In my previous blog post I described a simple approach for handling snow accumulation in an erosion simulator. In this post I will present new improved results, as well as a mechanism for reducing snow accumulation in areas with significant amounts of fluid. Now it looks like this:

The most obvious difference versus previous results is that the presence of water now reduces the amount of snow that can accumulate. Slightly less obvious is that the erosion simulation is now working properly after fixing a bug I introduced when adding snow. This bug reduced the fluid precipitation by a factor of 10, which is the cause of the underwhelming landforms produced in the previous results.

To implement the melting of snow by water, I consider the current fluid depth at each position when determining how much snow should fall. For fluid depths between 0.2 and 2.0 meters, I linearly reduce the amount of snow with increasing depth, such that beyond 2.0 meters depth, the snow amount is 0. This produces a nice effect which emulates how snow can be melted (or at least displaced) by flowing water. It also is a cheap way to simulate erosion of snow by water, without having to include the snow in the rest of the erosion simulator. Currently snow is just a single layer applied on top of the rock+sediment layers.

I also made some improvements to the tone mapper and automatic exposure control to make the images brighter by 1 stop and snow less gray. These results also use smoother interpolation between the per-vertex texturing, since the noise-based interpolation used previously is not working correctly at all scales and can look pixelated.