Predicting rarity and decline in animals, plants, and mushrooms based on species attributes and indicator groups



C. J. M. Musters, Institute of Environmental Sciences, Leiden University, P.O. Box 9518, 2300 RA Leiden, The Netherlands. Tel: +31 71 527 5618; Fax: +31 71 527 7434; E-mail:


In decisions on nature conservation measures, we depend largely on knowledge of the relationship between threats and environmental factors for a very limited number of species groups, with relevant environmental factors often being deduced from the relationship between threat and species traits. But can relationships between traits and levels of threats be identified across species from completely different taxonomic groups; and how accurately do well-known taxonomic groups indicate levels of threat in other species groups? To answer these questions, we first made a list of 152 species attributes of morphological and demographic traits and habitat requirements. Based on these attributes we then grew random forests of decision trees for 1183 species in the 18 different taxonomic groups for which we had Red Lists available in the Netherlands, using these to classify animals, plants, and mushrooms according to their rarity and decline. Finally, we grew random forests for four species groups often used as indicator groups to study how well the relationship between attribute and decline within these groups reflected that relationship within the larger taxonomic group to which these groups belong. Correct classification of rarity based on all attributes was as high as 88% in animals, 85% in plants, and 94% in mushrooms and correct classification of decline was 78% in animals, 69% in plants, and 70% in mushrooms. Vertebrates indicated decline in all animals well, as did birds for all vertebrates and vascular plants for all plants. However, butterflies poorly indicated decline in all insects. Random forests are a useful tool to relate rarity and decline to species attributes thereby making it possible to generalize rarity and decline to a wider set of species groups. Random forests can be used to estimate the level of threat to complete faunas and floras of countries or regions. In regions like the Netherlands, conservation policy based on attributes known to be relevant for the decline to birds, vertebrates or plants will probably also impact all aboveground terrestrial and freshwater macrofauna or macrophytes.