Discussion
Using a global dataset of phenotypic differences and differences in neutral genetic diversity for 1832 populations of 112 species studied comparatively in marine island and mainland systems, we showed greater differences in phenotypic traits between islands than between equivalent populations on the mainland, and no differences in the spatial patterns of neutral genetic diversity between the two systems.
As expected, mean phenotypic differences were higher between island populations than between mainland populations. On the mainland, more populations are likely to benefit from higher connectivity between habitat patches compared to islands (Pușcaș et al. 2008, Driscoll et al. 2013, Martín-Queller et al. 2017), which could lower the magnitude of spatial phenotypic variability. In island systems, the effective isolation due to the saltwater matrix and its consequences e.g., lowered gene flow, can amplify opportunities for phenotypic differentiation between populations, which has been linked to accelerated rates of speciation and high levels of island endemism (Whittaker and Fernández-Palacios 2006, Kier et al. 2009). The result could also be due to the potentially larger differences in population sizes between islands than between mainland populations. On islands, smaller population sizes are more frequent than on the mainland due to constraints of island size (Woolfit and Brohman 2005, Triantis et al. 2010). As a result, genetic drift is more frequent on islands (Woolfit and Brohman (2005), which can set populations on distinct evolutionary courses and enhance their phenotypic differentiation. Evolutionary pressure promoted by niche differentiation following colonisation of islands with different natural history may also underly the stronger phenotypic differentiation between islands compared to mainland populations (O’Connell et al. 2019). Due to our modelling framework we could not derive to what extent the phenotypic differentiation was due to genetic differentiation, because our genetic diversity metric quantified differences in neutral genetic diversity between populations. However, the demonstrated genetic differentiation between islands and mainland sites (review in Stuessy et al. 2014) strongly suggest that genetic differentiation may underly the accentuated phenotypic differences between individual islands compared to mainland systems. The effect of island system on phenotypic differentiation emerged despite us analysing oceanic islands together with continental islands. Continental islands have a different history (they are often closer to the mainland, benefiting from more frequent immigration opportunities that stabilize the selection on phenotypic traits) that may have lowered to some extent the effect of the island system type.
In line with our expectations we found a tendency for increased mean phenotypic differences between populations with increasing geographic distance on the mainland, but contrary to our expectations we did not find a similar trend in island systems, and we found no effect of macroclimatic distance on the phenotypic differences. While geographic distance and macroclimatic distance were correlated in our data, which is frequently the case in spatial analyses (Bahn and McGill 2007, Coutts et al. 2016), models excluding either the geographic or macroclimatic distance did not change the results. The lack of any geographic distance effects on phenotypic differences between islands reinforce that other spatial constraints as detailed above (isolation due to saltwater, island size, niche differentiation etc.) may be more effective at promoting phenotypic variability in island systems compared to the simple isolation by distance. On the mainland on the other hand, the signal, albeit weak, of a positive effect of geographic distance on phenotypes suggests that isolation by distance may play a relatively more important role in emerging spatial trait variability compared to island systems (e.g., De Vriendt et al. 2017). The lack of macroclimate effects in both systems suggests no effect of isolation by macroclimate in driving mean population-level phenotypic variation. However, evidence exists for the contrary at least for particular groups of organisms (e.g. in endotherm, but not in ectotherm vertebrates, mean temperatures were associated with smaller intraspecific body size globally; Henry et al. 2023). Therefore, the role of macroclimate in generating isolation is likely idiosyncratic in terms of the taxonomic groups it affects, and in contrast to geographic forces (spatial habitat structure, geographic distance) its effects on spatial phenotypic variability are harder to generalise. However, as sites for island-mainland population comparisons are primarily not selected to test variation determined by environmental differences, we suspect that in our dataset the macroclimatic distance between populations was too small, as the most frequent paired distance represented only 2% of the largest potential environmental distance found in our data (Fig. S3.1). Finally, macroclimate represents only one dimension of environmental distances between populations, while other environmental variables that more directly capture the environments experienced by the populations, such as the heterogeneity of vegetation types, could be potentially more influential on the measured phenotypic traits.
There was no effect of the system (island or mainland) on differences between population-level neutral genetic diversity, except when the geographic distance was omitted from the model. This is surprising, because we expected greater variation in neutral genetic diversity between islands beyond the effect of geographic distance due to e.g., disproportionate dispersal difficulties when traversing larger saltwater barriers, or the hypothesised larger differences in population sizes between different islands compared to populations on the mainland. The geographic and macroclimatic distances potentially underlying the variation in neutral genetic diversity had no system-dependent effects either, because the interaction between these variables and the system type was not significant. Other relevant factors for neutral genetic diversity not tested here such as effective population size or population dynamics and stability could still differ between island and mainland systems. Nevertheless, none of the potentially involved factors caused consistent between-population differences in neutral genetic diversity in island versus mainland systems in our study. Consequently, the spatial patterns of neutral genetic diversity are driven, at least partially, by different mechanisms compared to the spatial patterns of phenotypic traits, which are clearly governed by forces that differ between island and mainland systems (Whittaker and Fernández-Palacios 2006, Santos et al. 2016). These results support earlier findings showing similar levels of population neutral genetic diversity in island and mainland systems (García-Verdugo et al. 2015, De Kort et al. 2021) and also provide support for the universality of neutral processes across systems.
In line with our expectations we captured a weak signal of a positive influence of geographic distance on the mean differences in neutral genetic diversity between populations, which was similar across islands and mainland populations. While spatial isolation is typically a much stronger driver of genetic differentiation between populations due to limits to dispersal and genetic drift (Sexton et al. 2014), the effects of geographic distance on spatial patterns of neutral genetic diversity seem globally weak and may be more heavily influenced by organismal life histories combined with environmental conditions, as advanced earlier by Orsini et al. (2013) and Lira-Noriega and Manthey (2014). Nevertheless, geographic distance may still determine parallel patterns of neutral genetic diversity in both island and mainland system, despite the responses being overall weak.
The effects of environmental distance can override the effects of geographic distance on differences in neutral genetic diversity between populations (Lira-Noriega and Manthey 2014), but in our study, contrary to our expectations, macroclimate had no such effect in either system. As with the phenotypic differences, contrasting climatic requirements of different groups of species may make it difficult to distil generalisations over the course of global approaches. Extending the sampling design of island-mainland studies to evaluate responses across larger environmental gradients may be needed to strengthen signals of global macroclimate effects on neutral genetic diversity, as also suggested by the range of results in Lira-Noriega and Manthey (2014).
Reconciling island biogeography theories with complementary ecological and evolutionary theories has a high priority in the future agenda of island biology (Patiño et al. 2017). Our findings suggest that comparative tests of general isolation-by-distance and isolation-by-environment expectations in island and mainland systems, on populations of the same species, offer promise in achieving such a reconciliation. In such global comparative analyses, there is an outstanding amount of unexplained variability (e.g. 40-70% random species effect in De Kort et al. 2021). This was also the case for our dataset, with effects due to system type or variation associated with factors such as taxonomic diversity only capturing a small fraction of the variation between populations. While macroecological studies spanning global scales and across kingdoms, such as ours, typically have large levels of unexplained variation, they are expected to uncover fundamental spatial phenomena with large effect sizes. We found only relatively small effects of geographic distance in both island and mainland systems, indicating that the effect of this simple isolation measure, commonly used to explain between-population variation, is difficult to generalise or not as universal as previously thought. Context dependencies associated with different life histories, such as dispersal ability of particular species, biotic interactions, variable population sizes, specific colonisation and isolation histories etc. (De Kort et al. 2021) may weaken the effects of geographic distance on phenotypes and genetic diversity. For example, in our island systems the effect of geographic distance might have been lowered because we analysed oceanic islands together with continental islands that benefit more from the spatio-temporal connectivity with the mainland. We also expected stronger effects of kingdom (plant or animal), species or the phylogenetic relationships between species as in, e.g. Sexton et al. (2014), who found differences between plant and animal genetic responses to geographic and environmental distances. Because this was not the case in our dataset, we suspect that the low sample size (e.g., only seven plants in the phenotypic differentiation models) and the large range of traits each more or less responsive to geographic distance and/or correlated with each other to different extents might have blurred the general patterns. The magnitude of the differences between populations varied largely depending on the response type, but the amount of data available for particular response types was generally low, with the exception of body size for the phenotypic traits and heterozygosity for the neutral genetic diversity (Fig. S2.2). We therefore call for a careful investigation of context-dependent drivers of population variability across fundamentally different geographic systems, commensurate with the life history of particular organisms. Comparative functional connectivity studies are a promising avenue in this direction (Juhász and Oborny 2020, Kimberley et al. 2021).
The knowledge transfer between island and mainland systems is still limited. In a horizon scan of the state-of-the-art of island biogeography by Patiño et al. 2017, only 10.2% of respondents worked in both system types. We echo earlier calls (Patiño et al. 2017) for a better replication of the control mainland populations, which may massively improve the applicability of island biology studies in developing global biogeography models. Alternatively, studies could investigate spatial isolation mechanisms comparatively across different types of mainland systems ranging from ecological islands to continuous habitats, while perhaps benefiting from larger datasets.