Coming from UDK, the occlusion works automatically, where if it doesn’t see the mesh it won’t render it so they encourage you to build levels with many different parts. In Unity it seems they want you to keep as many objects as one mesh as you can that are close together. I’m not exactly sure yet how Unitys occlusion is differnt than UDK’s, but is it still correct with Unitys new occlusion to try and have models as single objects in Unity?
Merging meshes is only vaguely related to Occlusion + LOD, in that it optimises performance.
It reduces Draw Calls… more you have the slower things get, which is the same in any engine.
1000 objects of 1 tri - slooooww
1 object of 1000 tris - fast!
For good occlusion + LOD you need to keep meshes separate for them to do their magic. You wouldn’t merge individual rooms together or interiors with exteriors.
Occlusion Culling works really well together with static batching. With static batching, we can invidiually cull objects but we render all meshes in one static batch together to reduce draw calls. It’s magic and basically the best of both worlds. Low draw calls and but you still get very high culling efficiency from both occlusion culling and view frustum culling.
This doesn’t mean you should seperate meshes in pointless ways. Objects that are spatially very close should always be mesh combined if they can be and you should attempt to make them share materials too.
Sharing materials is necessary for static batching.