The strengths and weaknesses of species distribution models in biome delimitation. Issue 10 (7th July 2020)

• Record Type: Journal Article
• Authors: Moonlight, Peter W.; Silva de Miranda, Pedro Luiz; Cardoso, Domingos; Dexter, Kyle G.; Oliveira‐Filho, Ary T.; Pennington, R. Toby; Ramos, Gustavo; Särkinen, Tiina E.
• Editors: Guisan, Antoine
• Is Part Of: Global ecology & biogeography. Volume 29:Issue 10(2020)
• Page Start: 1770
• Page End: 1784
• Publication Date: 2020-07-07
• Subjects: biome delimitation -- cluster analysis -- diversity patterns -- macroecology -- savanna -- species distribution modelling; Ecology -- Periodicals; Biogeography -- Periodicals; Biodiversity -- Periodicals; Macroevolution -- Periodicals; 577
• DOI: 10.1111/geb.13149
• Languages: English
• ISSNs: 1466-822X

Abstract

Abstract: Aim: The aim was to test whether species distribution models (SDMs) can reproduce major macroecological patterns in a species‐rich, tropical region and provide recommendations for using SDMs in areas with sparse biotic inventory data. Location: North‐east Brazil, including Minas Gerais. Time period: Present. Major taxa studied: Flowering plants. Methods: Species composition estimates derived from stacked SDMs (s‐SDMs) were compared with data from 1, 506 inventories of 933 woody plant species from north‐east Brazil. Both datasets were used in hierarchical clustering analyses to delimit floristic units that correspond to biomes. The ability of s‐SDMs to predict the identity, functional composition and floristic composition of biomes was compared across geographical and environmental space. Results: The s‐SDMs and inventory data both resolved four major biomes that largely corresponded in terms of their distribution, floristics and function. The s‐SDMs proved excellent at identifying broad‐scale biomes and their function, but misassigned many individual sites in complex savanna–forest mosaics. Main conclusions: Our results show that s‐SDMs have a unique role to play in describing macroecological patterns in areas lacking inventory data and for poorly known taxa. s‐SDMs accurately predict floristic and functional macroecological patterns but struggle in areas where non‐climatic factors, such as fire or soil, play key roles in governing distributions.