A weighting method to improve habitat association analysis: tested on British carabids

Analysis of species’ habitat associations is important for biodiversity conservation and spatial ecology. The original phi coefficient of association is a simple method that gives both positive and negative associations of individual species with habitats. The method originates in assessing the association of plant species with habitats, sampled by quadrats. Using this method for mobile animals creates problems as records often have imprecise locations, and would require either using only records related to a single habitat or arbitrarily choosing a single habitat to assign.

We propose and test a new weighted version of the index that retains more records, which improves association estimates and allows assessment of more species. It weights habitats that lie within the area covered by the species record with their certainty level, in our case study, the proportion of the grid cell covered by that habitat.

We used carabid beetle data from the National Biodiversity Network atlas and CEH Land Cover Map 2015 across Great Britain to compare the original method with the weighted version. We used presence‐only data, assigning species absences using a threshold based on the number of other species found at a location, and conducted a sensitivity analysis of this threshold. Qualitative descriptions of habitat associations were used as independent validation data.

The weighted index allowed the analysis of 52 additional species (19% more) and gave results with as few as 50 records. For the species we could analyse using both indices, the weighted index explained 70% of the qualitative validation data compared to 68% for the original, indicating no accuracy loss.

The weighted phi coefficient of association provides an improved method for habitat analysis giving information on preferred and avoided habitats for mobile species that have limited records, and can be used in modelling and analysis that directs conservation policy and practice.