A probability-based reanalysis of a classic dataset from biogeography and community ecology

Background/Question/Methods

Since the very first analysis in the 1970’s, Mayr and Diamond’s dataset on the geographical distribution of birds in the Bismarck Archipelago and Solomon Islands has had a profound impact on the development of concepts and theory in biogeography, macroecology, and community ecology.  The early papers by Jared Diamond have influenced thought on competition, community assembly, equilibrium versus non-equilibrium organization, and colonization/extinction dynamics.  Diamond’s papers led directly to the development of null models as statistical tools to test for non-random pattern in ecological data.  Much debate quickly ensued on the relative importance of competition and proper use of null models.  Recently, the debate has flared again as two new and opposing analyses of the dataset have been published.  However, both of these analyses retain the traditional approach of testing for non-random pattern (i.e., competition) at the level of an entire species presence-absence matrix instead of by explicitly analyzing species pairs.  The question of how well the data represent widespread pairwise competition remains unanswered.  I used the probabilistic model of species co-occurrence to test for negative, positive, and random associations among the 150 species of the Bismarck Archipelago and the 141 species of the Solomon Islands. 

Results/Conclusions

In the Bismarck Archipelago, only 3% of the 5,515 species pairs represented negative associations whereas positive and random associations represented 52% and 45%, respectively.  Similarly, in the Solomon Islands, only 2.2% of 4,490 pairs represented negative associations, 79% and 23.8% represented positive and negative associations respectively.  However, further statistical testing revealed that the negative associations tend to involve species with the same habitat preference more often than expected – indicative of pairwise competition.  In each set of islands, there was a group of 6 – 10 species that were involved in many of the negative associations.  Many of these were the so-called supertramp species identified by Diamond as having good dispersal but poor competitive ability.  These species and the fact that many of the negatively-associated species represented different taxonomic families lends support to Diamond’s conclusion of diffuse competition.  Therefore, although this classic dataset does not provide evidence of widespread pairwise competition, the geographic distribution of some species pairs is consistent with one species exerting a competition exclusion effect on the other.  Overall, the results suggest that positive associations due to shared habitat preference may be a pervading mechanism affecting species co-occurrence and community composition.