Previously it was using this 'peak_height' which is the total height of the model in scale_vector.y. However it was then multiplying the height map with that scale vector and adding the base height again a second time. So the scaling part was too thick and included the base height, and the total height of the mesh was also too big.
I also reduced an unnecessary re-calculation of the height_from_base parameter. And as a result we don't need the peak_height variable at all any more.
Fixes#8902.
d53568c5db