To create a matrix of species presences on islands and island groups, we used raw data from GloNAF version 1.1 and raw data from PIER (updated 2^nd of June 2018). Both PIER and GloNAF list their sources for all records of a naturalized species on an island.

We harmonized species names using the LCVP (Freiberg et al. 2020) and the associated R-package ‘lcvplants’ (https://github.com/idiv-biodiversity/LCVP). Subspecies and varieties were aggregated to the binomial level, which we refer to as “species” level for simplicity. If a hybrid taxon was not recorded in the LCVP database, it was pooled with the first parent species, affecting 20 species. Forty species names were identified by the LCVP as synonyms for more than one possible species. In these cases, we chose the first species name provided by the LCVP as assigned name to assure reproducibility. Twenty species could not be linked to an accepted name by the LCVP with certainty and were assigned ‘NA’ values for species name, family, and order. We kept these unassignable species in the list format of PacIFlora to allow for future name resolution.

Islands often have many names or different spellings. We used web research and other information in our sources to identify synonyms, and chose one name from the sources (usually the most commonly used name). Island refers to the smallest available unit of reference and is therefore sometimes used for several unnamed islets of an atoll in close proximity which are not or only sometimes (tidally) connected by land. 488 islands were recorded. IslandGroup is a group name assigned by us. The inclusion of island groupings is useful for many types of research questions, as a complete species list at a broader spatial scale decreases problems of data deficiency for individual islands in an archipelago. Our groups are largely based on political borders, such as municipalities or states. If political borders did not reflect geographic borders, we used the distance between islands and ocean trenches to assign each island to one of 50 island groups. We excluded 146 records that could not be linked to any island group. This particular grouping is useful for questions related to the influence of dispersal barriers on biological invasions, as distance creates a natural barrier and political borders are known to influence dispersal via human imports (either intentional or accidental). However, islands also vary in age, size and geomorphology, and thus we make it possible to regroup the islands in our database into formats that might be better suited for other research questions (e.g. on establishment barriers).