Co‐occurrence‐based measure of species' habitat specialization: robust, unbiased estimation in saturated communities

Aims Large vegetation plot databases enable the estimation of niche width from species co‐occurrence data. Different indices have been proposed for this purpose, but do not give unbiased (i.e. independent of species pool size) and robust estimates over a wide range of conditions. The aims of the paper are to: (1) demonstrate the limitations of different methods, and (2) propose a new algorithm that results in unbiased and robust estimates. Results Whittaker's β‐diversity, calculated from raw data, is an unbiased niche width measure only if the relationship between γ‐diversity and local richness is linear. However, this requirement is satisfied only in specific conditions, if both γ‐diversity and local richness are linear functions of the species pool size with zero intercept. I propose the use of Beals smoothing to estimate species pools. It has been proved through analysis of simulated data that Whittaker's β calculated from species pool data is an unbiased estimate of niche width. I have shown that the robustness of the estimate can be improved by excluding extremely species‐rich plots. The relative role of methodological decisions during niche width estimation was explored through analysis of a large field data set (>8000 relevés). Conclusions The proposed algorithm results in robust, unbiased estimation, even in saturated communities, thus it avoids the drawbacks of the co‐occurrence‐based niche width measures proposed earlier.