Seasonal species richness of birds on the world's islands and its geographical correlates

Abstract

The presence of migratory birds on islands results in seasonal variation in species richness. These patterns and their geographical correlates within the context of island biogeography theory have not been examined. We used 21 years of bird observations on 690 islands from eBird to determine how seasonal species richness estimates vary as a function of island area, isolation and latitude. Species richness was highest on islands within the northern mid-latitudes during migration and on islands within tropical latitudes during the non-breeding season. Area defined positive, nonlinear relationships with species richness across seasons, with the steepest slopes occurring with islands greater than 1100 km². Distance to mainland defined negative, nonlinear relationships with species richness across seasons, with the strongest slopes occurring with islands located greater than 150 km from the mainland. Species–area relationships were weakest for the most remote islands and strongest for islands at intermediate distances to the mainland. Intermediate proximity to other islands was a poor predictor of species richness. Our findings emphasize the presence of seasonally dynamic geographical relationships, the enhanced role of evolutionary processes on larger islands, the unique ecology of the world's most remote islands, and the importance of islands as stopover sites and wintering grounds for migratory bird species.

1. Introduction

Islands are unique places for biodiversity [1]. Although they only make up some 5% of the global land area, islands contain ca 20% of the world's biodiversity [2]. Birds are well represented across the world's islands due in large part to their strong dispersal abilities [3], and the important role islands have played in bird evolution [4,5]. In addition to resident bird species that occur year-round on islands, islands can also host migratory bird species during the breeding or non-breeding seasons, or as transients during migration [6]. Island stopover events by migrants allow individuals to rest and refuel before continuing their migration journey [7–14]. Seasonal variation in the occurrence of migratory birds on islands should therefore result in seasonal variation in species richness. These seasonal patterns and their geographical correlates within the context of island biogeography theory, however, have not been broadly explored.

Through the dynamics of colonization, speciation and extinction, island biogeography theory predicts that species richness should be positively correlated with island area and negatively correlated with island isolation [2,15,16]. These predictions have been broadly corroborated for breeding bird assemblages on islands globally [17,18] and for breeding bird assemblages within island archipelagoes [19–21]. A double sigmoidal species–area relationship has been proposed as the expected outcome for island biota based on how stochastic, ecological and evolutionary processes operate as a function of island size [22,23]. Stochastic processes are expected to be more pronounced on the smallest islands, resulting in species richness being independent of island area (small island effect). Ecological processes (immigration/extinction dynamics) are expected to dominate with intermediate sized islands, resulting in a deterministic relationship between island area and species richness. With the largest islands, evolutionary processes are expected to increase in prominence through internal geographical isolation, enhancing the strength of the species–area relationship through increased opportunities for in situ speciation [24]. Current evidence for breeding bird assemblages on islands globally, however, suggests species–area relationships lack a double sigmoidal form and are largely linear [17,18].

In addition to the individual relationships with species richness, island biogeography theory predicts an interaction should exist between island area and isolation where the strength of the species–area relationship increases with increasing island isolation [2,15,16]. Here, differences in immigration rates between near-shore and remote islands is expected to result in different species–area relationships [15]. Higher immigration rates on near-shore islands are expected to maintain higher species numbers on small islands by reducing the chances of extirpations, resulting in weaker species–area relationships [25]. On remote islands, the occurrence of these ‘rescue effects' is unusual, resulting in fewer species on small islands and stronger species–area relationships [2]. Evidence for an interaction between island area and isolation, however, has not been broadly documented [26], including for breeding bird assemblages on islands worldwide [17].

How species–area and species–isolation relationships are defined for birds outside of the breeding season has not been explored. During the non-breeding season, we would expect similar relationships to those documented globally during the breeding season [17,18]. How these relationships will be affected by the presence of transient species during migration will depend on how island area and isolation affect visitation rates. Larger islands should represent more obvious stopover targets. We would therefore expect the species–area relationship to become steeper during migration as migratory species occur in greater numbers on larger islands. Islands that are located in closer proximity to other islands should also represent more obvious stopover targets, which increase the slope of the species–proximity relationship during migration. Islands that are located closer to the mainland are more likely to be encountered by birds travelling within migration flyways, which are situated over continental landmasses [27]. We would therefore expect the slope of the species–isolation relationship to weaken during migration as the number of migratory species increases on near-shore islands.

It has been well documented for birds and other taxa that species richness across the world's continents peaks within the tropics and declines as you move towards the poles [28,29]. This pattern is complicated by the seasonal occurrence of migratory bird species. Most migratory bird species spend the non-breeding season (boreal winter) within tropical latitudes and spend the breeding season (boreal summer) within the temperate latitudes of the Northern Hemisphere [30]. This seasonal redistribution of migratory species from the tropics to northern temperate regions creates a reverse latitudinal gradient in species richness where species richness peaks within the mid-latitudes of the Northern Hemisphere during the breeding season (boreal summer) [31–33]. We would expect these seasonal patterns of species richness to be replicated across the world's islands. Specifically, species richness during the boreal summer should be higher than that of the boreal winter for islands located within the mid-latitudes of the Northern Hemisphere. Species richness during the boreal winter should be higher than that for the boreal summer for islands located within the tropics. During migration, islands located within the mid-latitudes of the Northern Hemisphere should contain the greatest number of species during migration and the fewest species during the boreal winter. Deviation from these predictions would suggest the seasonal movements of migratory birds between the tropics and northern temperate regions does not occur across islands in a similar manner as it does across the mainland.

To test these predictions, we use bird occurrence information from eBird over a 21-year period (2002–2022) to estimate avian species richness [34] during four seasons of the annual cycle (boreal winter, spring, summer and autumn) across a large number of well-surveyed islands (n = 690). eBird is an online database initiated in 2002 that allows volunteers to submit their observations from any location and time, creating a global full-annual cycle source of occurrence information on the world's birds [35]. To date, eBird has sampled over 20% of the world's islands and, based on availability, eBird participants have a propensity to sample islands that are more isolated from the mainland and other islands [36]. By testing these predictions, our objective is to document how the relationships between species richness and island area, isolation, proximity and latitude are defined across seasons, and where and when migratory birds are visiting the world's islands.

2. Methods

(a) . Bird occurrence data

eBird contains bird obervations in checklist format where species detected by sight or sound are recorded by one or more observers during a sampling event [37]. eBird represents a semi-structured ‘big data' resource where volunteer observers select from a number of predefined sampling protocols where sampling effort is determined by the observer. We compiled bird occurrence information from all available eBird checklists globally from 1 January 2002 to 1 May 2022. We queried the database on 2 May 2022, and we included all sampling protocols, all levels of sampling effort, and both designations (yes/no) for the field ‘all observations reported'. We only considered observations that were identified as valid in the database, we combined observations in grouped checklists into single checklists, and we excluded all pelagic seabird species from the checklists. A total of 70 001 315 checklists were available for analysis containing 10 148 unique species.

(b) . Island spatial data

We acquired a shoreline vector coverage of the world's continents and 21 813 of the world's islands greater than 1 km² in area from [38]. We identified 4958 islands that contained bird observations from eBird (figure 1a). From these, 4529 were named islands. Across the 4958 islands, a total of 6 521 894 eBird checklists were available for analysis containing 6411 unique species. We estimated the distance to the nearest continental shoreline (excluding Antarctica) for all 21 813 islands by rasterizing the shoreline vector coverage of the world's continents to ca 10 km spatial resolution [36]. We first calculated the distance of each non-continental pixel to the nearest continental shoreline pixel. We then found the minimum distance to the continental shoreline for each island. Following the methods of previous island biogeography assessments, we log transformed island area and minimum distance to the continental shoreline to improve their distributional properties for analysis. We estimated the spatial proximity of other islands within a 2000 km radius from the edge of each island using a proximity metric based on an inverse distance weighting scheme [36]. We applied a power transformation (0.25) to the proximity index to improve its distributional properties for analysis.

Figure 1.

Estimated seasonal species richness of birds on 690 islands during the boreal winter, spring, summer and autumn. We defined the four seasons based on the following dates for boreal spring (15 March to 15 June) and boreal autumn (15 August to 15 November). (Online version in colour.)

(c) . Seasonal species richness estimates

We used the methodology developed by Lobo [34] to estimate seasonal species richness on the 4958 islands, a method used in previous eBird studies investigating global patterns of species richness [35,36]. Specifically, we estimated survey completeness of bird occurrence information for all years combined during four seasons (boreal winter, spring, summer and autumn) on the 4958 islands using the ‘exact’ estimator from [39]. We defined the four seasons based on the following dates for the boreal spring (15 March to 15 June) and boreal autumn (15 August to 15 November). We estimated the relationship between the accumulated number of species and survey effort for each island using the eBird checklists as a unit of survey effort, replicating the approach used in previous studies [35,36]. We defined survey completeness for each island as the percentage of observed species richness captured by the species richness estimates derived from the species accumulation curves [34]. We removed poorly sampled islands where the ratio between the number of occurrence records and the number of observed species was less than 3, the slope of the species accumulation curve was greater than 0.3 and survey completeness was less than 50% [34]. We then removed islands where species richness estimates were not available for all four seasons. We estimated seasonality in species richness for each island based on the maximum seasonal difference in species richness between the migration and non-migration periods. Specifically, the maximum species richness during either the boreal spring or boreal autumn minus the minimum species richness during either the boreal summer or boreal winter.

(d) . Statistical analysis

We examined how species richness varied among seasons using a one-way repeated measures Poisson mixed-effect ANOVA with a log link function and Tukey corrected post hoc pairwise comparisons with island included as a random effect. We examined the individual relationships between species richness and island area, distance, proximity and latitude using Poisson generalized additive mixed models (GAMM) with a log link function and exponential spatial correlation structure using the great circle (Haversine) distance [40,41]. We examined the individual relationships between our estimates of seasonality in species richness (defined above) and island area, distance, proximity and latitude using the same method. We selected a basis dimension (k) of six as a starting value for the smooth terms in the GAMM fits. We selected this value after it was determined that higher values did not result in statistical important changes in the results [42]. We tested five GAMM spatial correlation structures [43], which we compared using Akaike information criteria [44]. In all cases, the exponential correlation structure performed best. We used the locations of the GAMM fitted lines and 95% confidence bands to assess the level of evidence that the slope of the fitted lines differed significantly from zero and the level of evidence for significant differences among seasons. We report the effective degrees of freedom (edf) and approximate significance of the smooth terms from each GAMM fit to assess the nonlinear complexity of the GAMM relationships. We examined the level of evidence for an interaction between island area and distance to mainland for each season using a smooth interaction term in a Poisson GAMM with a log link function and exponential spatial correlation structure using the great circle (Haversine) distance. We interpret the presence of nonlinear and non-parallel contour lines in the plot of the smooth interaction term as evidence for an interaction. To assess the nonlinear complexity of the smooth interaction term and contour lines, we report edf and approximate significance of the smooth interaction term.

All analysis was conducted in the R software environment for statistical computing and graphics, v. 4.2.0 [45]. We estimated distance to mainland using the distance function in the raster R package [46]. The survey completeness analysis was implemented using the KnowBPolygon function in the KnowBR R package using the default parameters [34]. We implemented the Poisson mixed-effect ANOVA using the glmer function in the lme4 R package [47], and the Tukey corrected post hoc pairwise comparison using the glht function in the multcomp R package [48]. We implemented GAMM using the gamm function in the mgcv R package [42].

3. Results

From a total of 21 813 islands, 4958 islands (22.7%) contained bird occurrence information in eBird. From these, 3410 islands (15.6%) contained sufficient occurrence information to estimate survey completeness for at least one season. After removing poorly sampled islands, 690 islands (3.2%) contained enough occurrence information to estimate species richness for all four seasons (figure 1). The 690 islands occurred across the globe with the densest concentrations occurring along the Atlantic and Pacific coasts of North America (figure 1).

When the seasonal species richness estimates for the 690 islands were compared across seasons, species richness was highest during the boreal autumn for 293 islands (42.5%) located along the coasts of North America and Europe, within Southeast Asia and within Australasia (figure 2). Species richness was highest during the boreal spring for 207 islands (30.0%) located primarily within the mid-latitudes of the Northern Hemisphere (figure 2). Species richness was highest during the boreal winter for 179 islands (25.9%) located within the tropical latitudes and within the Southern Hemisphere (figure 2). Lasty, species richness was highest during the boreal summer for 11 islands (1.6%) located across the globe (figure 2).

Figure 2.

The islands containing the highest estimated seasonal species richness of birds from a total of 690 islands during the boreal winter, spring, summer and autumn. The per cent of the total number of islands for each season is shown in parentheses. (Online version in colour.)

Seasonality in species richness for the 690 islands was generally higher for islands located in the Northern Hemisphere (figure 3a). Island area had a nearly uniform relationship on average with seasonality in species richness (edf = 2.6, F = 7.2, p < 0.001; figure 3b). Seasonality in species richness displayed on average a negative relationship with distance to the mainland for islands that were located greater than 150 km from the mainland (edf = 3.0, F = 38.0, p < 0.001; figure 3c). Seasonality in species richness displayed on average a nonlinear relationship with proximity to other islands with seasonality peaking at intermediate levels of proximity (edf = 4.0, F = 8.68, p < 0.001; figure 3d). Seasonality in species richness displayed on average a strong nonlinear relationship with latitude with seasonality peaking within the mid-latitudes (centred on ca 30° N latitude) of the Northern Hemisphere (edf = 5.7, F = 41.0, p < 0.001; figure 3e).

Figure 3.

(a) Seasonality in species richness of birds on 690 islands. Seasonality is defined as the maximum seasonal difference in species richness between the migration (boreal spring and autumn) and non-migration periods (boreal summer and winter). Fitted lines with 95% confidence bands from generalized additive mixed models of seasonality in species richness of birds on 690 islands by (a) island area, (b) distance to mainland (excluding Antarctica), (c) spatial proximity to other islands (0.25 power transformed) and (d) the latitude of the centre of the island (degrees). (Online version in colour.)

Seasonal species richness estimates for the 690 islands varied significantly on average among the four seasons (χ² = 1083.9 d.f. = 3, p < 0.001; figures 1 and 4a). Species richness was higher on average during the boreal spring compared to the boreal autumn (z = 2.63, p = 0.042), boreal winter (z = 11.98, p < 0.001) and boreal summer (z = 26.80, p < 0.001). Species richness was higher on average during the boreal autumn compared to the boreal winter (z = 14.81, p < 0.001) and boreal summer (z = 26.80, p < 0.001). Species richness was higher on average during the boreal winter compared to the boreal summer (z = 14.81, p < 0.001).

Figure 4.

(a) Boxplot of estimated seasonal species richness of birds on 690 islands during the boreal winter, spring, summer and autumn. Fitted lines with 95% confidence bands from generalized additive models of estimated seasonal species richness of birds on 690 islands during four seasons by (b) island area, (c) distance to mainland (excluding Antarctica), (d) spatial proximity to other islands (0.25 power transformed) and (e) latitude of the centre of the island (degrees). Note the different scales on the y-axes. (Online version in colour.)

Species richness estimates increased in a nonlinear fashion on average with increasing island area for all four seasons (edf = 5.0, F = 1151.31, p < 0.001; figure 4b). Species richness was consistently higher on average during the boreal spring and boreal autumn with increasing island area except for the largest islands where values converged across seasons (figure 4b). The strongest species–area relationships during the boreal winter, spring and autumn occurred on average for islands ca less than 150 km² and islands ca greater than 1100 km², and the strongest species–area relationship during the boreal summer occurred on average for islands ca greater than 1100 km² (figure 4b).

Species richness decreased in a nonlinear fashion on average with increasing island isolation for all four seasons (edf = 5.0, F = 785.37, p < 0.001; figure 4c). Species richness was consistently higher on average during the boreal spring and boreal autumn with increasing island isolation except for the most isolated islands where the values converged across seasons (figure 4c).

Species richness increased in a nonlinear fashion on average with increasing proximity to other islands for all four seasons (edf = 5.0, F = 188.76, p < 0.001; figure 4d). Species richness was highest on average across seasons when proximity was greatest, with a peak in species richness occurring at intermediate levels of island proximity, especially during the boreal spring and autumn (figure 4d).

Species richness decreased at higher northern and southern latitudes on average for all four seasons (edf = 4.9, F = 410.20, p < 0.001; figure 4e). Species richness reached its highest levels on average during the boreal spring and boreal autumn within the mid-latitudes of the North Hemisphere (centred between 40° and 50° N latitude; figure 4e). Species richness reached its highest levels on average during the boreal winter within the tropics north of the Equator (centred on ca 10° N latitude; figure 4e). Species richness reached its highest levels on average during the boreal summer within the tropics south of the Equator (centred on ca 10° S latitude) and within the mid-latitudes in the Northern Hemisphere (centred on ca 50°N latitude; figure 4e).

The species richness response surfaces describing the interaction between island area and distance to mainland for each season contained nonlinear and non-parallel counter lines whose arrangements differed across seasons (edf = 28.3, F = 393.63, p < 0.001; figure 5). Species richness was lowest across all four seasons on the smallest, most remote islands (figure 5). Species richness was highest across all four seasons on the largest islands located at intermediate distances from the mainland (figure 5). These large, species rich islands were centred on ca 150 km from the mainland during the boreal winter and boreal summer and were centred on ca 250 km from the mainland during the boreal spring and boreal autumn (figure 5). With increasing distance to the mainland, the positive effect of area on species richness decreased in strength across seasons (figure 5). With increasing island area, the negative effect of distance on species richness decreased in strength across seasons (figure 5).

Figure 5.

Contour plots of estimated seasonal species richness on 690 islands during the boreal winter, spring, summer and autumn as a function of the smooth interaction between island area and distance to mainland (excluding Antarctica) from generalized additive mixed models. The contour interval scales are identical across the four contour plots. (Online version in colour.)

4. Discussion

Across the world's islands, we identified seasonally dynamic relationships between species richness and island area, isolation and latitude that did not completely follow our expectations. For the majority of islands (72.5%), species richness was highest during migration, emphasizing the role of islands as stopover sites for migrating birds, especially within the mid-latitudes of the Northern Hemisphere. Species richness tended to be higher during the boreal winter compared to the boreal summer, emphasizing the role of islands as wintering grounds for migratory birds within the tropics and the limited role of islands as breeding grounds for migratory birds within the Northern Hemisphere. Seasonal variation in species richness was greatest on islands located near the mainland (less than 150 km) and on islands located within the mid-latitudes of the Northern Hemisphere. Island area and proximity to other islands had little effect on seasonal variation in species richness. Contrary to previous findings [17,18], we found nonlinear species–area relationships across all four seasons with evidence for a major threshold at 1100 km² and a minor threshold at 150 km² during the boreal winter, spring and autumn. These findings highlight the role of evolutionary processes on the largest islands [22,23] and the importance of small and intermediate sized islands as stopover sites and wintering grounds for migratory birds. We found nonlinear relationships with distance to the mainland with a threshold occurring at 150 km from the mainland beyond which species numbers declined across seasons. These findings emphasize the ability of birds to dampen the species–isolation effect with near-shore islands and the role of near-shore islands as stopover sites and wintering grounds for migratory birds. We found nonlinear relationships with proximity to other islands where intermediate proximity was a poor predictor of species richness across seasons. These results suggest the effects of increasing proximity are strongest for species poor islands that are isolated from other islands and species rich islands that are not isolated from other islands, especially during the boreal winter, spring and autumn. Contrary to previous findings [17], we found evidence for an interaction between island area and isolation across all four seasons that diverged from the predictions from island biogeography theory [2,15,16]. Here, the species–area effect was weakest on small islands and strongest on islands located at intermediate distances from the mainland.

The relationships documented in this study between species richness and latitude broadly followed our expectations. The latitudinal species richness gradient during migration peaked within the mid-latitudes of the Northern Hemisphere, where the majority of the world's migratory birds occur during migration [30], emphasizing the importance of islands as stopover sites during migration. The latitudinal species richness gradient during the boreal winter peaked within the tropical latitudes, the region where the majority of the world's migratory birds spend the non-breeding season [30], emphasizing the importance of tropical islands as non-breeding grounds for migratory birds. Latitudinal species richness gradients during the boreal summer contained two peaks, one just south of the Equator and one within the mid-latitudes of the Northern Hemisphere. These results reflect the occurrence of large numbers of migratory birds within the mid-latitudes of the Northern Hemisphere during the breeding season [30]. These findings also replicate the reverse latitudinal gradient that has been documented on the North American continent during the boreal summer [31–33].

The species–area relationships documented in this study did not display the double sigmoidal form predicted to occur based on how stochastic, ecological and evolutionary processes are expected to operate as a function of island size [22,23]. We identified a significant threshold at roughly 1100 km² beyond which the slope of the species–area relationship increased substantially. These findings suggest the presence of stronger evolutionary processes on larger islands where internal geographical isolation increases the chances of in situ speciation [22,23]. We also identified a weaker threshold at roughly 150 km² during the boreal winter, spring and autumn. These findings suggest transient species occur in large numbers on small to intermediate sized island during these seasons, increasing the strength of the species–area relationship. Our results did not provide strong evidence for a small island effect where the species–area relationship is weakened through stochastic processes. There is a chance the small island effect was not well represented in our data due to the absence of very small islands (less than 1 km²) and the absence of islands where no birds were observed (empty islands) [49]. In total, our findings suggest we failed to capture the transition from stochastic to ecological processes (ecological threshold T₁) but we did capture the transition from ecological to evolutionary processes (evolutionary threshold T₂) [22,23]. By capturing the evolutionary threshold, our findings emphasize the ability of speciation to enhance the species–area relationship on larger islands [24].

Our analysis of seasonal variation in species richness did not find a significant effect of island area. This outcome could reflect the use of broad migration fronts [27] where all islands, independent of size, have the potential of being encountered during migration. By contrast, seasonal variation in species richness showed a negative relationship with island isolation, which was most pronounced for islands located further than 150 km from the mainland, replicating the threshold identified with the species–isolation relationship. These results suggest that the occurrence of migratory bird species on near-shore islands is an important source of both species numbers and seasonal variation in species richness. For the most remote islands, our estimates of seasonal variation in species richness were very close to zero, suggesting that the occurrence of migratory bird species on these islands is highly unusual. The 150 km threshold may reflect the unwillingness of migrants to deviate from the safety of the coastline or could reflect the use of transoceanic ‘shortcuts' between continents. For example, migrants crossing the Gulf of Mexico and Caribbean Sea between North and South America [33,50,51] and migrants crossing the Mediterranean Sea between Europe and Africa [52,53]. Current evidence suggests vagrancy is common on near-shore islands, indicating birds are willing to cross relatively short oceanic barriers, especially when the island can be seen from the coast [54,55]. Travelling further out to sea where islands tend to be smaller and more isolated would substantially increase the chances of mortality.

The interactions between island area and isolation documented in this study did not follow expectations based on island biogeography theory [2,15,16]. We did not find weaker species–area relationships for nearshore islands and stronger species–area relationships for more remote islands. Rather, we found a stronger species–area relationships for islands at intermediate distances to the mainland and a weaker species–area relationships for the most remote islands. For islands at intermediate distances to the mainland, our findings suggest that the unique combination of stochastic, ecological, evolutionary processes on these islands is responsible for the stronger species–area relationship, with evolutionary processes on the largest islands likely playing a significant role. The weak species–area relationship for the most remote islands is likely a product of several factors including the relatively small size of the islands, the rarity of immigration events on the islands, the dependence on in situ speciation to populate the islands, and the recent extinctions of many endemic island bird species [56,57].

The presence of migratory species on islands could have several possible consequences for island biogeography theory. If a migratory species is purely transitory, the presence of these species would likely have little effect on an island's immigration/extinction dynamics. A possible exception could occur if the migratory species interfered with the survival or fitness of resident species, increasing the chances of their extinction. For example, through predation of young or adults, interfering with mating or foraging behaviour, the depletion of food resources, or the degradation of habitat. Another exception could occur if the migratory species forms a founder sedentary population. This could occur on islands where, once landfall has been achieved, continuing the migration journey is no longer advantageous due to the island's remoteness and the physical state of the bird. The formation of a sedentary population could also lead to speciation through a variety of processes [58]. Once established, the influx of new individuals during migration could mitigate the risk of extinction through demographic and genetic contributions [25]. Testing the relevance of these different effects would clarify the role of migratory birds as a factor affecting an island's speciation processes and immigration/extinction dynamics.

In summary, our study provides the most comprehensive empirical assessment to date on the patterns and geographical correlates of the seasonal species richness of birds across the world's islands. One limitation of our approach is that the dates we used to divide the year into four seasons are not absolute across the world's islands. Using eBird or other data sources to generate island specific migration phenology estimates would likely improve the quality of our seasonal species richness estimates. The thresholds identified in our analysis for island area and distance can be used to inform and guide research, monitoring and conservation efforts for the world's migratory birds. Evidence for an interaction between area and isolation that is contrary to expectations highlights the ecological isolation of the world's most remote islands and the continuing vulnerability of these systems to effects of anthropogenic activities, and the importance of evolutionary processes on the world's largest islands. Acquiring additional insights into the nature of these seasonal patterns and the role of stochastic, ecological and evolutionary processes will provide a basis to better understand how resident and migratory bird species function within island systems across their annual life cycles.

Data accessibility

The bird occurrence data are available from https://ebird.org/science/use-ebird-data and the shoreline vector coverage of the world's continents and islands is available from [38]. All data used in the analysis are available at https://doi.org/10.5061/dryad.m0cfxpp67 [59].

Authors' contributions

F.A.L.: conceptualization, formal analysis, methodology, visualization, writing—original draft; M.S.: conceptualization, writing—review and editing; A.M.D.: writing—review and editing; E.T.M.: writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by The Wolf Creek Charitable Foundation and the National Science Foundation (ABI sustaining: DBI-1939187).