01.28
I’ve replaced the simple sinus function for latitude temperature with a bit more complex one :
sqrt(0.5*(1+sin(x-1.6)^5))
Go there and copy/paste the function to see it. Put 0 and 6.28 in the x min and max.
The result is that there is more temperate climate, tundra and desert bands are thiner as seen on this biome map.
I think it’s better for the gameplay. We have a bit of tundra, a bit of desert, some tropical zones, but the main part of the land is Tolkienish temperate climate.
I spent also a good amount of time breaking my teeth against river algorithms… I tried slope based simulation, A* and dijkstra path finding using slope or altitude as movement cost and finally rapidly-exploring random trees. None of them gave a good result… I’ll probably fall back to a more pragmatic, less simulationistic solution…
Have you seen James McNeil’s posts on using RRTs for river network creation in games? They probably won’t be of direct assistance – the river network is built prior to creating a height map – but the screenshots sure are pretty.
http://playtechs.blogspot.com/2008/11/rapidly-exploring-random-trees.html
http://playtechs.blogspot.com/2008/11/rrt-with-obstacles.html
Have you tried identifying drainage basins? Real rivers tend to stick near the center of their basins, sure (with tributaries coming down smaller indentations that you might call sub-basins), but meander and lay down silt. A river’s volume will always be the rainfall in its basin minus evaporation there, and its path within the basin will be more random than efficient.
Yeah I discovered RRT thanks to James’ post
As for drainage basins, I think any heightmap based method will fail because of the coarse resolution of the heightmap (compared to the river). I’d need a lower resolution heightmap but I don’t want to because I want the map generation to be as fast as possible (it takes currently less than 1 sec). That’s why I also avoid pure simulation technics and rather use hacks to reach something realistic enough