[Mod][Merged] Valleys Mapgen [valleys_mapgen]

[duane]

Rivers running through my deserts annoy me. I've tried getting the biomes to cooperate and put just a bit of vegetation around them, but it always adds too much complication to the code. So, I went the opposite route -- rivers gradually dry up as you approach desert regions, leaving nothing but desolate channels and endless thirst.



I'm still testing this, but if you'd like to try it, it's on github.

[Ivà]

+1 @duane. Good idea.

[Gael de Sailly]

Is that normal to find a dry river in the jungle?

screenshot_20160212_121623.png (1012 KiB) Viewed 657 times

(imgur won't upload this image, maybe because that's my 101th one?)

[duane]

Is that normal to find a dry river in the jungle?

That's a case of a high-heat, low-humidity jungle. It's another balancing act between losing water in hot jungles or losing water in dry grasslands. Once I get the basics working, I'll try to adjust it for the extremes.

Needless to say, it'll be a lot easier to add it to the lua.

I also found another bug in the humidity handling, though it doesn't seem to have much obvious effect.

[duane]

Just when I think I've got the riverbeds working, I realize that I've made the jungle tree problem even worse.



Back to the drawing board.

[Gael de Sailly]

I've made a beautiful screenshot but I can't upload it on imgur and on minetest forums :-/

Just when I think I've got the riverbeds working, I realize that I've made the jungle tree problem even worse.

Back to the drawing board.

:-(

I think I've found the explanation for the incoherent biome repartition around rivers (where there is a dry savanna next to the river and a grassland higher). The current biome system uses Voronoi diagrams to define biomes. I've made a diagram with GeoGebra.

Spoiler

Vertical axis: humidity ; Horizontal axis: temperature

That's not due to humidity, but to temperature. The temperature is decreased (leftward move), so it may leave the "savanna" zone and enter the grassland or forest zone, without humidity change. So that's these biomes that are incoherent.

[paramat]

Thanks for the diagram, i added a background making the best attempt at keeping all biome names readable.
The biome system being a voronoi diagram makes it very difficult to tune and balance, months of work and many problems, for example rainforest can meet coniferous forest at extreme humidity.

[duane]

I think I've found the explanation for the incoherent biome repartition around rivers (where there is a dry savanna next to the river and a grassland higher)...
That's not due to humidity, but to temperature. The temperature is decreased (leftward move), so it may leave the "savanna" zone and enter the grassland or forest zone, without humidity change. So that's these biomes that are incoherent.

You're right, of course. The biomes would probably work better without any extra temperature/humidity changes, using stricter altitude limits.

However, the problem in my screenshot turns out to be the decrease in altitude in the riverbed raising the humidity. So, at least that was easy to fix.

[Gael de Sailly]

The biomes would probably work better without any extra temperature/humidity changes, using stricter altitude limits.

I don't think so. I think we should modify the biomes. I think the savanna biome is way too high in humidity.

@paramat, if you're interested by the diagram, here is the source file, (geogebra file, rename it to biomes.ggb).

Biomes.zip

(7.64 KiB) Downloaded 42 times

[Fixer]

So, at least that was easy to fix.

Is it possible to encounter same savannah but more hot one, so river will be without jungles? or it is always jungles in savannah river?

[duane]

This is kind of interesting. Now the jungle is surrounding the savanna.

Code: Select all

diff --git a/mods/default/mapgen.lua b/mods/default/mapgen.lua
index c92514c..fe50cbb 100644
--- a/mods/default/mapgen.lua
+++ b/mods/default/mapgen.lua
@@ -672,7 +672,7 @@ function default.register_biomes()
 		y_min = 5,
 		y_max = 31000,
 		heat_point = 85,
-		humidity_point = 20,
+		humidity_point = 10,
 	})
 
 	minetest.register_biome({
@@ -690,7 +690,7 @@ function default.register_biomes()
 		y_min = -112,
 		y_max = 4,
 		heat_point = 85,
-		humidity_point = 20,
+		humidity_point = 10,
 	})
 
 
@@ -708,8 +708,8 @@ function default.register_biomes()
 		--node_river_water = "",
 		y_min = 1,
 		y_max = 31000,
-		heat_point = 85,
-		humidity_point = 50,
+		heat_point = 65,
+		humidity_point = 30,
 	})
 
 	minetest.register_biome({
@@ -726,8 +726,8 @@ function default.register_biomes()
 		--node_river_water = "",
 		y_min = -3,
 		y_max = 0,
-		heat_point = 85,
-		humidity_point = 50,
+		heat_point = 65,
+		humidity_point = 30,
 	})
 
 	minetest.register_biome({
@@ -744,8 +744,8 @@ function default.register_biomes()
 		--node_river_water = "",
 		y_min = -112,
 		y_max = -4,
-		heat_point = 85,
-		humidity_point = 50,
+		heat_point = 65,
+		humidity_point = 30,
 	})
 
 

[duane]

Is it possible to encounter same savannah but more hot one, so river will be without jungles? or it is always jungles in savannah river?

Well, check the diagrams above. If you have hot + humid, it's always going to be jungle. Savannah, desert, and jungle share the same heat range.

[Gael de Sailly]

The same diagram with your changes:

Spoiler

Biomes2.png (69.71 KiB) Viewed 657 times

It looks strange at first but I think it could be better than the current biomes.

[sofar]

many problems, for example rainforest can meet coniferous forest at extreme humidity.

This is entirely valid, and happens in many locations in real-life. For instance, not more than 80km away from me, there is a coastal rainforest that is directly backed by conifers.

[Gael de Sailly]

many problems, for example rainforest can meet coniferous forest at extreme humidity.

This is entirely valid, and happens in many locations in real-life. For instance, not more than 80km away from me, there is a coastal rainforest that is directly backed by conifers.

The tropical conifer forest exists, but that's not the same conifers than on tempered forests.
Anyway I think that's not a major problem for MT's biomes.

[duane]

I like having a green border around jungles, rather than going straight to savanna. Of course, ethereal had this long ago.



It's getting complicated.

[prof-turbo]

This looks amazing °.°

[Gael de Sailly]

[…] the altitude chill code decreases temperature logarithmically (whereas, reality decreases it linearly).

The Lua VMG uses only positive values. That's not the case for the C++ VMG. Temperature is from -20 to 120, so the exponential decreasing of the temperature (valleys_mapgen.cpp:621-623) is inverted on very cold places. It should be linear.

[duane]

[…] the altitude chill code decreases temperature logarithmically (whereas, reality decreases it linearly).

The Lua VMG uses only positive values. That's not the case for the C++ VMG. Temperature is from -20 to 120, so the exponential decreasing of the temperature (valleys_mapgen.cpp:621-623) is inverted on very cold places. It should be linear.

Well, how about this?

Code: Select all

noise_heat->result[index_2d] -= 20.f * t_alt / altitude_chill;

The drawback is that biome temperature fluctuations are much more prominent, causing the snow line to be more ragged. It doesn't look as pretty to me. On the plus side, it's a bit faster to generate.



I don't think many mods are going to be too particular about what happens below zero temperature, so it probably won't be a major issue.

[Gael de Sailly]

I've tried to change the altitude chill code with what you have proposed, and I've not seen these strange snow lines, even with the same world seed (from your dry_rivers branch):


But I've found another problem that only occurs with dry_rivers: jungle trees growing in the river where the river meets the sea, at y=0 only.

I've uploaded the images on GitHub by dragging-dropping on a comment field. I don't know whether they will be deleted since they're not used in GitHub.

[duane]

I've tried to change the altitude chill code with what you have proposed, and I've not seen these strange snow lines, even with the same world seed (from your dry_rivers branch)

Well, I probably made a mistake at some point. It's not a major issue for me. Both methods give similar results within the area most people will be focused on. If you and paramat like this method, I'll try to put it into the dry_rivers patch. This is what I'm testing at the moment, though I haven't had much time to focus on it yet:

Code: Select all

diff --git a/src/mapgen_valleys.cpp b/src/mapgen_valleys.cpp
index 94abed3..77a9785 100644
--- a/src/mapgen_valleys.cpp
+++ b/src/mapgen_valleys.cpp
@@ -597,7 +599,8 @@ int MapgenValleys::generateTerrain()
 			// Derive heat from (base) altitude. This will be most correct
 			// at rivers, since other surface heights may vary below.
 			if (use_altitude_chill && (surface_y > 0 || river_y > 0))
-				t_heat *= pow(0.5f, MYMAX(surface_y, river_y) / altitude_chill);
+				//t_heat *= pow(0.5f, MYMAX(surface_y, river_y) / altitude_chill);
+				t_heat -= 20.f * MYMAX(surface_y, river_y) / altitude_chill;
 
 			// If humidity is low or heat is high, lower the water table.
 			float delta = noise_humidity->result[index_2d] - 50.f;
@@ -661,7 +664,8 @@ int MapgenValleys::generateTerrain()
 			// This is similar to the lua version's seawater adjustment,
 			// but doesn't increase the base humidity.
 			if (t_alt > 0)
-				humid *= pow(0.5f, t_alt / altitude_chill);
+				//humid *= pow(0.5f, t_alt / altitude_chill);
+				humid -= 20.f * t_alt / altitude_chill;
 			noise_humidity->result[index_2d] = humid;
 		}
 
@@ -674,8 +678,8 @@ int MapgenValleys::generateTerrain()
 				// The altitude hasn't changed. Use the first result.
 				noise_heat->result[index_2d] = t_heat;
 			else if (t_alt > 0)
-				noise_heat->result[index_2d] *=
-					pow(0.5f, t_alt / altitude_chill);
+				//noise_heat->result[index_2d] *= pow(0.5f, t_alt / altitude_chill);
+				noise_heat->result[index_2d] -= 20.f * t_alt / altitude_chill;
 		}
 	}
 

I think I'll have to tone down the humidity change, since I'm seeing more cold desert than I'd like, even when it's set down to zero humidity. One good thing I've noticed is that the biome noises are more prominent, causing the trees clump together better, rather than trying to follow rivers.

[Gael de Sailly]

I would like to keep the logarithmic humidity increase near the rivers. The influence of the rivers on humidity is important when you are close to the river, and becomes quickly insignificant when climbing up, contrary to the altitude chill that is linear. But I have not tested your code. If you think that's better, you can commit it.

[duane]

I would like to keep the logarithmic humidity increase near the rivers. The influence of the rivers on humidity is important when you are close to the river, and becomes quickly insignificant when climbing up, contrary to the altitude chill that is linear. But I have not tested your code. If you think that's better, you can commit it.

I agree. I mean to keep the logarithmic function near rivers. The humidity change there is, of course, for the altitude function that (sort of) takes the place of the sea-level humidity and soil thickness functions in the lua. It's helpful, but a bit too strong as I posted it.

[duane]

In case anyone missed it, the mainline minetest code now features dry riverbeds where the humidity is low. We also managed to correct some technical problems that may have been causing caves and dungeons to spawn in weird places, made swamps look less sandy and more swampy, removed river water from the seas, figured out a more robust way of varying humidity and temperature by altitude, and made the heightmap more accurate for mods.

All of these changes should appear in the upcoming minetest version.

[Don]

Nice work!