Thanks, guys. I like the second one better too, but the first was a commission that needed to show certain archaeobotanic things, so this POV was the only one possible really.
The photo I took was from a piece of ground with quite flattened grass, moss and stuff, so nothing really higher than a few centimeters. Poor sandy heathland soil. From a tile of 2x2m I made a bumpmap in Pixplant, which took only 1cm of displacement in the image map shader in TG. I guess the greyscale of the leaves and raised parts also played a role in the realism. If you have dark leaves over light soil, it would have been far more complex. Which is why it's actually a bit of a cheat to make bumpmaps from photo's. You really need the photogrammic method.
Mind you, there's one object of heath in middle foreground, and one branch object. Rest is photo. The course snow remains are additional PF masked, with a bit of distort by normal, and a slope restraint.
The snow clumps are indeed 2 small fake stones, the prints voronoi billows 1 octave unclamped color followed by a smooth step, and then of course some minus displacement.
For the displacement intersection I used a 1.5m patch in the compute terrain and set intersection zone to 0.02 or 0.01, shift to -1 (I think, but maybe far less). The terrain patch decides the largeness of the snow remains. If set to 20, it would probably be all snow, or not at all. To make the snow softer, I used an initial terrain PF with 10cm as smallest size before compute, and later added an extra PF (after compute) with tiny detail, which would then be smoothed out in the snowy areas of course.
So, there you go.
