LOD or mesh combining?

In a race game, where the track is made of modules, is it better to apply LODs or combine all the meshes of the track in one?
Well of course the answer is “it depends”, but generally speaking which tecnique has the best results? And which tools / indicators are best to evaluate which solution is the best?

Thanks.

I see it’s been 5 years you’ve been wondering about it, but I would go for LOD in this case as combining the meshes would force GPU to constantly render the whole track even if you are facing only a fragment of it - which would result in a constant and unnecessary waste of resources.