PS 22-222
Which and how many functional traits should be selected to measure the functional diversity of plant communities?

Monday, August 10, 2015
Exhibit Hall, Baltimore Convention Center
Linhai Zhu, State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, China
Fu Bojie, Research Center for Eco-Environmental Science, Chinese Academy of Sciences, Beijing, China

Functional diversity index summarizing the trait information of a community into a measure, is essential to understand the relationships among biodiversity, ecosystem functioning and environmental constraints. Although trait choice is crucial to quantify functional diversity appropriately, the methods of trait choice are rarely discussed. Meanwhile, very little is known about how the trait choice might affect conclusions about ecological processes drawn from functional diversity.

We will present four methods of trait selection as alternatives to the ordination axis-based method, which is mainly adopted at present and use ordination axes as “traits” to measure functional diversity. The proposed methods are based on RM coefficient, Yanai’s Generalized Coefficient of Determination and RV coefficient, as well as the highest loads in each ordination axis. Contrary to the ordination axis-based method, these approaches identify a subset of key traits to represent the total variation of all the measured traits. To evaluate their performance, six data sets from different ecosystems were used and species richness is predicted by four functional diversity indices (FAD, FD, Q, FDis), which were calculated by different methods. The evaluation was also benchmarked against the best combinations of traits determined by calculated functional diversity indices using all the combinations of traits (the complete search).


In the twenty-four analyses (six data sets and four functional diveristy indices), the four methods of trait selection were comparable to the ordination axis-based method. Only in the six analyses, the predictive power of these two types of methods was close to these of the best combinations of traits determined by the complete search, which were also totally different from the combinations of the four trait selection methods. The best trait numbers of the complete search ranged from one to seven, whereas these of the trait selection methods were four or two based on the intrinsic dimensionality of traits. Additionally, a high incidence existed that the low dimensional function diversity indices outperformed those high dimensional ones. Therefore, trait identity might be more important than trait number in the measurement of functional diversity. Without appropriate trait selection, a spurious ecological conclusion might be drawn via a biased assessment of functional diversity. We advocated computing all the possible functional diversity indices using different numbers and identities of traits to assess the rationale of different trait choice methods.