Why do most vegetation monitoring programs rely on permanent plots?  If plots are randomly sampled over a homogeneous vegetation community, why not just establish new plots each time, rather than locating and re-sampling the same plots over and over again?