Hypothesis 4: Species interactions lead to the presence of non-random combinations of traits, hence defining the properties of the island trait spaces
We compared the observed metrics for each island’s trait space with the values obtained from a random sample of the species from the Canary Islands’ subterranean species pool. We hypothesised finding larger differences between observed and random values on mature islands, where community and evolutionary processes have optimised the use of the trait space via the establishment of stronger species interactions (Wilson 1969; Borregaard 2016; 2017). Therefore, we expected that functional richness and evenness would resemble the value expected by chance in young islands, as these offer more ecological opportunities for speciation and new species will evolve occupying all the vacant niche space. However, as the island reaches its maximum topological complexity and the species richness gets to its equilibrium, we expected functional richness to decrease and functional evenness to increase, reflecting how the establishment of more interactions among species at the community level will promote a reorganisation the use of the island’s trait space in a non-random manner. In senescent islands where extinction dominates the island dynamics, we expected the strength of species interactions to relax, leading to values of functional richness and evenness again similar to those expected by chance alone.
We applied null modelling to test whether the hypervolume metrics for each island were different than expected from a random sample of species from the Canarian subterranean species pool. We expressed the observed value as the actual functional richness and evenness of the hypervolume for each island. Then, we repeatedly (999 times) randomly subsampledn species from the species pool, where n equalled the total species richness in each island, and extracted the same two functional metrics. For each permutation, we estimated standard effect sizes and associated p-values using the BAT function ses to assess the significance of the deviation from the null expectations.