Analysis of biodiversity experiments: A comparison of traditional and linear-model-based methods

作者：Fibich Pavel^*; Rychtecka Terezie; Leps Jan

来源：Acta Oecologica, 2015, 63: 47-55.

DOI：10.1016/j.actao.2015.02.002

• Biodiversity effects
• Diversity-interaction models
• Species identity
• Species interactions
• Sown density
• Species pool

The relationship between diversity and ecosystem functioning is often studied by biodiversity experiments. Traditionally, the mechanisms behind biodiversity effects observed in these experiments have been evaluated by relative yield, overyielding and Loreau and Hector's additive partitioning of net biodiversity effect, and more recently by linear-model-based methods such as Kirwan's diversity-interaction model and Bell's random partition model. We compared data required and results given by these traditional and linear-model-based methods using three data sets: a simulated data set and two pot biodiversity experiments. For the simulated data, we also compared expected outputs based on defined ecological species traits with actual outcomes of the methods. BEF experiments were designed to answer five fundamental questions: Q1) How does sown species richness change ecosystem functioning? Q2) What ecological mechanisms cause this change? Q3) How does the importance of ecological mechanisms change with sown species richness? Q4) Which species are responsible for given ecological mechanisms? Q5) How do other experimental treatments change answers to all questions above?
We show that all methods were capable of answering Q1 and to some extent also Q5 although different methods use different procedures to reach the answer. Concerning Q2 Q4, we found that traditional methods provide more detailed insight into the ecological mechanisms than the linearmodel-based methods which leave us just with brief information. A direct comparison between traditional biodiversity effects and effects from linear-model-based methods showed that species interactions from the diversity interaction model were significantly positively correlated with the net effect, while species identity effects from diversity interaction model were related to the species relative yield. The selection of an appropriate method for BEF experiment analysis thus depends on the questions we ask which in turn also affect the design of the BEF experiment.

• 出版日期2015-2