Triangulated irregular network (TIN) can produce terrain meshes with a reduced triangle count compared toregular grid. At the same time, TIN meshes are more challenging to optimize in real-time in comparison to otherapproaches. This paper explores efficient generation of view-dependent, adaptive TIN meshes for terrain duringruntime with no or minimal preprocessing. This is achieved by reducing the problem of mesh simplification to thatof inexpensive 2D Delaunay triangulation and lifting it back to 3D. The approach and its efficiency is validatedwith suitable datasets.