Introduction

Habitat fragmentation plays a fundamental role both in shaping biological communities and in intraspecific diversification. As barriers to migration increase, or as population sizes decrease, the risk of overall extinction of a taxon increases (Hanski 1989, Hanski & Gilpin 1991, Reed 2004). Moreover, the isolating matrix itself may facilitate colonization from outside the system, by species better able to survive in the new conditions created (Cacabelos et al. 2022, Hobbs 2001). However, in some situations, the isolation created by fragmentation can lead to species undergoing local adaptation (White et al. 2022), and potentially even speciation when fragments merge after longer time periods (Flantua et al. 2019). The latter situation is a key premise in the theory that much of the evolution and diversification of life has been driven by the coevolution of species with their ever-changing ecological communities (the mosaic theory of coevolution; Thompson 2005), and the concept of constantly changing (evolving) metacommunities (Urban & Skelly 2006). However, because fragmentation changes the diversity and composition of entire biological communities, establishing very different selective regimes in the different fragments (Kuli-Révész et al. 2021), there are multiple feedbacks that affect the interplay of colonization and extinction versus adaptation and speciation (De Meester et al. 2016). Only through study at the scale of entire communities can we start to develop insights into the role of fragmentation in restructuring local communities on one hand and differentiation on the other (Leibold et al. 2022).
Such large-scale analyses have long been limited by the sheer taxonomic diversity of studied communities. Recent developments in high throughput DNA barcoding now allow for the rapid inventory of entire communities at multiple levels, from taxonomic composition to intraspecific genetic variation (Arribas et al. 2022). This opens up unprecedented opportunities to study community assembly in response to environmental differences and habitat isolation (Emerson et al. 2022). These techniques have been particularly useful in the study of arthropods, the most species-rich animal taxon. Though arthropods are compelling to study because of their ecological diversity and their centrality to the food web of most terrestrial ecosystems, their overwhelming taxonomic diversity makes the manual classification of community samples impractical and means that this taxon is particularly underdescribed (Emerson et al. 2016).
The current study couples a unique natural fragmentation experiment on the Hawaiian Islands with a novel bulk community metabarcoding approach to infer how entire arthropod communities change across discrete spatial scales in a fragmented landscape, both in terms of species composition and intraspecific differentiation. The Hawaiian Islands are ideal for examining the effects of isolation: The archipelago itself is the most isolated in the world, which means that there were very few successful colonists (Wagner and Funk 1995). Within the archipelago, each island is isolated, with many of the endemic lineages showing a pattern of progression that follows the geological chronology of the islands from older to younger (Shaw & Gillespie 2016). Within an island, isolation is created by separation of volcanoes and, over the most recent time scales, new lava flows that can carve up the existing forests (Vandergast and Gillespie 2004, Vandergast et al. 2004). In the latter case, lava flows from volcanic eruptions lead to isolation of forest fragments for extended time periods, with forest regrowth on lava substrate taking anywhere from 300 to 3,000 years depending on the abiotic environment (Drake & Mueller-Dombois 1993, Clarkson 1998). The early history of each island has been characterized by perpetual cycles of such dramatic fragmentation (Carson et al. 1990). The effects of these cycles can be observed in real time on the youngest island of Hawaiʻi, where lava from active volcanoes has led to multiple episodes of forest fragmentation. This pattern is striking on the slopes of Mauna Loa Volcano, where flows have repeatedly covered the landscape over the last 400,000 years, creating a patchwork of habitats in different successional stages (Carson 1995). Here, mature forests have often been surrounded by new flows, resulting in patches of forested “islands” (kīpuka) of various sizes and distances from each other and from the continuous forest. The contrast between forest and lava is stark, with the forest patches being continuously wet and cool, while the surrounding lava is dry and hot during the day and cold at night. Hawaiian forest arthropods are well known for their narrow climatic niches (Hiller et al. 2019, Lim et al. 2022) and hence expected to be strongly isolated by the inhospitable lava (Vandergast et al. 2004, Vandergast & Gillespie 2004).
A particularly well-studied kīpuka system is located along the northwestern slopes of Mauna Loa, where a volcanic eruption in 1855 has left a network of multiple forest patches separated by barren lava. Studies in this location have provided evidence that kīpuka differ in community composition (Knowlton et al. 2017, Mueller 2015), and that arthropods that are specialized for the wet forest are largely isolated in a kīpuka, with the isolation sufficient to allow populations in different kīpuka to demonstrate genetic differentiation (Carson & Sato 1969, Vandergast et al. 2004).
It has long been speculated that a kīpuka-like landscape dynamic, fostering repeated isolation events, may serve as a “crucible of evolution” (Carson 1990). According to this hypothesis, the isolation of populations by lava flows has played a key role in fueling adaptive radiation on the Hawaiian Islands. At the same time, it provides a natural experiment for testing how fragmentation affects the species composition of habitat patches of different size and isolation; not only are kīpuka sometimes very small, so unlikely to sustain a high diversity of forest-specialists over time. This can be exacerbated by a substantial “edge” of low native species diversity that penetrates an estimated 20 m-deep into kīpuka forest (Vandergast & Gillespie 2004). The declines in species diversity towards kīpuka edges and with declining kīpuka size can make these fragments more prone to colonization from non-indigenous species (Mueller 2015).
Here, we use the Mauna Loa kīpuka system to examine how entire communities of arthropods have responded to habitat fragmentation over a 170 year period. We use a vegetation beating protocol (Lim et al. 2022) to sample forest undergrowth arthropod communities at the core and edge of 13 kīpuka of varying sizes and distances to each other. As a comparison to the kīpuka, we sampled two transects in continuous forests on the eastern and western slopes of Mauna Loa. We applied a semiquantitative metabarcoding protocol, which allowed us to study communities from the scale of taxonomic diversity to the scale of intraspecific genetic differentiation. We hypothesize that
1) Due to the pronounced climatic specialization of Hawaiian forest arthropods, lava will have presented a strong barrier to their dispersal, resulting in a community-wide species-area effect. Specifically, we expect smaller kīpuka to have lower taxonomic and genetic diversity. We expect to see higher community turnover over similar distances between kīpuka than between continuous forest sites. For the same reason, species that are limited to the forest core of a kīpuka will, on average, show more evidence for genetic differentiation than species sampled across similar distances in continuous forest.
2) Given the environmental differences between forest at the core of the kīpuka compared to the edges (Vandergast & Gillespie 2004), we expect markedly different arthropod species composition between them. In individual kīpuka, we expect forest-specialized taxa to dominate core habitats. Meanwhile, we expect that fewer forest species will persist in kīpuka edges, and that both edge habitats and adjacent lava matrix will have been colonized by higher proportions of non-native taxa than kīpuka cores or native forest habitats. Consequently, kīpuka should host distinct communities in their cores versus their edges, excepting in the case of the smallest kīpuka, which may behave almost entirely as edges.
3) The species-area effect of reduced species diversity with decreasing kīpuka size should provide less biotic resistance, resulting in an inverse relationship between kīpuka size and the proportion of non-native taxa detected in both core and edge habitats.