worries. Maybe I should try working with 16-bit images then, I've had some problems with lower elevations that maybe get fixed by doing that.

It will only help. The difference is really hard to see with grayscale. Here's a comparison with a gradient filter on:

8-bit:


16-bit:


That stepping is definitely going to cause problems if you are trying to post-process your terrain in wilbur or similar terrain software.

[Skalimoi]

It will only help. The difference is really hard to see with grayscale. Here's a comparison with a gradient filter on:

8-bit:


16-bit:


That stepping is definitely going to cause problems if you are trying to post-process your terrain in wilbur or similar terrain software.

I see! Thank you. I thought it was just a resolution problem, something like the files not being high-quality enough and that affecting lower altitudes, but it seems that isn't the case.

[Michi il Disperso]

Wow.. that files are near 1Gb EACH!

EDIT: ... but they are damn beautiful!

[Rwhyte]

This is great! Thanks for documenting. And the Q&A along the way is great too

If it's ok to interject one more option...

...
At this point, you may be tempted to open up the tif in Photoshop and be off to the races. More likely than not, Photoshop will greet you with an entirely black image if you do this. That is because the data in the tif is scaled differently than what photoshop expects.

Yes, as you open these Tifs from QGIS (32-bit images) in Photoshop, you are greeted by a black or white image, but, there are some options... (At least, with Photoshop CS5? and later).
Under Image --> Mode --> You can convert from 32 to 16 bit:


And in that process, you're presented with options on how to re-map the 32-bit values to the 16-bit scale, defaulting to Local Adaption.


There are other options as well, Exposure adjustment would be like a linear offset. Or, Histogram Equalize stretches the image contrast to the 16-bit scale.
If it's just the elevation gradient as an image that's important, this is great. This is less ideal if the original elevation gradients need to be maintained, since this can shift in non-linear ways


One more caveat. While this is convenient, Photoshop is not equipped to handle signed (negative) values. I believe these are essentially truncated to 0 in the process. So, if the tif elevations contain negative values, and those are important, re-scaling like you've done in GDAL would be best. Here's a tif example with negative, below sea level elevations, truncated to zero in Photoshop, and lost forever, below the waves.
However, there may be many cases where negative values can safely be treated as zero, without loosing real information. But it does depend on the source data and location (eg Death Valley, bathymetric data, etc)

[morne]

This is great! Thanks for documenting. And the Q&A along the way is great too

If it's ok to interject one more option...

It is always okay to interject more opinions! There is always another way to do things. Doing the conversion in Photoshop is something I had not thought of, and could definitely be a more convenient way, especially if you don't want to work with the command line. Good note about the negative values, I think with the GMTED2010 data you wouldn't be truncating anything important, save for land depressions.