LandFrag : a dataset to investigate the effects of forest loss and fragmentation on biodiversity

Abstract

Motivation: The accelerated and widespread conversion of once continuous ecosystems into fragmented landscapes has drivenecological research to understand the response of biodiversity to local (fragment size) and landscape (forest cover and fragmen-tation) changes. This information has important theoretical and applied implications, but is still far from complete. We compiledthe most comprehensive and updated database to investigate how these local and landscape changes determine species composi-tion, abundance and trait diversity of multiple taxonomic groups in forest fragments across the globe.Main Types of Variables Contained: We gathered data for 1472 forest fragments, providing information on the abundanceand composition of 9154 species belonging to vertebrates, invertebrates, and plants. For 2703 of these species, we obtained more than 20 functional traits. We provided the spatial location and size of each fragment and metrics of landscape composition andconfiguration.Spatial Location and Grain: The dataset includes 1472 forest fragments sampled in 121 studies from all continents exceptAntarctica. Most datasets (77%) are from tropical regions, 17% are from temperate regions, and 6% are from subtropical regions.Species abundance and composition were collected at the plot or fragment scale, whereas the landscape metrics were extractedwith buffer size ranging from a radius of 200–2000 m.Time Period and Grain: Data on the abundance of species and community composition were collected between 1994 and 2022,and the landscape metrics were extracted from the same year that a given study collected the abundance and composition data.Major Taxa and Level of Measurement: The studied organisms included invertebrates (Arachnida, Insecta and Gastropoda;41% of the datasets), vertebrates (Amphibia, Squamata, Aves and Mammalia; 44%), and vascular plants (19%), and the lowest levelof identification was species or morphospecies. Software Format: The dataset and code can be downloaded on Zenodo or GitHub.