why are heightmaps subsampled?

edit

I'm using a terrain heightmap and I notice that there's a fixed 2x subsampling built into gazebo. The subsampling is done using simple linear interpolation. Why is this done? I mean, if I wanted to subsample I could do that myself too before loading the heightmap (an image in my case) into gazebo... If the subsampling could be controlled then maybe it would be a nice feature, but as-is it just quadruples the number of vertices for what gain?