I have to admire the level of persistence that it takes to do this sort of thing by hand. Looking at the contour map, it actually looks pretty good to me. A little bit of smoothing shows that the mountain spine is actually fairly broken up:
I am regularly amazed when I look at an area that I've studied on a topo map and then look at a 3D image. My mental model from the topo map tends to be way off.
This looks amazing. What process did you use? Every time I try to work on wilbur with a contour map I find myself suffering at trying to remove the terraces that are formed (The deterrace command doesn't quite do it).
Though I'm still not convinced, I mean, it does look good for a small region, but the region I'm working on is relatively large (upwards of 1000km north-south)
I split the input file (Advance_3.png) into a set of masks by altitude:
I then loaded each mask as a selection and filled it with the appropriate altitude that you specified in your previous e-mail.
To ensure that I got the rivers in the right places, I loaded a coast mask with rivers (see below), inverted the selection, and offset the terrain by -50.
From there, it was add noise (25), fill basins, incise flow (amount=2, exponent=0.2), and then about 30 iterations of precipiton erosion. The terrain was about half its original height by then, so I scaled it by 2 to get back to the original height. For convenience (and to mask the fact that rivers are slightly off from their original location), I draped the original image over the surface in Wilbur (Shader Setup, Blending, select the original image in the Texture field).
At first, I tried a 2-level mask based off the Advance_2 image (I selected the red ridge lines and then expanded the selection to get the base mountain mask) I tried the CSU Johnsondale style of processing and some assembly in Photoshop and got the following result:
Then I noticed that you had a different elevation profile in mind.
Anyhow... I get carried away because this sort of thing is fun!
Unfortunately, I can't get really plausible results for much more than areas about 100km or so wide (maybe 100m per pixel). Getting results on a much larger terrain (say, 1000 km) is difficult because there isn't much in the way of detail at that point unless you're willing to go very large (say, 10000 pixel). Even then, the processing time becomes very large. For those sorts of surfaces, getting the basic features set up at low resolution and then resample by 2x and process for a few iterations gives niceish results (the Numenor map in this month's challenge was done this way).
Thanks,Waldronate!
When in doubt, research. I looked up maps of different mountainous regions and scaled them to match the scale I'm working at, from there I analyzed the shapes of the mountains and how they can be constructed.
Regular mountains are grouped in blocks, such as the ones that are marked here (The Alps, Google Earth):
These blocks tend to have a "Spine", a ridge or two from which other, smaller ridges, stem from. I studied the shapes and tried to replicate them in the map below:
Now I have to make the contours again based on this new map, hoping to get more accurate results.
A word of minor caution: those "blocks" are portions of watersheds. The boundaries of those blocks are river valleys. One of the reasons that those particular valleys are so prominent is that glacial action reduces the initial V shaped valleys cut by streams to U shaped valleys.
The Alps map you posted spawned a little niggle that grew into a fully-fledged idea. I could generate the high-frequency details (the river valleys) and then raise the baseline to look more mountainous. In the example below, the apparent blocks of mountains are emergent features of the river network, not any designed property.
In this case, I started with a 256x256 flat block of altitude, added noise, and did about 20 passes of precipiton erosion. I then clipped the terrain at about 75% of the total height, added more noise, and filled basins. I used the river map finding feature in Wilbur to generate a white map for rivers and black elsewhere. I then used the Texture>>Transfer>>Texture To Selection feature to get the rivers into the selection and then filled that selection with 0 altitude to deeply incise the rivers. Then back for a few passes of precipiton erosion, resampled larger and repeated the river finding/transfer to selection/fill/precipiton. After doing that a couple more times I ended up with a good approximation of the details. I then drew a selection around the area, added a mound, and added a gradient to get the central parts even higher.
OK, I really need to stop hijacking your thread. I should probably invest in video capture software and record some tutorials. Or fix the next version of Wilbur so that the operation recording things work.
Hijack it all you want, I'm learning a lot from your examples. So much that I'm actually considering making the maps in wilbur and then derivind the contours from that.
I still think your original elevations were really good, BUT the new version is better
great thread, just commenting to subscribe. keep up the good work
Using the info from your Advance_4 image (including the mountain ridges), the aforementioned process produced this:
This is showing a bit of promise. Good source material makes ALL the difference!
There are still issues with missing foothills and the like, but that's easily corrected.
I'm glad to hear my maps are good source material
Here is the new contour version made from the previous ridge and river map. And the old map for comparison:
Once more, the scale is logarithmic: 125m, 250m, 500m, 1000m, 2000m and 4000m. I didn't add the rivers this time because they made the map hard to read.
The southwest area is kind of underdeveloped because I need the adjacent maps to figure the elevations out.
Posting Permissions
Reply With Quote
