November/December 2024 Challenge: The archipelago of endless wrecks

[Diamond]

Well, it looks pretty good so far. I think if you can get the bathymetry right, this will kick a$$.

[Redrobes]

I had another go at that and I don't think its going to happen. The issue is an old perennial one in that in order to get the bathymetry right it needs the sea bed raised and to do that means getting the undersea height to within a few m or 10's of m from the surface. If I break through the surface then it changes the coastline. Its so much easier to start with a 3D landscape and mess about with it as required and then at the end you have whatever coastline it gives for sea level. Trying to do this where you are setting the sea level coastline and trying to figure out the above sea level and below sea level relief to make that sensible is almost impossible.

I have been running the erosion passes and that in itself alters the landscape and changes the shape of the coastline too. A further complication is that because I created the relief from blurs and so on then its a bit blobby. I am finding that there are multiple lakes and spots which are highly unrealistic even with a sort of basin fill. That and my erosion app is a total pain to operate. Its like fitting a saddle to a velociraptor. On the one hand its quite powerful but one false move and it will eat you alive. Its good at generating deserts and flooded terraforms which look nothing like what you wanted from it.

I think I will bin my latest attempts to mess with the bathymetry and go back to where the last images here showed. Perhaps its best if I render it as an atlas style and not worry about the 3D effect of it so much. I'll keep going for a bit and see if leads to a dead end or not.

[Chashio]

Interesting progress. The process sounds almost more painful than doing it by hand, but the results have a nice look to them.

[Diamond]

I would pay a large amount of money to watch RedRobes saddle a velociraptor.

[Greason Wolfe]

I would pay a large amount of money to watch RedRobes saddle a velociraptor.

As would I.

Also, digging the effort and approach.

[Redrobes]

Interesting progress. The process sounds almost more painful than doing it by hand, but the results have a nice look to them.

Its certainly more challenging than normal. If you went at it from the 3D terrain and no restriction to keep the shape of the coastline matching the provided pattern then its a lot easier. Usually doing 3D is a lot more effort up front and then you get a lot for free down the line.

I would pay a large amount of money to watch RedRobes saddle a velociraptor.

As would I. Also, digging the effort and approach.

Thx - I would pay a lot to not have to ride one !

So this eves effort got limited success again and I got a bit frustrated with the rivers not being thin enough so I grit my teeth and hit the code base a bit more. I made it get the vegetation to suck up more water as a param and added another param to the ever growing huge list of knobs I can twiddle. I can now also set the vegetation rotting amount so it accumulates height where vegetation grows.

The idea here is that, where warm enough, usually next to rivers it has the conditions to grow veg well. When water flows, if its too much it kills the vegetation. So you have these wider flows of wet mud and the water spreads out. If the veg is allowed to grow next to it then it sucks up some of the water on the edge of the water flows and dries it out. If its thin enough water then the veg persists and now it drops height down there and builds up a bank then it forces the water to route around the veg and the water flow gets thinner. If all of the water is squeezed then it stays deep enough to keep the vegetation killed so it keeps the thin center of the water running.

Heres the latest run...

### Latest WIP ###
ThinnerRivers1.png ThinnerRivers3D_1.png

Theres a lot of nice things in the image but if I can highlight this area:

Regrowth.png

This is where the river used to be wider, then by turning down the rain and snowfall and letting it drain out a bit, the vegetation has regrown into that river bed forcing the river into a line again. Theres instances of this all over but most of them you need to know how bad the river was before it has regrown because in most cases it looks normal whereas it would have looked quite bad.

I had a few more ideas to add to it which might help but its a bigger coding effort. I wrote this app in 2007 so quite a while back when we had a lot less memory and you had to be a bit conservative with what data values per pixel you would store. But now with the 32 or 64 Gigs of RAM you can go quite bananas or have bigger terrain tiles to process at once.

What would be nice, and effective I think, is to have two types of height. One is rock, the other is soil. Soil is on top of rock and has wildly different erosion amounts. So once you have eroded through the soil you are on to bedrock and then it slows down a lot. So you can force the water into channels quite quickly without having these deep gorges slicing their way through the map - especially at the base of the mountains and big slopes which have accumulated a lot of water from large area and snow melt.

The rotting veg and sediment deposition can add to the soil height not the rock height so hopefully you get better river bed braiding without it cutting it into a deep channel. I don't think this app will ever do ox-bow lakes and that serpentine river flow because I don't erode sideways on this app. I think that would be quite hard to do but I have seen people write something where that does work.

Well I guess I have to decide what I am going to do with this map now !

[Diamond]

That is crazy that you can finesse this thing to take into account vegetation, river sediment, etc. Pretty amazing stuff.

By the way, I like where it's at right now.

[Redrobes]

Thanks. My app is sort of a terrain simulator. It has an emphasis on modeling water so that it can place down where the rivers are but its about progressing the terrain into the most realistic that I can get it to and spitting out image maps containing as much info about it as I can squeeze so that I have a load of source maps for the texturer.

It works by having an array of pixels each with a few parameters. Currently its height, fluid, sediment, some flow params, vegetation, temperature, distance, and some flags. The fluid can be water, ice or lava and there are a whole load of params acting on stuff overall. Theres some more param maps as well such as a temperature bias and rainfall then we can isolate some areas and have more or less rain there.

The general thrust of it is that you use the rain and snowfall to dump extra water onto the pixel and then move the fluids around based on the terrain gradient at that pixel. If you have too much gradient then you can erode the cliff directly and move a bit of the height to its neighbouring pixel.

As water flows over some height and erodes it away, you add the eroded amount to the sediment value. When the water is slow moving and deep you drop out sediment and add it back to the height at the new location.

Where its warm enough and not too high an altitude and has a little bit of water but not too much it grows veg. If its too dry or too wet the veg dies off. When you have veg growing then it sucks up a little of the water and also drops out a tiny bit of height for the vegetation rotting down.

If the temperature is below zero then it marks the pixel as ice and uses ice flow params. If that pixel with ice gets too warm it melts and converts to water and then uses water flow params. Ice doesn't flow too quickly but water does. As water flows it carries some temperature with it so its possible to have glaciers lower down the slope of a mountain than the basic altitude cooling would have formed it directly. If lava gets too cold it turns it all into height. Lava is only created via an image map of lava sources.

It's just lots and lots of that processing going on for every pixel over and over. It takes a while to run and consumes a lot of CPU time but it runs on many cores of a CPU so if you have 8 or 16 then it goes faster than 1 or 2. Each pixel of terrain is independent of all of the others except that it takes in and dumps off amounts to its neighbours. So you can break up the map into bits and process a bit per CPU core. Ideally it should run on the GPU then it would go a lot faster as you might have 8 powerful CPU cores or 1024 simple GPU cores to do it with and this processing is pretty simple really so it would work well on a GPU. It would be awesome to get the feedback to your parameter twiddles in real time. You could home in on your desired type of terrain climate quite easily.

But as you can imagine, it can easily get out of hand. A bit too dry and all the veg dies and you have a desert. Too much rain and your waterlogged all over the map. So its a lot of trial and error and may runs tweaking parameters until you get it just right. The real world has better feedback control and will sort itself out better than this app does. With all of this climate change stuff they keep saying the earth will die but that's ridiculous. The earth will be just fine. Its all of us humans who have adapted to just the right levels of oxygen, temperature, rain etc that will die and all of the other lifeforms that we are symbiotic with dying that will cause us the grief. A new evolution of life form will adapt to the new values and live quite happily with its newly adapted symbiotic lifeforms.

This app does not model wind nor tides. When water hits the sea its taken out of the processing so a tiny spot of sea can look odd. It only stores values as a vertical stack up on the pixel, so no caves or undercuts. There is only height - not rock types and strata. I don't erode sideways or have many effects that one pixel affects another. I don't model water pressure erosion and, well, there's a million other simplifications to whats happening in the real world. A pixel is often about 10m x 10m and one processing step is like several years of time. And water viscosity is way lower than I can get processed. Its trying to be realistic but in all honesty it just cant be anything close.

It all runs on the command line and has no user interface other than text. It so bad that quite often I need to get the code debugger out and print out just why its doing what it does because I don't understand what the hell its doing ! So its not for the feint hearted...and I have a bit of a love hate relationship with it.

So that's it in a nutshell.


Looking at the map I keep looking at the middle bit of it. The top left and top right have too many mountains and the bottom has too much sea - esp since its a bit featureless. So I cut off a bit of border to it. And I ran the original shape image into my "Put a ring on it" processing (see sig) and got some wave patterns out of that. I touched that up manually so that bays and inlets wont have big waves hitting it. Then added that to the texture engine.

I rather like this as an image but I don't know what to do to proceed to make it into a useful map. I am not sure whether to go with a 3D map which would be nice to look at but lower res. Or keep with the 2D one with all of its detail. And I am not sure whether I should take this terrain and reprogram the texturer to serve up an olde parchement map with ink lines for all of the bits. That would be more in keeping with the wrecks title. But it could be that its a modern map of WWII or similar battle naval wrecks. It would seem a waste to throw most of the nice colour and 3D effects into the bin to make an atlas style of map. Maybe I do a map which has the realistic and archaic style side by side.

### Latest WIP ###
ColourLandMiddle1.jpg ColourLandMiddle3D_1.jpg

[Redrobes]

I spent a bit of time programming up a new shader and come up with this sort of "old school module print in fixed colours" type of thing. It has a little texture effect to simulate the inconsistent print quality that you used to get in the 70's / 80's period.

Visually I like the realistic rendering but for this challenge I think I need to make a proper "map" not a satellite view. I am split in my mind about which way to go. What do you guys think ? Is the print style more in keeping with the challenge ?

### Latest WIP ###
GraphicDesignMap1.jpg

[Chashio]

You could do wrecks that are more sci-fi in nature, crashed spaceships and semi dead airships that may still even float. That could work well with the 3d view or 2d. You could have both side by side to show different aspects if you wanted. I don't think any particular direction is more in keeping with the challenge. Just do something you would love to see.

Also that software you've designed sounds amazing in what you can fiddle with. Would be very cool to see the changes in real time.