Question: Understanding functional change in vegetation and how it might impact on vegetation change and ecosystem function entails measuring plant traits and attributes in situ. However, it is not realistically possible to sample every species for every trait, hence we need to address the consequences of not sampling all species.

Location: Nine arable and grassland sites from the machair of the Western Isles, UK and twenty grassland and woodland sites from Nynäs Nature Reserve, southern Sweden.

Methods: The effects of progressively reducing the proportion of species used to estimate the weighted mean of a range of continuous and qualitative traits were assessed.

Results: Relative abundance and species traits were related, and hence there is a cost in accuracy in reducing the number of species sampled in estimating the weighted mean for the vegetation. This cost was higher for qualitative traits than quantitative ones.

Conclusion: The analysis suggested that for the quantitative traits a minimum of those species that make up at least 80% of the vegetation should be sampled if traits do not vary greatly between species. If this variability is high and the trait is likely to be correlated to abundance then greater effort in sampling species for traits is required. Qualitative information on the rarer species should still be entered into the analysis if it can be accessed in any way, even if quantitative information for that species is unavailable.

Abbreviations: RMSE = Root mean square error; SLA = Specific leaf area