Anthropogenic and environmental drivers shape diversity of naturalized plants across the Pacific

The Pacific exhibits an exceptional number of naturalized plant species, but the drivers of this high diversity and the associated compositional patterns remain largely unknown. Here, we aim to (a) improve our understanding of introduction and establishment processes and (b) evaluate whether this information is sufficient to create scientific conservation tools, such as watchlists.


| INTRODUC TI ON
The Pacific ocean covers almost half of the earth's surface and contains many isolated islands with a rich and highly endemic flora that has become threatened by non-native plant species introductions (Loope, 1998;SPREP, 2016). Island floras in the Pacific region show a strong turnover in naturalized species among islands leading to the steepest regional species-area relationship in the world (van , meaning that the number of naturalized species sharply increases with the number of islands analysed. Unique features of the Pacific, such as its geographic isolation, biogeographic history and spatial variation in natural and anthropogenic factors, could explain the naturalized species distribution patterns for this region. Understanding patterns and drivers that shape naturalized plant richness and composition in the Pacific is necessary for a better global understanding of naturalization processes. Further, region-specific knowledge is a prerequisite for guiding conservation planning, such as compiling regional watchlists to mitigate future invasions (Sherley et al., 2000).
Previous studies have shown that most naturalized species occur in one or a few Pacific islands and that most islands have low naturalized species richness (Denslow et al., 2009;Meyer, 2004). Speciespoor islands tend to have a nested subset of the species present on species-rich islands (Traveset et al., 2014). Rarity indices that quantify the range size of species were developed for conservation research to identify areas with many endemic species (Usher, 1986;Williams, 1993). This same index can be used to quantify whether the naturalized species that are present on an island group are those with small or large ranges. If species-poor island groups have rare naturalized plant species, this could indicate strong environmental filtering for unique local conditions, whereas if they are inhabited by widespread naturalized plant species, this could indicate that dispersal limitation so far limits the arrival of rarer species due to stochastic factors. Quantifying patterns of rarity in combination with patterns of beta diversity will elucidate the drivers of naturalized species composition, with strong implications for conservation.
The success of naturalized plant species in the Pacific is often attributed to the weak competition ability of native plant species (e.g. Gillespie et al. 2008;MacDougall et al. 2009). The extreme isolation of many island groups in this region has led to phylogenetically clustered compositions of native species, whereby some native plant lineages are absent because of dispersal limitation and other lineages are over-represented due to high rates of in situ diversification and local extinctions (Cabral et al., 2019;Cavender-Bares et al., 2009;Crisp et al., 2009;König et al., 2020;. As a result, the Pacific region might have under-exploited resources (i.e. absence of species-trait combinationsDenslow, 2003;Elton, 1958;Gillespie et al., 2008)  For native plants, climate is known to strongly influence species richness (Kreft & Jetz, 2007;Wright, 1983), and both geographic distance and climate dissimilarity strongly influence dissimilarity in species composition (often measured as beta diversity) (Keil et al., 2012;König et al., 2017;Soininen et al., 2007).
Naturalized plant diversity patterns are shaped disproportionately by other drivers (Denslow et al., 2009;Moser et al., 2018;Qian et al., 2008;Winter et al., 2010), and globally, naturalized plant plant families that contributed many naturalized species globally also did so in the Pacific, particularly Fabaceae and Poaceae. However, many families were significantly over-or under-represented in the Pacific naturalized flora compared to other regions of the world. Naturalized species richness increased primarily with increased human activity and island altitude/area, whereas similarity between island groups in temperature along with richness differences was most important for beta diversity.

Main conclusions:
The distribution and richness of naturalized species can be explained by a small set of drivers. The Pacific region contains many naturalized plant species also naturalized in other regions in the world, but our results highlight key differences such as a stronger role of anthropogenic drivers in shaping diversity patterns. Our results establish a basis for predicting and preventing future naturalizations in a threatened biodiversity hotspot.

K E Y W O R D S
anthropogenic drivers, beta diversity, island biogeography, naturalized species, Pacific Ocean, plant invasion richness is strongly linked to human activity (e.g. represented by human population density and economic activity, Pyšek et al., 2017). Humans import and export species, purposefully and accidentally, through trade activities (Hui et al., 2017;Hulme, 2009;van Kleunen et al., 2018van Kleunen et al., , 2020 and directly or indirectly create disturbances that favour their establishment (Frenot et al., 2001;Hulme, 2009;Marvier et al., 2004;Merlin & Juvik, 1992). The broad range of biogeographic (area and isolation) and anthropogenic (e.g. human density and GDP) factors present in the Pacific is uniquely suited for studying naturalized species distributions, as it includes both highly invaded (Kueffer et al., 2010) and almost untouched islands.
To understand the climatic, biogeographic and anthropogenic factors underlying diversity patterns of naturalized plant species across the Pacific, we quantify (a) naturalized species richness per island group, average distribution ranges of naturalized species per island group and naturalized species sharing between island groups in the Pacific, (b) the rarity of naturalized species present on each island group, (c) the species richness of naturalized plant families in the Pacific in relation to worldwide total and naturalized species richness, and (d) the anthropogenic, environmental and biogeographic drivers of naturalized species richness and beta diversity across Pacific island groups. Combined, these analyses provide a description of current naturalized species richness and composition in the Pacific and give insights into the forces that shape these patterns, serving as guidelines for applied conservation.

| Compilation of naturalized plant species occurrence data
In our study, the Pacific is defined as all islands between 40°N and 40°S, excluding large landmasses such as Japan, New Zealand, Indonesia, the Philippines and Papua New-Guinea as well as small islands on the Pacific-American or Australian coast. Japanese and South East Asian Islands west of Bonin/ Palau are also excluded (  Moser et al. 2018;van Kleunen et al. 2015). The GloNAF database contains information using an island grouping based on political borders rather than the groupings we develop here (see below). We therefore used raw data at the (mostly) island level from GloNAF version 1.1 (van  and aggregated them at the level of the island groups defined in this study. Table S1 gives a list of all 481 islands with species records. Our data mostly contained information at island level, but we aggregated occurrences to island group level to ensure high completeness of the species lists analysed. We acknowledge that this reduces but does not eliminate the issue of data deficiency. To create island groups, we started with (a) political borders, including municipals or states. If political borders poorly reflected geographic borders (e.g. Samoa and American Samoa), we used (b) distance between islands and (c) ocean trenches to assign each island to one of 50 island groups. We excluded species x island records that could not be assigned to any island group with absolute certainty (e.g. when location information was imprecise). Our island groups reflect both geographic and political borders, because both are known to be important for non-native species movement. A list of all islands included in each island group is provided in Table S1.
Naturalized species were assigned accepted binomial species names using the Leipzig Catalogue of Vascular Plants (Freiberg et al., 2020) and the associated R package lcvplants (https://github. com/idiv-biodi versi ty/lcvpl ants). Different subspecies and varieties of the same species were merged to the species level. Hybrids were treated as separate species, except if no valid entry could be identified in which case they were treated as the first parent species.
Species records were removed when they could not be assigned to an accepted plant species (e.g. information only at plant family level).
In the rare case of more than one corresponding accepted species for a synonym, we selected the first output of the lcvplant function to ensure reproducibility.
Some species x island records in the databases were listed as cultivated (incl. for ornamental purposes) or native. For some species, naturalization status varies among species x island combinations (i.e. a species can be naturalized in one island group and cultivated or native in another), which we attribute to the large area of the Pacific, and the temporal nature of naturalization (i.e. a species may have escaped cultivation only in some island groups). When aggregating to island group level, we considered a species as naturalized for the island group if it has naturalized on any island within the group. When classifying the native status of species records, we averaged the expert evaluation in the sources for each island. We considered a species to be native if it is listed as native on any island within the group. Species that are not explicitly listed as naturalized in any record on the Pacific but listed as cultivated/ native at least once were excluded from our study. With these criteria, 2,672 naturalized species were included in our analyses which can be found in Table S2.
For most analyses, we evaluated our questions excluding records of species on island groups classified as native and cultivated. We show the results of the full data set (i.e. including those records) in the Figures S1-S3. Results of both approaches were comparable, indicating the robustness of our results. For beta diversity and shared species analyses, excluding records of island groups with species classified as native and cultivated would create false absences (i.e. type I errors) and may generate false negatives (i.e. type II errors). Thus, we present analyses using all data for these analyses. We did not categorize species as "invasive" or similar (sensu Blackburn et al. 2011;Richardson et al., 2000), because this classification is not available for all species in the study region.

| Patterns of naturalized species richness, range size, rarity and composition
We first quantify basic patterns of naturalized plant species distributions in the Pacific. Compositional species sharing between each pair of the 50 island groups was quantified in two ways: (a) the total number of shared species, "(b) beta Simpson pairwise dissimilarity , with a being the species two island groups have in common and b and c being the richness of the individual island groups; this metric considers species turnover without the influence of species richness differences (monotonic transformation as in Baselga (2010), using the betapart package in R (Baselga et al., 2020).
We used our naturalized species by island group matrix to calculate species richness per island group (n g ), species occupancy (c i , range size of species i) and compositional species sharing. Average (geographic) rarity per island group is defined as , that is the average over all species present in an island group of the inverse of their occupancy (Usher, 1986;Williams, 1993). We assessed whether there is a relationship between the average rarity of the naturalized species and the naturalized species richness of an island group using Pearson's product-moment correlation to assess whether species-poor island groups tend to have more common or more rare species, indicating a strong role of dispersal or environmental filtering.

| Naturalized plant family representation in the Pacific
We investigated whether the Pacific has a different naturalized plant family composition compared to other regions of the world.

| Naturalized species occurrences outside the Pacific
Using GloNAF, we quantified the number of alien species in the Pacific that are also naturalized in other continents or countries. For each continent or country, we then computed the proportion of the total number of naturalized species that are shared with the Pacific.
As spatial units, we used (biogeographical) continents as defined by the World Geographical Scheme for Recording Plant Distributions developed by the Biodiversity Information Standards (TDWG). We also used countries, as information exchange concerning potential non-native species exchange occurs primarily at the country level (e.g. Global Invasive Species Database; http://issg.org/datab ase/ refer ence/index.asp). Additionally, we quantified the proportion of naturalized plant species in the Pacific that are unique naturalizations to this region and compared this to other continents, to test whether there are proportionately more naturalized species exclusive the Pacific. To complete this picture, we used Pearson's correlation coefficient to test whether species that are widespread in the Pacific (occupy many island groups) are also widespread globally (occupy many countries).

| Drivers of naturalized species richness and beta diversity
We assessed the relative importance of anthropogenic, biogeographic and climatic drivers in explaining patterns of naturalized species richness and beta diversity across island groups in the Pacific. One of our main goals was to investigate which factors promote naturalized richness and compositional similarity between island groups. We used GIFT (Global Inventory of Floras and Traits, Weigelt et al., 2020), a global archive of regional plant checklists and floras including physical, geographic, bioclimatic and anthropogenic characteristics, which are computed based on the spatial polygons and summary statistics for each island group (islands with naturalized species records not yet included were added to GIFT for this purpose): land area (combined area of all islands belonging to the same island group), distance to mainland (closest distance from coast to coast) and land area in proximity (SLMP: standardized land mass in proximity, averaged among islands, that is partly reflecting the spatial arrangement of a group (Weigelt & Kreft, 2013)). Further, we extracted other variables from additional resources and assigned them to the island groups.
These included the following: airport capacity (number of airports x airlines visiting them as a proxy for their size; https://openf lights.org/), harbour capacity (number of harbours x their size in three classes (World Port Index, https://msi.nga.mil/Publi catio ns/WPI)), mean annual temperature and precipitation (Chelsea (Karger et al., 2017), elevation (maximum) (WorldClim 1.4 digital elevation data (United States Geological Survey, 2011)), human population number and density (Doxsey-Whitfield et al. 2015) and Human Footprint and Human Influence Index (WCS, 2005).
Finally, we extracted a GDP Index which estimates GDP based on night light emission (NOAA et al., 2010). A low proportion of values were missing in our data set (18% for GDP Index and Human Population, 4% for precipitation and temperature), and for these, we imputed missing values using a random forest algorithm that is trained on observed values to predict missing values, using the R package "missForest" (Stekhoven & Buehlmann, 2012). We imputed missing values with 100 trees in each forest and five variables randomly sampled at each split; the estimated out-of-bag error imputation error rate across all variables was low (Out-ofbag (OOB) error = 0.144).
We assessed which variables promote naturalized species richness on Pacific island groups. Because variable importance may be biased towards correlated variables (Strobl et al., 2007), we used a two-step variable selection procedure when fitting random forest models using the R package "VSURF" (Genuer et al., 2015). The first step of this procedure ranks variables by importance and eliminates the redundant ones, where the threshold value is an estimate of the standard deviations of redundant variables. The second step selects variables that are important for interpretation, which retains variables that are strongly associated with the response variable but allows for some degree of redundancy. We used 100 trees in each forest, 10,000 forests in each interpretation step. This selection process is visualized in Figure S6.
To assess which drivers promote beta diversity of Pacific island groups, we calculated the absolute difference of each environmental and anthropogenic variable between each pair of island groups, and the geographic distance of the island groups to each other (instead of distance to continent). We included richness difference between island groups as a predictor in the random forest model to test whether pairwise beta Simpson depends on how similar or different the pairwise island groups are in their species richness.
This factor must be interpreted carefully as it does not represent a biogeographical, anthropogenic or strictly environmental driver.
Rather, richness difference provides meaningful information on the possible pathways in which species disperse in the Pacific. As beta Simpson examines species turnover without the influence of species richness differences, it is meaningful to examine how species richness differences influence beta Simpson. If species with similar richness have similar beta Simpson values, this could indicate that species-poor island groups all have the same composition of widespread naturalized species. Alternatively, if species with different richness values have high beta Simpson values, this could indicate that species-poor island groups contain a nested subset of the naturalized species found on species-rich island groups. When running the random forest model without the factor richness difference included, results changed only to a minor degree ( Figure S7), showing the robustness of our results. To support the interpretation of our results, we included partial dependence plots of a standard random forest model in the Appendix S1, as there is no indicator of direction or shape of the drivers influence in VSURF. This function draws the regression, which represents the effect of our drivers on naturalized species richness ( Figure S8-a) or beta Simpson diversity ( Figure S8b, randomForest package in R, Liaw & Wiener, 2002). This allows us to consider trends in these regressions for a better interpretation of our results.

| Patterns of naturalized species richness, range size and composition
On average, there were 258 species per island group (median of 182, Figure 1a). The lowest number of naturalized species was recorded for Howland and Baker (6) and the highest number for the Hawaiian Islands (1,544; Table S3). Most naturalized species were present on one or a few island groups (Figure 1b). Of the 2,672 naturalized species found in the Pacific region, only 1,377 (52%) occurred on more than one island group and only 367 (14%) occurred on more than 10.
Island groups with many naturalized species also tended to share many naturalized species with other species-rich island groups in absolute terms (Figure 2a), but this relationship was not observed for beta Simpson diversity (Figure 2b) likely because naturalized species-rich island groups had proportionally more rare species ( Figure 3). Indeed, we found a positive association between the average rarity and the species richness of island groups in the Pacific (p < .001; Pearson's correlation estimate = 0.78; Figure 3).

| Naturalized species occurrence outside of the Pacific
Twenty-one per cent of naturalized species found in the Pacific were not recorded as naturalized elsewhere in the world ( Figure S4), but this is neither exceptionally high nor low (within the standard deviation). Many naturalized species in the Pacific were also found as naturalized species in South America and Africa, and in the large countries bordering the Pacific including the United States, Australia and Mexico (Table 1). The top eight countries which shared the most naturalized species with the Pacific are shown in Table 1, and these countries serve as indicators for their larger regions (i.e. Portugal for the Mediterranean, Puerto Rico for the Caribbean). Naturalized species that were present as naturalized species in many countries outside the Pacific were also widespread within the Pacific ( Figure S5).

| Drivers of naturalized species richness and beta diversity
Human footprint and airport capacity were the most important variables explaining naturalized species richness of island groups, closely followed by other drivers that also reflect human influence and economic activity (Figure 5a; R 2 = .88). The difference between island groups in naturalized species richness and temperature were the most important variables explaining beta Simpson across island groups (Figure 5b, R 2 = .95). The partial dependence plots we included to foster interpretation ( Figure S8) do not provide a slope, but we observe clear trends. We can observe a generally positive trend for all drivers influencing richness (e.g. the higher the human footprint index and the land area of an island group, the higher the naturalized species richness). We observe a generally positive trend for differences in temperature and in geographic distance influencing F I G U R E 1 Histograms of the frequency of (a) naturalized species richness for all island groups and (b) naturalized species range (number of island groups each naturalized species is present on) for all naturalized species

| D ISCUSS I ON
In our comprehensive analysis of the naturalized flora of the Pacific region, we find 2,672 naturalized plant species from 228 different plant families. We show that the richness of naturalized plant species on island groups in the Pacific is strongly linked to anthropogenic factors, reflected in several drivers, while compositional similarity F I G U R E 2 Pairwise patterns of shared species between island groups (more yellow colours indicate more shared species). Top (a): Total richness and total species shared between island groups. Bottom (b): Beta Simpson dissimilarity between island groups and average rarity of species on each island group F I G U R E 3 Average rarity of naturalized species on each island group in relation to naturalized species richness for each island group in the Pacific. High average rarity indicates the presence of many naturalized species that are located on few other island groups. A linear model (black line) and its 95% confidence interval are displayed. Names of many island groups with low naturalized richness are not displayed for visibility reasons  alone. We found an even more skewed distribution of naturalized plant richness across island groups than reported by Denslow et al. (2009) on 1,332 naturalized plant species. This is likely due our larger data set that includes more rare naturalized species.
The four most widespread species in our data set are all associated with anthropogenic factors and disturbed habitats, including the following: Eleusine indica (45/50 island groups), an agricultural weed that is resistant to glyphosate (Lee & Ngim, 2000), Cynodon dactylon (44/50), a graminoid which is resilient to drought, salt and trampling (Fischer et al., 2008;Kaffka, 2009), Portulaca oleracea (44/50), a common agricultural weed, and Euphorbia hirta (42/50), a weed with a wide native and naturalized ranges. Other generalist species known to thrive in disturbed ecosystems, such as species of the genus Chenopodium, are currently rare in the Pacific, but could spread further in the future.
Pacific island groups with few naturalized plant species are populated by "common" naturalized species (i.e. these island groups have a low average rarity score), indicating a set of notorious species that are able to establish readily and are frequently transported, even to the most poorly connected places. This pattern has been observed for invasive naturalized species on islands globally (Traveset et al., 2014), and for native species (native species-rich islands have more endemic species) on tropical archipelagos (König et al., 2017) and in the Aegean archipelago (Kallimanis et al., 2010).
As the processes that result in native species patterns are likely to be very different than those of naturalized species, a direct comparison of patterns of rarity for native and naturalized species in the Pacific and in other regions could provide insights into the underlying processes. Further, our result indicates that it is important and feasible to prioritize the prevention of the (further) spreading of common naturalized species on islands with currently low naturalized species richness and that anthropogenic introduction barriers such as trade by ship and plane, but also the availability of human-altered ecosystems, are very important.
We expected the Pacific region to contain more unique naturalized species compared to other regions due to its unique native floras F I G U R E 5 Variable importance from VSURF random forest models explaining species richness (a) and beta Simpson (b) of naturalized plant species in the Pacific. Explanatory variables are ranked by relative variable importance. Anthropogenic, biogeographic and climatic drivers are shown in red, gold and blue, respectively. Richness differences are in shown in grey. Only variables retained by the variable selection process are included. Light colours (least important) indicate that these drivers were not selected as suitable, that is not included in the calculation of the R 2 , but is included in this figure to put them in a broader context and potentially lower resistance to biological invasions (Atkinson & Cameron, 1993;L. Loope & Mueller-Dombois, 1989 . This provides evidence that human activity overcomes geographic barriers (di Castri, 1989(di Castri, , 1990 and that more species in these regions could be exchanged with the Pacific, as saturation in the accumulation of naturalized species has not yet been reached . Vitousek et al. 1987). Many naturalized species may have traits that favour their accidental and purposeful introductions by humans or their abilities to thrive in degraded or anthropogenic ecosystems (Doyle, 2001). There are many ecosystems or whole islands in the Pacific that have been degraded, altered or transformed entirely by human colonialization (Braje et al., 2017), and this may explain the over-representation of these families and highlights the association of naturalized species with human activity.
Variation in naturalized species richness among island groups in the Pacific was explained largely by human activity and to a lesser degree by island group size. Island size was linked to naturalized richness in the Pacific in a previous study (Denslow et al., 2009), but the other variables we consider here were heretofore largely unexplored in the Pacific. In a global analysis of islands, elevation and distance to the mainland were also linked to naturalized plant species richness (Moser et al., 2018). Our results highlight human activity (e.g. human footprint and airport capacity) as the main driver of non-native species richness, while biogeographic variables (e.g. land area and SLMP) play a secondary role. We could not show any influence of distance to mainland on naturalized species richness.
This finding suggests that either naturalized species follow the similar pathways and thrive under similar conditions as humans (e.g. island groups with more area are more attractive for human activities and harbour more habitats that may host non-native species), and/ or the presence of humans increases the availability and strength of introduction pathways (i.e. harbours, airports and trade volume) and suitable habitats (i.e. urban/agricultural/disturbed areas).
We find that island groups are compositionally similar to other island groups that are different in their species richness, GDP and area, but are similar in their mean annual temperature. Speciespoor island groups are inhabited by widespread naturalized species.
However, there is high variation in the beta diversity of speciespoor islands groups, with some pairs having the exact same widespread species and others having completely different compositions.
Species-poor island groups tend to have a nested subset of the species that are present on species-rich island groups (see also Traveset et al. 2014), and thus, pairings of species-poor with species-rich island groups have consistently high values of Simpson pairwise beta diversity. These results suggest that species dispersal likely follows a pathway by which species-poor and monetarily poor islands receive their naturalized species not directly from outside of the Pacific, but rather from dispersal from species-and monetarily rich island groups, which are in turn the initial introduction points of non-native species to the Pacific ("hubs" for invasion, as in Traveset et al., 2014).
Naturalized species that arrive will be more likely to establish if they come from an island group of similar temperature, due to the high importance of niche constrains for establishment. In this context, it is interesting that mean annual precipitation was not considered relevant in our model. Temperature and other niche constraints such as maximum elevation are important factors shaping compositional similarity, while factors related to anthropogenic dispersal define richness.
Our analysis identified three island groups that were notable outliers in terms of richness and composition: Juan Fernandez, the Hawaiian Islands and the Solomon and Bismarck islands. Juan Fernandez has many naturalized species that are not found anywhere else in the Pacific. This could be due to the combined effects of its (a) unique colonization history, first being discovered by Spanish seafarers rather than Polynesians and later strengthening ties to Chile (Cuevas & van Leersum, 2001), (b) high isolation from other island groups and therefore poor connection to major trade routes, (c) unique biogeographic factors that shaped its highly endemic flora (Cuevas & van Leersum, 2001) and (d)  The large number of naturalized plants in the Pacific gives reason to be worried, as some of them are, or will likely become invasive (Jeschke & Pysek, 2018;Traveset et al., 2014, sensu Blackburn et al., 2011Richardson et al., 2000). Even if they do not become invasive, the presence of naturalized species reduces regional floristic distinctiveness through biotic homogenization (Vitousek et al., 1997). Our results highlight the human-mediated connectivity between the Pacific and other parts of the world (see also Capinha et al., 2015) and highlights potential entry points for non-native species. There are groups of naturalized species that may be particularly suited to establish in the Pacific (e.g. because they are pre-adapted to climatic conditions or have traits associated with human use or disturbed ecosystems) and identifying additional common traits of these species holds much promise for predicting future establishment. Naturalized species patterns in the Pacific are intrinsically bound to human activities, but economic growth and naturalized species richness increase do not have to be coupled in the future, as this is a human-made connection that can be mediated using human-made measures. Management strategies to prevent undesired introduction and establishment of non-native must be put in place or strengthened and the same increased global connectivity that leads to the introduction of naturalized species can also limit it Pyšek et al., 2010;Seebens et al., 2015).
Cooperative initiatives to prevent and manage non-native species (e.g. "The Cooperative Islands Initiative"; http://www.issg.org/cii/) could be used to create specific watchlists for the Pacific, especially for invasive species, to make screening in trade easier (SPREP, 2016;Sherley et al., 2000). In this study, we provide a comprehensive overview of patterns and drivers of naturalized species establishment in the Pacific, as a foundation for future work on the processes of naturalized species spread and on applied management of naturalized species.

ACK N OWLED G EM ENTS
We appreciate the helpful comments of two anonymous review-

Pe e r Rev iew
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13260.

DATA AVA I L A B I L I T Y S TAT E M E N T
Our data are compiled from GloNAF and PIER which are both open source. Species names and island/island group names can be found in the Table S1 and S2. We plan to publish our aggregated data set in an upcoming data publication.