Oct / Nov '17 Lite Challenge: X Marks the Spot

[waldronate]

I decided to go with the squarish one rotated to align with the map edges for plausibility reasons. I'll start with a mask like I usually do. Mountains arranged to form a central area. I'm aiming for roughly 300 feet per pixel, maybe with some decorative bits.

### Latest WIP ###
wip0.gif

[Bogie]

Welcome to the Challenge,, off to a good start!!

[waldronate]

I think that I may want to scale back a little. Maybe 50 miles or so across. Possibly I'll go with a different rendering style. It depends on what I get time to do.
### Latest WIP ###
wip1.jpg

[tilt]

scaling back is always good LOL

[waldronate]

I really need to try something different one of these days.
### Latest WIP ###
wip2.jpg

[Redrobes]

Squar ?!!

As usual your names are epic ! That is some very nice pattern on the snowy slopes too. I can never seem to prod my app into getting anything as nice as that.

[waldronate]

It's not completely Square, you see...

The pattern comes about from a combination of the process-inflate loop with the processing as incise flow (find flow at each pixel, subtract flow^exponent) followed by precipiton (gradient-following automata that move altitude from higher to lower) erosions. If I recall correctly, yours only does the precipitation part, but actually tries to do volume-preserving operations and keeps the water levels and sediment elements from pass to pass. Incise+Precipiton is waaay stupider and results in a lot of physically implausible parts that turn out to be visually plausible.

[Redrobes]

It's not completely Square, you see...

I didnt catch that and thats even funnier.

Yeah I drop rain and find the difference between height+water at each pixel and then move some of each waters pixel to the lowest next neighbour pixel. It maintains some momentum about the flow as well and the flow plus some forumula about the flow & height of the water (fast, thin = lots) erodes the height down and converts it into sediment. The sediment is carried with water and a different formula based on the flow & height of the water (thick, slow = lots) drops some of the sediment back to height again and removes that volume from the water. I find that with the sediment I get nice wibbly rivers but I seem to always get one river which cuts deeper and deeper into the rock and I dont get that nice pattern that you get on your mountains. As more water channels up then I get more flow and more erosion and its a positive feedback situation. I dont think my ravines being formed are all that plausible unless its like limestone.

[Mouse]

Very funny map, Waldronate

I love your sense of humour!

Being the little devil that I am, I sometimes wonder what kind of app would result if you and Red ever got together to make a sort of combined erosion thingummy

o...k... I think I should.....RUN!!!!

[waldronate]

I dont think my ravines being formed are all that plausible unless its like limestone.

One of the major differences of both our systems from real-world conditions is that the material to be eroded is of uniform resistivity. In real-world conditions, sediment is much easier to erode than bedrock, underlying bedrock is most commonly non-uniform in resistance to erosion, there is an ongoing reduction in resistance of the outer layer of exposed bedrock due to chemical weathering, there are mass-wasting effects (e.g. rainfall diffusion on sediment, landslides), and finally there is an ongoing uplift of the terrain along certain lines due to tectonic forces. Then there is a problem that plants affect erosion rates dramatically and that the particle size of eroded materials makes a huge difference in final eroded appearance. The very simple fluid dynamics approach results in things eroding like big mounds of fine mud, which both systems show.

The fun frilly bits that show up in the process that I use look suspiciously like the dendritic drainage pattern formed in tidal mud flats. That's an effect of the noise (removes straight paths) + basin fill (establishes connectivity) and incise flow (entrenches continuity based on assumed uniform flow) technique. If you look carefully at the results after incise flow, you'll see that there are unnatural basins near the juncture of river segments that are caused by the non-physically-based blur option on incise that smooths and widens areas. The way that a lot of those things stay in the landscape is that the scale-and-process loop establishes large-scale features using the same processes on a smaller surface. There isn't enough energy available in the processing parts to get rid of those major features when more pixels are processed and the common use of a bicubic interpolation scheme acts as a nice smoothing function between levels (using nearest neighbor gives a subtly different effect). I also tend to get nasty spikes out of the processing because the precipiton implementation randomly drops agents onto the surface and is likely to missing some elements. My final step of morphological erode (local minimize) cleans out the spikes.

The stuff that I tend to do is really just a big ball of trial-and-error. It's interesting (it was for me, anyhow) to compare my recent results with my earlier ones and see what sort of change each successive addition caused. Sometimes, seeing people flail about resulted in better features in Wilbur (e.g., the abuse of incise flow to get canyons as river proxies in the Israh map led directly to the Find Rivers feature in Wilbur) and sometimes, better implementations led to better results (e.g. the original distance metric computation in Wilbur used Manhattan distance calculations and had peculiar artifacts; switching to a bit of code that used a parabolic metric gave much nicer results).