Wilbur: Incise Flow vs Fluvial Erosion for World Creation

[Eusebio Ptolomeu]

(sorry for bad english)

If one wants to create the topographic map for an entire world, what's the best method to use in Wilbur? From my experiments, generally speaking, Incise Flow tends to generate better-looking mountains, with noticeable river valleys, watersheds, etc. However, to use the Incise Flow operation, it's advisable to use the Fill Basins operation, and this have the downside of connecting all the basins with the ocean, in a way that you won't get any endorheic basin. With fluvial erosion, there's no need to use the Fill Basins option, and you can keep endorheic basins, but this method does't create nice looking mountains as far as I seen. Is there a suggested way to mix both of these methods in a single map? I've been trying to find a way to do so. I've tried playing around with the different tools for slection in Wilbur (select by height range, basins, facing direction, etc.) but would liek to know if there's a proper tutorial to deal with this. In a sense, I could say that I kind of wish to mix the Volume 1 and Volume 5 of the Fun With Wilbur tutorials.

[johnvanvliet]

If one wants to create the topographic map for an entire world, what's the best method to use in Wilbur?

i remap the north and south poles to polar-steriographic ( using Gimp 2.10 )

then run the erosion cycles on the poles and the center

when i am done i then remap the poles back to simplecylindrical ( using gimp) and blend the 3 images together

wilbur dose not do the polar area well , this is why i erode the poles and the middle separately


i work with 32 bit floating point height maps ( bt format) in wilbur and use GDAL to convert them into a tiff ( and to bt ) to use in gimp
now you NEED to have sufficient height -- i use 0 to 65536 - the 16 bit image range in my Height Maps

This gives me a rather good erosion look after only 3 erosion's and one incise flow


trial and error is a bit of your friend here

try a few things and if you do not like the output
just restart from the beginning


this is a close up of just 3 erosion cycles on a full planet , no incise flow
test.png

[waldronate]

Wilbur's Precipiton Erosion feature is very broadly similar to a fluvial erosion system, but differs in a few critical ways: there is no global water memory and it's purely altitude-delta (difference in altitude between adjacent samples) based. The lack of a water layer results in a lack of lake formation and no deltas (things that look like deltas are usually due to the landscape's slope flattening near sea level). I stole the name Precipiton (it was that or waterbot) from some group that published a paper or two. The implementation in Wilbur is pretty simple: drop an agent on a random location on the surface and let it push its way downslope, moving some fraction of the altitude difference along with it. No hardness layer, no sediment layer, no water layer, no water deposition, nothing but single-minded agents plowing their way downhill. I made it multi-threaded a while back and it's totally the wrong implementation: each thread walks along without regards to the others, pushing one agent at a time. The agents might walk on each other, but they are statistically unlikely to. It's not an algorithm that most folks would describe as fluvial, despite the insistence of many folks that the menu says "precipitation".
Incise flow is just about as simple minded: at every point on the surface, an agent walks uphill from cell to cell, updating a count for number of visits at each cell. The final result is a map of the count of the number of cells upstream of that point. That count is defined as proportional to the amount of "flow" upstream, which would be true if "rainfall" was uniform across the surface. The logarithm (base enterable in the dialog) of the count (possibly averaged with its neighbors via blur) is used as the amount removed from the surface. Due to the way that the count is computed, however, the result is that there are little divots near where flow streams merge.
I recommend looking at https://www.cartographersguild.com/s...ad.php?t=29412 (especially the CSU Johnsondale one).

The algorithms in Wilbur aren't really suited for whole-world operation because they assume cell-to-cell connectivity that's planar. If you can get smaller areas that are reasonably planar, then you'll be fine. Otherwise, you'll need to reproject your areas to reduce the errors.

[Eusebio Ptolomeu]

Thanks for the answers! I was aware of the limitation of Wilbur generating distortions in a planet-size topographic map, since it doesn't take the spherical geometry into account. However, my question here was more of the correct methods to achive good-looking results in such a map. Basically speaking, I was looking for a way to preserve endorheic basins even after using Incise Flow. For example, would it be possible to (even if outside Wilbur, trough image editing programs) create a map containing the nice-looking valleys generated by the Incise Flow feature in Volume 1, while at the same time preserving the inland sea that we get with Volume 5? We won't get the lakes feature in Volume 5 doing the procedure with Volume 1, because the Fill Basins tool connect the basins. Is there a way to have both of them together?

[waldronate]

Incise flow can only work on connected flow networks. Precipiton erosion is likely to eventually create a connected network, but it will take a long time and fill up a lot of basins along the way. Filling basins is a shortcut to creating a connected network. The only way that I have found to preserve basins is via the "basins as lakes" technique of volume 5 and that technique only preserves the appearance of basins to make it appear that rivers are flowing through lakes. Wilbur lacks a good way to blend two surfaces, meaning that reinstating basins by subtracting the basins from lakes image from the filled and incised image needs to be done in an external program.
I described the algorithms to show that doing worldwide flow in wilbur isn't going to be easy or particularly plausible in many cases. It's just not what it was built for

[Eusebio Ptolomeu]

In your opinion, what's the best way to use Wilbur if oe wants to create a world map? I saw the tutorial involving Wilbur and FT3, and I actually liked the results there, only problem was the lack of endorheic basins, which is necessary for Incise Flow to properly work. I'll try the CSU Johnsondale tutorial here, he managed to get those dendritic-like mountains that I wasn't being able to generate with Precipiton erosion alone.
One more question about these algorithms: do they take into account selected/unselected areas? I know that if we select a region of the map, the effects of Incise Flow (plus everything else) will appear only on this region, but does the algorithm knows to only "calculate the erosion" in this region as well? For example, in the Precipiton erosion as you mentioned: if there was a region of the map selected, would the algorithm only "drop" agents inside this region, and do all the erosion calculations based on this limit?

[waldronate]

Incise Flow should calculate its flow on the entire map, but only write in the selected area. Operating in this way is important to ensure consistency with or without selected areas.
Precipiton erosion, however, only drops agents inside the selection and only lets them do their job inside the selection (they terminate when they pass out of a selected area). Just programmer laziness, I guess.

If you're set on using Wilbur as a world creator, start by creating a world full of islands with no large land masses. Then output the individual islands using a projection centered on the island and process each individually. Seriously, though, that sort of processing is about as good as you'll get with Wilbur, even if you apply it to more continental-scale things.

[Eusebio Ptolomeu]

Do you know if is it possible to use Precipiton Erosion to get mountain ranges with these dendritic-like patterns we achieve with Incise Flow? For example, the mountain range you achieved here, with the watersheds and the blocks of mountains. I'm trying to play around with the parameters in the Precipiton menu: reducing the Delta value to something like 0.1 or 0.05 seems to work in the right direction, but all river valleys are eventually smoothed-out by sucessive passes. On a side note, is it possible to "blur" the results from the precipiton erosion the same way we have with Incise Flow?

Edit:
I've just realized there was o Incise Flow on the post I linked, only precipiton erosion I tried the procedure you suggested there, and it really did generate the watershed-like features I was looking for. Only problem is that it generated some "random spikes" troughout the figure, little spiked elevations the size of a pixel. Doing an morphological->erode pass or a blur pass solve this problem, though.

[waldronate]

The major problem with Wilbur's flow calculation is that it's calculated as either 8-way or 4-way connectivity to the nearest sample. Either of those connectivity options will pretty quickly lead to up/down, left/right, and diagonal artifacts. The purpose of adding noise is to confuse the flow finder to minimize the visibility of the artifacts.

The river map finding feature referenced in the pseudo-Alps description in the Muna thread is pretty much the first step of Incise Flow (finding flow) with processing to go to the texture channel rather than the altitude channel. It has the same problems with short channels that Incise Flow does. Note that you can make a block of mountains with rivers generally where you want them by drawing your own mask with the main river trunks already present (it's pretty much the same as the regular operations).

One interesting tool that you may not have played with is the Max Length option on the Precipiton Erosion feature. Changing it to a positive value greater than 1 (for example, 2) will move altitude downhill by a limited amount (I think it's n-1 steps). This operation results in what is pretty much heat diffusion (it's a randomized heat diffusion that gets little spikes due to missed items). It's a rough approximation of thermal erosion, just as the precipiton erosion is an approximation of fluvial erosion.

You can make a few mountains and use those as brushes for stamping rough shapes, if that idea helps any. Being able to pick brushes randomly from images in a directory is an idea on the wish list.

[Eusebio Ptolomeu]

I've come up with a procedure to use Incise Flow and no Fill Basins, therefore allowing some endorheic basins to persist. I would like to know if the results are good, and where does it fails short of realism. The procedure goes as follows:

1- First, I generate a heightfield using this equation (with Mathematical -> Calculate Function):
if((lt(pow(pow(sinc(r/5),0.25),10),0.25),pow(pow(sinc(r/5),0.25),10),0.25)
This generates some sort of circular plateau. It's basically a radial gradient capped at 0.25m (to be fair, this is just so I create an Island-like landmass).
WilburWorldNew.png

2- Then, I multiply this heightmap with some fractal noise:
WilburWorldNew.png

3- I then use Mathematical -> Span of 0 to 1 (for some reason, smaller spans seems to generate less straight rivers for me) and do 25 passes of precipiton erosion, with 1% noise per pass.
WilburWorldNew.png
This generates some nice plateaus, though I see their boundaries with the mountains are pretty sharp, and not smooth, because of the small height span I used.

4-Now, I do one more pass of Precipiton Erosion, one single pass with no noise, just to connect the system a little, but not in it's entirety, for I want to preserve some lakes.

5-Then, I use the Incise Flow feature. I like to use an exponent of 0.1 to get more rivers (I'm probably overdoing it, though), and a Pre Blur of 0.75. The result is this:
WilburWorldNew.png

6- Finally, I do another 25 Precipiton Erosion passes, without any noise this time. I end up with the following result:
WilburWorldNew.png

The result seems to preserve the nice-looking river valleys carved by Incise Flow, while allowing some lakes to remain. That big yellowish plateau is an unconnected basin, as can be atested by the Compute Basin Deltas result below:
WilburWorldNew.png

So far, this procedure is the closest I got to have both Incise Flow nice-looking rivers and allowing some endorheic basins to still exist. I would like to know what do you guys think of the result, does it look ok?