The island biogeography of human population size

Abstract

For decades, biogeographers have sought a better understanding of how organisms are distributed among islands. However, the island biogeography of humans remains largely unknown. Here, we investigate how human population size varies among 486 islands at two spatial scales. At a global scale, we tested whether population size increases with island area and declines with island elevation and nearest mainland, as is common in non-human species, or whether humans escape such biogeographic constraints. At a regional scale, we tested whether population sizes vary among islands within archipelagos according to the positioning of different cultural source pools. Results illustrate that on a global scale, human populations increased in size with island area, similar to non-human species, yet they did not decline in size with elevation and distance to nearest mainland. At a regional scale, human population size often varied among islands within archipelagos relative to the location of different cultural source pools. Despite broad-scale similarities in the geographical distribution of human and non-human species among islands, results from this study indicate that the island biogeography of humans may also be influenced by archipelago-specific social, political and historical circumstances.

1. Introduction

Humans occupy almost every corner of the globe. Islands are no exception, despite their geographical isolation. However, some archipelagos are densely populated (e.g. Long Island New York, Singapore), while others are nearly uninhabited (e.g. Svalbard Islands, Tristan da Cunha). Although the geographical correlates of human population sizes are well known on continents [1–6], an overarching explanation for global variation in insular population sizes has yet to be established [7–10].

Other species occupy islands in predictable ways [11,12]. Larger islands typically support bigger populations, because they contain more inhabitable area and resources [11,12]. In humans, larger islands contain greater linguistic diversity and have been occupied more commonly and continuously historically [10,13]. By contrast, population sizes might be expected to decline with island isolation. Isolated islands have lower immigration rates, potentially producing smaller population sizes [8,11,12]. Isolated islands were often colonized more recently and less consistently by humans [8,9]. In addition, humans might prefer flat islands, which have larger surfaces suitable for agriculture [1,2], thus facilitating larger population sizes. While these drivers of island distributional patterns have been thoroughly explored for other species [14–21], they remain largely unknown for humans [8,10].

Humans have undergone profound cultural transformations through time. Resources are now extracted and relocated at unprecedented pace and scale [22]. Larger islands generally offer more resources; however, greater external supply might aid island populations circumvent dependency on local resources [22]. In turn, the size of a population might be unrelated to that of an island. Next, technological advancements in transportation (i.e. mechanized vehicles) allow for effortless and rapid movement across islands [23], reducing geographical barriers. Thus, population sizes might be independent of island isolation. Finally, island economies are now increasingly reliant on tourism rather than agriculture [24,25] and therefore topography might have little importance in shaping modern island inhabitation. Owing to such economic and technological transformations, humans might escape biogeographical constraints.

In many species, biological traits are complemented by cultural traits [26–28]. In humans though, cultural traits play a much more extensive role than in any other species [29–31]. Cultural traits are social attributes that both originate and transmit through non-genetic pathways [32,33]. When sets of cultural traits are shared among individuals, distinct societies are generated [27]. Societies turn over through both space and time [33,34], often resulting in complex economic, historical and political circumstances [10,35]. Events such as wars [36], famines [37], migrations [38], displacements [39], colonialism [40], diseases [41], socioeconomic changes [42] and, more recently, tourism [25] might all determine abrupt changes in human population size on islands. Thus, cultural differences may affect spatial variation in human population size on islands.

Humans progressively colonized many archipelagos across the planet [8,43]. However, in non-human species, directionality in the colonization process of each archipelago varies with the number of source pools and their location. The contribution of two distinct source pools to an island biota may increase with proximity to each of them. As the distance from each pool source increases, species will be gradually filtered out in either direction, a process known as two-way filtering [44]. Similarly, human populations within an archipelago might vary in size depending on proximity to the respective cultural source pool.

Here, we investigate the island biogeography of human population size on 486 islands within 10 archipelagos distributed globally (figure 1). At a global scale, among archipelagos, we tested whether human population sizes vary geographically as predicted by earlier work on non-human species (i.e. increase with island area and decline with isolation and elevation) or whether humans escape these biogeographic constraints. At a regional scale, we tested whether population sizes vary among islands within archipelagos according to the positioning of different cultural source pools.

Figure 1.

Map of the globe displaying the investigated archipelagos.

2. Methods

(a) . Study system and island characteristics

We selected 10 archipelagos across the globe that encompassed a wide range of climatic, environmental and geographical conditions (total n = 486 islands; table 1, figure 1). Archipelagos range from tropical to subarctic climates, some are oceanic and some continental. There is at least one archipelago per continent, spanning both hemispheres (table 1, figure 1). We included both habited and uninhabited islands. Population size was extracted from the latest available census (electronic supplementary material, Appendix S1 and table S1). Census information was unavailable for small islands in Northern New Zealand, so in this instance, we estimated human population size by counting the total number of dwellings visible from aerial imagery [45].

Table 1.

List of archipelagos investigated and associated number of islands, nearest mainland, hemisphere, continent and climate.

archipelago	number of islands	nearest mainland (within 500 km)	hemisphere	continent	climate
Seychelles	79	na	Southern	Africa	tropical
Samoa	22	na	Southern	Oceania	tropical
Hawaii	45	na	Northern	Oceania	tropical
Virgin Islands	48	Hispaniola	Northern	America	tropical
Northern New Zealand	70	New Zealand North Island	Southern	Oceania	temperate to subtropical
Zhoushan	66	Southeast Asia	Northern	Asia	temperate to subtropical
Kuril Islands	33	Kamchatka Peninsula, Honshū island	Northern	Asia	continental to subantarctic
Channel Islands	40	Mainland Europe, Great Britain	Northern	Europe	temperate
Ionian Islands	48	Greek and Italian peninsulas	Northern	Europe	temperate/Mediterranean
Shetland	35	Great Britain, Scandinavian peninsula	Northern	Europe	temperate to subpolar
Total	486

To assess geographical variation in human population size, we measured four island characteristics (see the electronic supplementary material, Appendix S1 and table S2). Estimates of island area (km²) and elevation (m) were obtained from freely available sources [46]. When estimates of island area were not available, islands were manually digitized and their total planar area calculated using geospatial software [47]. At a global scale, we measured isolation as the distance from the mainland (hereafter ‘nearest mainland’, km), which we defined as the closest landmass to the archipelago that was at least three orders of magnitude larger. The same metric was employed at a regional scale, which we defined as an area within a 500 km radius from an archipelago. At a regional scale, we employed the same isolation metric but, to account for multiple population sources, we included up to the two closest landmasses (first and second nearest mainland). We defined the regional scale as an area within a 500 km radius from an archipelago. If no landmass was detected within 500 km, no nearest mainland was included in the regional scale analysis. In the case of the Kuril Islands, both the two nearest landmasses were encapsulated within the first nearest mainland metric, thus no second nearest mainland was measured.

(b) . Statistical analyses

To assess global-scale variation in human population size, we conducted a generalized linear mixed model. Population size was included as the dependent variable and island area, elevation and nearest mainland were included as independent variables. The archipelago was included as a random factor to account for the spatial aggregation of islands. All variables were log-transformed to conform to the assumptions of normality and linearity. One was added to each replicate to avoid undefined values. Analyses were conducted using lmer [48] and lmerTest [49] packages in the R environment [50]. Prior to analyses, we used Moran's I coefficients to test for spatial autocorrelation in both dependent and independent variables in ArcGIS 10 [47]. None of the variables was significantly spatially autocorrelated (electronic supplementary material, Appendix S1 and table S3).

To assess how population sizes vary among islands within archipelagos according to the positioning of different cultural source pools, we conducted separate generalized linear models for each archipelago. We used population size as the dependent variable and island characteristics (i.e. area, elevation, first and second nearest mainland) were included as independent variables. Population size was modelled using Poisson or, in cases of overdispersion, quasi-Poisson models (electronic supplementary material, Appendix S1 and table S4). Model fit was calculated by dividing explained deviance by null deviance. Predictor variables were variously transformed to conform to assumptions. Island area and elevation were log-transformed. First and second nearest mainland were not transformed. First nearest mainland was square root-transformed for Northern New Zealand and log-transformed for the Ionian Islands (electronic supplementary material, Appendix S1 and tables S5 and S6).

Before analyses, we tested for multi-collinearity among predictors using Pearson's product motion correlation, with cut-offs at r = 0.80 [51]. On a global scale, none of the island characteristics was strongly correlated (electronic supplementary material, Appendix S1 and table S7). On a regional scale, elevation was often correlated with area (electronic supplementary material, Appendix S1 and tables S5 and S6). Under these circumstances, we chose area because it is the most commonly used variable in island biogeography [11,12]. Analyses of multi-collinearity using variance inflation factors yield consistent results (electronic supplementary material, Appendix S1 and tables S5–S7).

3. Results

Human population size on islands across the globe varied from 0 to 953 207, with an average of 6988 individuals per island. The most populated archipelagos were Zhoushan (mean population size = 16 399.59), Hawaii (m = 30 242.20) and Samoa (m = 11 138.00), while the least populated were Northern New Zealand (m = 77.59), Kuril Islands (m = 581.46) and Shetland (m = 661.91) (electronic supplementary material, Appendix S1 and table S2). Global-scale analyses showed the size of human populations varied only with island area and that the trend was positive (figure 2; electronic supplementary material, Appendix S1 and table S8). No significant relationship was detected between population size and both elevation and nearest mainland (figure 2; electronic supplementary material, Appendix S1 and table S8).

Figure 2.

Relationships between human population size (partial residuals) and island (a) area, (b) elevation and (c) nearest mainland on 486 islands worldwide. Symbol colours and shapes represent different archipelagos. Solid lines indicate significant relationships, while dashed lines non-significant relationships. In light grey are confidence intervals. All variables were log-transformed. Archipelago's names are abbreviated as follows: SEY (Seychelles), SAM (Samoa), HAW (Hawaii), VIR (Virgin Islands), NNZ (Northern New Zealand), ZHO (Zhoushan), KUR (Kuril Islands), CHA (Channel Islands), ION (Ionian Islands) and SHE (Shetland).

At a regional scale, population sizes increased with island area in all 10 archipelagos (table 2). Area alone explained at least two-thirds of variability in all models (electronic supplementary material, Appendix S1 and table S4). Elevation was strongly correlated with island area in most cases and was therefore included only in three cases. Population size was unrelated to elevation in the Seychelles and Kuril Islands but declined with it in Samoa. For three archipelagos, Seychelles, Samoa and Hawaii, no landmass was detected within 500 km and thus no nearest mainland was included in the models. Three archipelagos were within a 500 km distance of a single landmass. Population size declined with isolation in Northern New Zealand, while it increased in the Virgin Islands and Zhoushan. For the last four archipelagos, the Kuril Islands, the Channel Islands, the Ionian Islands and Shetland, there were two landmasses within the 500 km radius. Since they lay as a chain between two landmasses, the nearest mainland encompassed both the first and second mainland in the Kuril Islands, and population size was unrelated to it. In all other cases, the relationship between distance from both nearest mainlands and population size was negative (figure 3).

Table 2.

Generalized linear models exploring the relationship between human population size and island characteristics on 10 study systems worldwide. (Entries are estimates (± s.e.) and t-values. Significant relationships are in italics (p < 0.05). Area and elevation were log-transformed. The nearest mainland was not transformed, except for Northern New Zealand (square root-transformed) and the Ionian Islands (log-transformed).)

archipelago	area		elevation		nearest mainland		second nearest mainland
	estimate ± s.e.	t-value	estimate ± s.e.	t-value	estimate ± s.e.	t-value	estimate ± s.e.	t-value
Seychelles	1.26 ± 0.59	2.13	−0.86 ± 0.67	1.29
Samoa	1.10 ± 0.25	4.42	−1.06 ± 0.45	−2.36
Hawaii	0.59 ± 0.14	4.18
Virgin Islands	1.15 ± 0.10	12.01			0.03 ± 0.01	4.39
Northern New Zealand	1.13 ± 0.17	6.84			−0.51 ± 0.17	−3.11
Zhoushan	1.37 ± 0.05	30.43			0.02 ± 0.00	5.50
Kuril Islands	1.26 ± 0.30	4.14	0.00 ± 0.00	0.02	0.00 ± 0.00	0.43
Channel Islands	1.88 ± 0.02	123.60			−0.00 ± 0.00	−2.37	−0.02 ± 0.00	−28.7
Ionian Islands	0.75 ± 0.08	9.37			−0.14 ± 0.06	−2.23	−1.16 ± 0.30	−3.84
Shetland	0.86 ± 0.08	10.97			−0.02 ± 0.01	−4.07	−0.02 ± 0.01	−2.31

Figure 3.

Map representing two-way filtering in three archipelagos across the planet. Channel Islands (a), Ionian Islands (b), and Shetland (c). Arrow direction indicates directionality in the colonization process. Arrow colour shows first (black) and second (grey) nearest mainland; t-values illustrate the effect of isolation on population size and are reported on the side of each arrow. Country codes follow the International Organization for Standardization alphabetic codes (ISO 3).

4. Discussion

(a) . Global-scale analysis

The number of people inhabiting islands was strongly structured geographically. On a global scale, similar to most species of plants and animals, human populations were bigger on large islands. Positive relationships between population size and island area are among the most common patterns in ecology [52–54], and they can arise from a variety of processes [55,56]. Perhaps the simplest explanation is that larger islands offer more inhabitable space [11,12,55]. Thus, they can house larger human populations than small islands all else being equal [55]. However, this result does not exclude the possibility that humans are less reliant on local resources than in the past. If larger communities initially settled on bigger islands, they might now favour further immigration by offering a wider range of services and resources, thus sustaining positive population size–area relationships.

Population size across all islands did not decline with isolation, suggesting humans might have escaped such biogeographic constraints. Human mobility has changed markedly in the past few centuries [23]. For instance, infrastructures that facilitate air access (airstrips, airports, etc.) can alter immigration pathways and reduce the overall effect of isolation [57,58]. New immigrants will more likely reach islands with such infrastructures first, irrespective of their distance to the mainland. Conversely, past migrations were reliant on water transportation and would probably reach the closest islands first [8], depending on currents and prevailing winds [59]. Future studies should incorporate accessibility and connectedness metrics in assessing the effect of isolation on insular population size. However, here we used the distance from the nearest mainland as a measure of isolation, as it is the traditional metric for investigating insular distributional patterns in other species.

Similarly, population size was unrelated to elevation. While flat islands allow more surface for agriculture [2,60], this activity had been often abandoned [61]. Instead, many islands are now highly reliant on tourism [24,25] or managed as reserves [61]. In summary, changes in land use from cultivation to leisure and conservation might explain a lack of relationship between population size and elevation.

(b) . Regional scale analysis

On a regional scale, population size–area relationships were consistently positive across all archipelagos. Island area was the most important predictor in all models. A significant effect of elevation was only observed in Samoa, where population size declined with elevation. Agriculture might be still playing an important role in Samoa's economy. However, this archipelago is also partitioned into two distinct political units, Samoa and American Samoa [56], perhaps influencing population patterns.

Archipelagos could be categorized into three groups depending on their relationship with the nearest mainland. First, for three archipelagos, no landmass was detected within a 500 km radius (Seychelles, Samoa and Hawaii). These archipelagos are all constituted of highly isolated oceanic islands. A different approach to isolation might be necessary for these island groups, perhaps by employing alternative isolation metrics. Additionally, technological advancements that reduce travelling times to islands may constrain isolation effects.

A single nearest mainland was detected for three archipelagos. The effect of isolation was significant in all cases, in contrast with global analyses. This suggests that perhaps humans do not always escape this biogeographic constraint. A negative relationship was detected for Northern New Zealand, a trend often associated with reduced immigration rates with increasing distance from the mainland [12]. Conversely, in the Virgin Islands and Zhoushan, population sizes increased with isolation. Positive relationships with isolation are consistent with the density compensation hypothesis, which posits that an island can host only a certain number of individuals irrespective of species richness [62]. If isolated islands have fewer species, then more individuals will be present per each species [62]. On the other hand, historical contingencies might play a role. Zhoushan historically had a predominantly maritime economy, characterized by fishing and piracy [63]. After suppressing piracy, Zhoushan became an important commercial port [64,65]. In this context, islands distant from mainland authorities and with easy access to the sea might have been preferred for settlement, promoting a positive population size–isolation relationship in this archipelago.

Two different nearest mainlands were identified for the last four archipelagos. In the Kuril Islands, population size was unrelated to either mainland. This archipelago was disputed by the Japanese and the Russians for over a century [66]. However, the Kuril Islands became part of Russia at the end of the last worldwide war. After that, mass deportation and replacement of entire island populations altered settlement patterns. Today's political circumstances might have overridden an initial two-way colonization process, southwards from Russia and northwards from Japan.

In three archipelagos, population size was higher on islands closer to either source pool, consistently with the two-way filtering hypothesis [44]. In the Channel Islands, population size declined with distance both from mainland Europe (first nearest mainland) and Great Britain (second nearest mainland). This archipelago was disputed between the French and the English for centuries and was acquired by the latter in 1259 [67]. It has been a British crown dependency ever since [68], which might also explain why isolation from Great Britain had a stronger effect on population size than the distance from France. In addition, islands closer to England were less exposed to potential invasions from continental Europe [68].

Population sizes in the Ionian Islands declined with distance from Greece (first nearest mainland), but also from Italy (second nearest mainland). While these islands are geographically closer to Greece, and politically part of it, they remained under Venetian rule for over 400 years between 1386 and 1797 [69]. In any case, these islands were either controlled by political powers located in mainland Greece or Italy for most of their known history [69], perhaps explaining the two-way negative effect of isolation.

Population sizes in the Shetland Islands declined with distance both from Great Britain in the south (first nearest mainland) and Norway in the east (second nearest mainland). Larger populations on southeastern islands might reflect two different immigration waves, one from the Scandinavian peninsula in the ninth century, and a second from Scotland, which began in the thirteenth century [70]. Alternatively, southern and eastern locations are, respectively, more hospitable and less exposed to oceanic currents [71]. Interplays between ecological gradients and historical contingencies would be a fruitful avenue for future research.

(c) . Avenues for future research and conclusions

Overall, cultural, economic and historical contingencies seem to explain differences in how humans inhabit distinct archipelagos across the globe. However, other factors might be at play. Colonizing new islands entails adapting to new ecological conditions, often resulting in niche shifts and life-history changes in colonist populations [8,72]. In particular, type, level and resource structure can shape island occupancy and relationships with other human populations [73]. For instance, the reliable presence of large amounts of resources might determine a stronger tendency towards territoriality, while, on the other hand, inconsistently sparse resources might generate population dispersion [73]. Investigating relationships between resource structure and human occupancy might help explain non-random within archipelago human distributions. Such investigations should also take into account external resources (tourism, subsidies, etc.), on which many of the studied archipelagos are often reliant.

Archipelagos were chosen for this study because they encompass a wide range of climatic, environmental and geographical conditions. Although this resulted in a total sample size of 486 islands, many more remain uninvestigated across the globe. Additionally, human inhabitation of islands has changed through time [13,72,74–76]. In the last century alone, many insular populations have changed dramatically in size worldwide [60,77,78]. However, this process was not homogeneous and while population sizes greatly increased on some islands, on others this process was less pronounced or even reversed [42,60]. As such, investigating historical differences in island occupancy would be an interesting avenue of future research. Finally, along with population size several other estimates of human inhabitation can be measured, such as dwellings, docks and farms [79,80], which can potentially yield interesting information on how humans occupy and use islands.

Overall results illustrate that, similar to other species, insular distributional patterns of human populations are strongly structured geographically [11,12]. However, trends varied idiosyncratically among many archipelagos and appear to be driven by site-specific cultural differences, resulting from social, political and historical factors [10,35,72]. Therefore, the island biogeography of human population sizes might differ fundamentally from that of other species.

Data accessibility

The data are provided in the electronic supplementary material [81].

Authors' contributions

F.M.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, visualization, writing—original draft, writing—review and editing; K.C.B.: conceptualization, supervision, writing—review and editing.

Both 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

Funding was awarded to F.M. by the Victoria University of Wellington through a Doctoral Research Scholarship.