Professor Yale University New Haven, Connecticut, United States
Background/Question/Methods
The volume of species occurrence records in global biodiversity databases is rapidly growing across a breadth of taxa. This includes occurrences for many species that have alien and invasive ranges, often documented in global alien species databases such as GRIIS. While the assigning of alien status to occurrence records within recognized alien regions is relatively straightforward, few techniques exist to assess the biogeographic status of records outside of both a species’ native rangemap (which may contain spatial and taxonomic uncertainty) and documented alien regions. This is despite the disproportionate importance of those records as potential first signs of damaging future invasions. An automated system for the classification and detection of alien species occurrences is thus urgently needed. If implemented carefully, such a system could greatly support global invasion detection and monitoring efforts, and would contribute to national and global indicators of biological invasion.
Results/Conclusions
We present a system for the large-scale automated classification of alien species occurrence records. Our system integrates occurrences, rangemaps, and species-group taxonomies from the Map of Life with national-level alien species introduction records in GRIIS to generate predictions of occurrence record biogeographic status across a breadth of taxa. Predictions are based primarily on occurrence distance to native rangemap, but additionally consider the dispersal ability of a species as informed by taxon-specific trait information. We introduce geohash-based techniques for the highly-efficient computation of occurrences’ distance to rangemap, enabling the system to classify hundreds of thousands to millions of occurrence records. Since occurrence records in large biodiversity databases are often of variable quality, and since our workflow is particularly designed to work with spatial outliers, we present methods for spatially quantifying occurrence data reliability for the accurate identification of true alien occurrence records. We aim to make the alien occurrence classification system public in the future, as a module within Map of Life’s biodiversity informatics platform.
