I question this because I have seen publications with descriptions based in less than two relevés and I think its uncomparable and very heterotonous (specially in grassland and scrubland vegetation series). Even if the relevés are small or big in size, you cannot compare with other geographic areas in which the same diagnostic spp may occur, it generates confussion for further investigation.
Less than 2 relevés is 1 relevé. It happened me once in very complex terrain and difficulties to return back to the same vegetation type.
But even 1 relevé is valid, but as a rule not applicable under normal conditions.The variation of that particular vegetation type will be missed.
I think under normal circumstances you should split your relevee in smaller plots to have some variance and be able to work it using at least basic statistics. If you have only 1 plot, how can you estimate uncertainty when interpretating the data (see, for example http://www.sciencedirect.com/science/article/pii/S0304377014000655)?
I think a single plot gives you only a set of "fixed" data that can not be statistically interpreted unless if is integrated in biogeographical analyses summing other plot in the landscape of biogeographical scale.
You can validly publish a association with 1 relevée, but I think it will cause much confusion in future research. As Antoine already stated, the variation will be missed. There is a more severe problem though, and it has to do with this variation, but also with the random distribution of species. In order to have a good idea of the species composition of a certain vegetation TYPE, you have to investigate what the total species composition is IN THAT TYPE. I for myself think of it as the constant species in a community under investigation, and it depends on the processes whether I take a 40% or a 60% threshold for constant species. In fact you have to filter out the species that are not normally growing in the community, the accompanying species with low frequencies. Although there can be many of such species, the are not the core of the association, but a kind of TYPOLOGICAL biass (NOTE: they can be of ecological importance nonetheless, of course!).
It's obvious that you get a better picture of the constant species if you base your description on more relevées. In two relevées you can not decide what is 'normal' for your association, and what are the accidental species (in a typological sense). The more, the better, I would say, although you can find an optimum for each vegetation type I guess, where every new relevée hardly adds any new information....
Hope this helps?
PS: if you want to talk a little on this topic, please feel free to come by for a cup of coffee, I'm in the Gaia-building....:D
Well, I quite disagree with the idea of filtering out rare species. In tropical ecosystems they are a normal component of any vegetation. Almost any tropical ecosystem you sample will be full of rare species and have a small set of "abundant species". This is a reality in field work and should be reported accordingly. If you filter rare species out, you will be changing real data for an idealized approach. I prefer presenting phytosociological tables where abundant species appear on the top of the table and rare species are ranked bellow but, nonetheless are also fully reported (as I have done here: http://www.sciencedirect.com/science/article/pii/S0304377014000655). Doing this you give the reader the idea of which are the abundant and which are the rare, but you provide them with your full set of data, so he can judge by himself the results (and even disagree with you and use a different approach using the same dataset).
I have applied once to have one single relevee. That was in Colombia in a remote summit area under control of the precursor of the FARC, a place difficult to reach, with only half an hour to perform a releve. But I was able to see there was repetition in the area of that particular vegetation type. At least for tropical areas I should say under difficult field conditions: yes one relevé suffices.
Thanks so much for all your answers. I also agree that rare species should remain in the dataset but they will not be used as characteristic species to describe a new unit. In the other hand, I think it will be depend on the type of ecosystem threated. For instance, in xerophytic, chasmophytic, scrubland and grassland ecosystems, several relevés are needed in order to have a feasible comparison on which characteristic species dominate, with only one or two relevés you cannot conclude a higher syntax unit approach, again heterogeneous. In tropical humid ecosystems, I consider that 1-2 relevés can be sufficient to describe which trees/shrubs are dominant and therefore can be usefull to describe a unit, also because the relevé size will be much higher. I agree with Antoine Cleef, Rense Haveman.