The data comprehends the list of bird species that occur on 3,713 across the globe. The data was compiled from two data sources:
1) BirdLife International and HBW (2018) (BirdLife hereafter). This dataset consists of expert range maps (polygons of geographical distributions) of all bird species globally (11,125). Every species has information related to its presence (e.g., extant, probably extant, possibly extant, possibly extinct, extinct), origin (e.g., native, reintroduced, introduced, vagrant, origin uncertain, assisted colonization), and seasonality (e.g., resident, breeding season, non-breeding season, passage, seasonal occurrence, uncertain). 2) eBird (2020). This database consists of occurrence points where birds were observed. This database documents the distribution, abundance, habitat use, and trends of birds through species lists collected by birders. These observations were collected on a global scale, and experts reviewed unusual records. For our focal islands, we obtained a total of 54,731,644 records with more than 947,862,800 observations).
To compile a birds-on-islands dataset, we separately intersected the expert range maps (BirdLife) and the occurrence points (eBird) of each bird species with the island polygons. Doing so resulted in two occurrence matrices with information on the presence and absence of each bird species on each island. For the Birdlife data, we kept all attributes that describe the status of the species, resulting in a total of 7,289 island bird species on 17,771 islands. For the eBird data, we kept the occurrence information, and we found a total of 7,318 species on 4,030 islands. The database is very comprehensive in many locations but does not provide systematic global coverage, so it is known to underestimate species’ presence in some regions of the world.
To increase the likelihood that our bird dataset only contains actual species presences with the correct presence, origin and seasonality, i.e., to reduce the number of false presences and increase the accuracy of the distributions, we only kept species records from the BirdLife range maps which were also confirmed by species presence records from eBird and vice versa. We decide this, because the distribution maps created by Birdlife estimate the distribution of the species, extrapolating data according to the habitat of the identified species. Thus, Birdlife maps do not provide the actual presence of the species but its potential distribution. These expert range maps are extent-of-occurrence estimates, so are known to overestimate species’ presence. However, these maps provide ecological and seasonal information on birds that is not available in eBird.
To match the two occurrence matrices in this way, we intersected them using the species names and the island identity, allowing us to validate the presence of birds on islands. Here we obtain a resulting common dataset of 5,123 species on 3,713 islands.
