Endemicity biases nestedness metrics: a demonstration, explanation and solution

Nestedness is frequently investigated to understand complex patterns of species occurrences. Temperature (T) is commonly used for comparisons of nestedness of different-sized datasets. However, the assumptions made for the standardization of this metric have not been fully explored, particularly the effects of endemicity. Here we show that T incorrectly indicates an increase in nestedness with the addition of non-nested endemics to matrices that are not perfectly nested – a consequence of standardizing matrix size by the product of species and sites. This problem is common both to test matrices and to real matrices that are typically subjected to nestedness analyses. The latter are often characterized by substantial numbers of endemics and by variation in the numbers of endemics in different taxa. Standardizing by occupancy resolves this problem, which is demonstrated using a derivative of discrepancy, d1. A small modification to T, such that it standardizes matrices by occupancy, would resolve the current problems with this nestedness metric.