Methods for detecting non-randomness in species co-occurrences: a contribution

Area of Science:

• Ecology
• Community Ecology
• Biogeography

Background:

• Assessing species co-occurrence patterns is crucial for understanding community structure.
• Existing null models, like Gilpin and Diamond's, may produce misleading results, indicating structure in random data.

Purpose of the Study:

• To propose and validate an alternative statistical method for analyzing species co-occurrence.
• To re-evaluate island community structure using a more robust null model and randomization test.

Main Methods:

• Developed a novel randomization test using Monte Carlo simulations to determine the expected distribution of species co-occurrences.
• Employed a null model that accounts for observed island and species occurrence totals, offering a conservative test.
• Applied the method to the Vanuatu bird dataset previously analyzed by Gilpin and Diamond.

Main Results:

• The proposed randomization test identified significant departure from the null model in the Vanuatu bird data.
• Contrary to Gilpin and Diamond's findings, this study found an excess of extreme negative species associations.
• The results highlight the limitations of previous methods in accurately detecting community structure.

Conclusions:

• The developed randomization test provides a more reliable approach to analyzing species co-occurrence.
• Island community structure can exhibit patterns not detected by standard methods, including an excess of negative associations.
• Further research is needed to elucidate the ecological drivers (autecology, biogeography, interspecific interactions) behind observed negative associations.