Negotiating a Fragmented World: What Do We Know, How Do We Know It, and Where Do We Go from Here?

Abstract

Genetic diversity determines evolutionary potential. Without a variable genome, natural selection cannot act. Habitat fragmentation is the single largest threat to global biodiversity, as it reduces or eliminates gene flow among populations, thereby increasing the erosion of genetic diversity through random genetic drift. The loss of adaptive capacity in small, isolated populations is irreversible without gene flow and the ensuing genetic rescue. Without habitat connectivity, populations cannot expand or contract into refugia, an increasingly vital capacity under climate change. Here, I review what we have learned from organisms found in naturally fragmented landscapes. Metapopulation theory has played a seminal role in this goal. However, extending this theory to anthropogenically fragmented habitats has been a challenge. Single-species approaches cannot elucidate the impacts of habitat fragmentation on entire communities, composed of species with diverse interactions—mutualisms, facilitations and predator–prey dynamics—and proper ecosystem functioning. To overcome the limitation of single-species studies, metacommunity and metaecosystem ideas have emerged. The spatial extent and configuration of habitat patches will determine which species remain in altered landscapes. Changes to species interactions, community structure and ecosystem processes will follow. Ecosystem function determines ecosystem viability, and losses of keystone or foundation species will have cascading effects. Genomic tools can track the effect of landscape changes on population and movement dynamics, the maintenance of genetic resources and the persistence probabilities of individual species in the context of the communities in which they are embedded. Landscape genetics combines landscape features and population genetics to quantify how species use diverse landscapes and is now a powerful tool to assess the causes and consequences of habitat fragmentation for interacting species in fragmented ecosystems.

1. Introduction

Evolutionary forces, including mutation, natural selection, genetic drift and gene flow, have shaped Earth’s biodiversity. Gene flow among populations spreads favorable mutations and provides connectivity among populations, which acts to maintain genetic diversity and large effective population sizes, thereby facilitating adaptive potential and counteracting the forces of random genetic drift [1,2,3,4]. Habitat fragmentation represents a major extinction threat for species across the globe, as it disrupts connectivity and impedes gene flow [5,6,7,8,9,10,11]. Fragmentation, loss and degradation impact habitats in every major biome and species of all major taxonomic groups [6,7,12]. Rapid global climate change exacerbates the already dire implications of habitat loss and fragmentation and further reduces the probability that viable refugia will exist [13,14,15,16,17]. Evolutionary potential will depend upon the genetic variants at adaptive trait loci as well as the adaptive landscape, where the ability to respond to environmental variation will determine a population’s fitness peak [18,19,20,21,22,23]. Population isolation and reduced genetic resources will ultimately constrain the ability of species to mount an adaptive response to rapidly changing environments under climate change.

The unprecedented human population growth beginning with the industrial revolution has placed humankind in competition for space and resources with all other organisms that share the Earth. Sheer habitat loss due to human population growth and encroachment accounts for staggering losses of biodiversity [24,25,26,27,28,29]. Between 1970 and 2020, declines in species found in terrestrial, freshwater and marine ecosystems have occurred on a global scale, with the relative abundance of monitored wild vertebrate populations decreasing by 73 percent [30]. The largest declines have occurred in Latin America, which has experienced a 95 percent decrease [30,31]. The extinction vortex is upon us.

On an individual species level, isolated habitat fragments not only result in reductions in genetic diversity but also expose species to novel predators, parasites and disease through edge effects, impinge on species’ ability to fulfill their life history and impair proper ecosystem functioning [32,33,34,35,36,37,38,39,40]. Patch size, number and level of connectivity will determine habitat viability for species, communities and whole ecosystems. How this plays out in landscapes will determine what genetic resources are available for adaptation to a changing environment, as species losses can have cascading effects on ecosystem function and ecosystem resilience [41,42,43].

The challenge we face then is piecing together patch networks within habitat types that provide the critical components necessary for species and community persistence and evolutionary potential in the face of the ever-shrinking intersection of fundamental, potential and realized niche space (Figure 1).

2. Habitat Spatial Structure and the Metapopulation Concept

Conservation biology emerged as a distinct scientific discipline with the formation of the Society for Conservation Biology in 1985 and the publication of Conservation Biology, the Society’s flagship journal, in 1987. Habitat fragmentation was quickly recognized as a global conservation concern, and there is now vast literature on this topic (see [5,9,11] for a comprehensive review). Early research focused largely on population viability analysis for species whose habitat was greatly reduced, giving rise to the minimum viable population idea [44,45,46,47]. The metapopulation concept also garnered the attention of the conservation biology community as a theoretic construct for potentially understanding how organisms might navigate fragmented habitats and providing guidance for protected reserve design (e.g., [48,49,50,51,52,53,54,55]).

2.1. Metapopulation Theory

Metapopulation theory describes a group of distinct, semi-independent and interacting populations where the persistence probabilities of individual subpopulations within the network are contingent upon a balance between population extinction and recolonization from extant populations [56,57,58,59,60]. Levins’ original model assumed a large number of habitat patches but did not specify the patch size or spatial configuration [56,57]. Harrison [61] and Harrison and Taylor [62] expanded on this original model to include five types of metapopulation configurations—Levins’ classic metapopulation, mainland–island, patchy population, non-equilibrium and an intermediate case, which combines aspects of all the models (Figure 2). Mainland–island scenarios consist of a large mainland habitat and multiple smaller habitat patches that support individuals but due to their size experience greater demographic and environmental stochasticity and are therefore more susceptible to local extirpation events [63]. The mainland acts as the source of colonists for the extirpated patches, and the persistence probability of the entire metapopulation depends upon the longevity of the mainland population. In a patchy population, although the habitat patches are spatially structured, movement among occupied patches is frequent enough that local extirpations are rare. The patch network is actually a pseudo-metapopulation in that the frequent movement among patches promotes consistent occupancy and panmixia [62]. Intermediate cases involve both mainland–island dynamics and a patchy population dynamic. The spatial arrangement of patches promotes frequent dispersal among proximate patches that form a type of mainland with the more distant patches acting as islands in a mainland–island sense. Non-equilibrium metapopulations are declining populations headed toward extinction, as extirpated patches do not get recolonized [61,62]. The utility of the metapopulation approach in a conservation framework will lie in its applicability to real-world population dynamics [60,64,65,66,67,68].

2.2. Tests of Metapopulation Theory

Tests of metapopulation theory have largely been conducted on single species but have included a wide variety of taxa from invertebrate, vertebrate and plant groups (e.g., invertebrates [69,70,71,72,73,74,75,76]; protists [77]; mammals [65,78,79,80,81]; plants [82,83,84,85]; birds [86,87]; fish [88,89,90,91]).

2.2.1. Classic Metapopulations

Among the first tests of metapopulation theory were Hanski’s work on the Glanville fritillary butterfly (Melitaea cinxia) [92,93,94,95] and Sjögren’s on the northern pool frog (Rana lessonae) [96,97,98]. Both species live in habitats with a Levins’ classic metapopulation structure, and both species are now range-restricted. Early papers on these species focused on extinction–colonization dynamics and network persistence probabilities but revealed complexity in movement dynamics associated with patch size, spatial configuration, the quality of the intervening habitat matrix and species’ life history [92,96,97,98,99,100]. This early work showed that the dispersal dynamics of the Glanville fritillary butterfly varied depending upon patch size, resource quality and butterfly density (Figure 3). Larger patch sizes, high within-patch density and flower abundance decreased emigration, while smaller patch sizes and lower densities with surrounding large and open landscapes increased immigration [99]. Ecological tradeoffs in dispersal behavior were evident, however, as dispersing females paid an energetic cost and had lower lifetime reproductive success [99]. Despite the importance of dispersal to metapopulation persistence, the reduced fitness of dispersing females could lead to selection against dispersal phenotypes [101,102]. Hanski [25] went on to show that the extinction–colonization metapopulation dynamics in the Åland Glanville fritillary network influenced allele frequency changes in the phosphoglucose isomerase (Pgi) gene, revealing strong associations between genetic variation at the Pgi locus and dispersal, recolonization and local population dynamics.

For the northern pool frog, pond permanence determined occupancy, with ephemeral ponds representing dispersal sinks [98]. The spatial configuration of ponds was also important in both the persistence and recolonization probabilities. Ponds at critical threshold distances from the nearest neighboring ponds had higher extirpation probabilities due to isolation, increased susceptibility to predation and the combined effects of demographic and environmental stochasticities [98]. The dispersal direction from occupied ponds was also non-random in that frogs tended to move toward neighboring ponds across hospitable marshland habitat, highlighting the importance of the landscape matrix surrounding occupiable habitat in facilitating dispersal success [98].

The studies on the northern pond frogs and Glanville fritillary butterfly show that despite the classic metapopulation habitat structure, few naturally occurring populations will likely fit the assumptions of Levins’ classic model. Not all habitat patches are equal, even those of equal size, and the extinction/recolonization dynamic will be determined by the interplay of life history, patch size and quality, spatial configuration and matrix characteristics [92,97,98,99,100,101]. However, a classic metapopulation dynamic is rare and largely limited to insect species, small mammals, weedy plants and amphibians in ephemeral habitats often at range margins with low persistence probabilities [103,104,105,106,107].

2.2.2. Mainland–Island Metapopulations

Most naturally occurring populations found in fragmented habitats have a mainland–island spatial structure or are intermediate cases with aspects of both mainland–island and patchy population dynamics [e.g., bay checkerspot butterflies (Euphydryas Editha Bayensis) [108]; North American pika (Ochotona princeps) [78]; tree hyrax (Dendrohyrax arboreus), blue duiker (Philantomba monticola) and samango monkey (Cercopithecus mitis labiatus) [109]; black bears (Ursus americanus) [110]; crested guineafowl (Guttera edouardi) [111]; micro-arthropod metacommunity, [112]; water voles (Arvicola amphibious) [113]; root voles (Microtus oeconomus) [114]; etc.]. The persistence probabilities for mainland–island metapopulations are largely determined by mainland extent but also by patch network characteristics including the patch number, size and spatial configuration that determine the individual patch residence time and recolonization potential [109,111,114,115,116,117].

Dispersal capacity will make a pronounced difference in connectivity among patches regardless of the metapopulation type. An increased dispersal capability and the ability to track resources facilitated colonization potential and residence times for the root vole in mainland–island habitats, where individuals were able to recolonize even remote small habitat patches of high quality [114]. Dispersing juvenile nuthatches, Sitta europaea, traveled farther in fragmented habitats but once settled in a fragmented patch tended to stay, as compared to more fluid movement dynamics in non-fragmented forest, thus potentially increasing patch isolation [118]. For less mobile species, such as Cunningham’s skink (Egernia cunninghami), habitat fragmentation reduced movement, increased the population genetic structure and increased the levels of relatedness in individual patches [119]. For the bay checkerspot butterfly, currently listed as threatened under the United States Endangered Species Act (ESA), which has seen precipitous declines over the last century due to anthropogenic disturbance, large mainland populations have proved key to the persistence of the remaining butterfly populations. Now found primarily in the San Mateo and Santa Clara regions of California, the butterfly is restricted to patches of serpentine grassland with the majority of individuals found in a single mainland–island habitat in Santa Clara County (Morgan Hill; Figure 4) [108,115]. Among the smaller patches, patch residence times were correlated with patch size, landscape features such as topography and patch-level resource abundance, as well as proximity to the mainland, the source of colonists to repopulate extirpated patches [115]. However, the negative impact of climate change on metapopulation persistence, even those with large mainlands, is evident, as seen in the bay checkerspot butterfly [115]. Disruption in the evolved asynchronies under increasing ambient temperatures between insect and plant host phenology, i.e., selection on fecundity to complete development within the lifespans of ephemeral hosts, is resulting in increased larval mortality in the bay checkerspot butterfly and population extirpations of even large habitat patches [120]. Low dispersal capacity results in the low recolonization potential of the more distant and smaller patches for this species, and increased larval mortality reduces the likelihood that even if colonized these patches will remain viable.

2.2.3. When Is Fragmented Habitat Fragmented? A Species Perspective

How species found in the same spatially structured habitat utilize that landscape varies widely across taxa [109,114,118,119,121,122]. In an example from a mainland–island habitat in fragmented Podocarpus forests of the KwaZulu-Natal midlands, South Africa, in three moderately sized mammal species with different life histories, the tree hyrax (Dendrohyrax arboreus), blue duiker (Philantomba monticola) and samango monkey (Cercopithecus mitis labiatus), habitat use varied considerably. Here, the hyrax and duiker, both primarily solitary species, utilized both the mainland habitat and smaller habitat patches (Figure 5) [109], whereas the smaller patches could not support the social samango monkey, found in groups of 10–40 individuals [109]. Although the duiker and hyrax exhibit a mainland–island metapopulation dynamic consistent with the habitat spatial structure, the social nature and low dispersal capability of the samango monkey resulted in a transient, non-equilibrium and declining metapopulation. Generalized linear modeling results show that the increasing isolation of the samango monkey on mainlands greatly increases extirpation probability [109]. Although mainland–island dynamics for the hyrax and duiker show continued occupancy and recolonization potential across patches of various sizes, they also show a gradual increase in isolation and decrease in patch occupancy [109]. In the same habitat, the Karloof forest complex, the forest-restricted and social crested guineafowl (Guttera edouardi) persist in both mainland and island habitats despite low dispersal rates. Large patch sizes are important for these birds, which have large home ranges and are found in flocks of between 10 and 30 individuals. Most of the large patches supported guineafowl; however, persistence on even the mainland is likely to decline, as continuing habitat fragmentation is increasing isolation across the range of this species [111].

Another example from a study of eight lemur species showed that a fragmented habitat with a mainland–island spatial configuration had strikingly different impacts on individual species ranging from the expression of a mainland–island metapopulation dynamic to patchy population dynamics (Figure 6) [122]. For beetle species, there was considerable variation in how beetles utilized continuous forest, small forest patches with a sedgeland matrix and streamside forest environments in naturally fragmented landscapes in Tasmania, Australia [67]. Despite a mainland–island habitat structure in the fragmented landscape, forty percent of the most commonly captured beetle species did not form metapopulations of any kind due to either a lack of dispersal capability or a dispersal ability greater than the distance among fragments precluding a metapopulation response. Only a small fraction of the species samples showed distributions consistent with a mainland–island metapopulation [123]. Similarly, metapopulation dynamics varied among native mollusk species that inhabit freshwater habitats on two islands of Guadeloupe. Dynamics ranged from stable and fluctuating to declining metapopulations among the native species, while non-native species were increasing ([124]; Figure 7).

Data for multiple butterfly species in fragmented landscapes in Britain and adjoining areas on the European continent showed that dispersal ability was key to maintaining species richness [125]. Butterfly species with intermediate dispersal capabilities experienced the greatest declines, followed by those of low mobility and little to no decline for highly mobile species [125]. The species with low dispersal capacity tended to remain in their natal habitat patch, and long-distance dispersers could move effectively among discrete habitats, but species with intermediate abilities that could not fly between patches would often end up in a matrix habitat that could not support them. A spatially structured habitat in and of itself does engender a metapopulation dynamic for all species, as these studies demonstrate.

2.2.4. Metapopulation Theory and the Real World

While early studies on metapopulation dynamics were informative, much of the empirical research on populations found in fragmented habitats over the past several decades has revealed a more complex dynamic than can be explained by the more restrictive assumptions of the first metapopulation models. Research shows that occupancy dynamics in spatially structured habitats are not solely determined by extinction and recolonization but that movement among extant patches in between extirpation events (rescue effect), the use of patches as stepping stones and age-specific or density-dependence-specific movement among patches are all components of real-world metapopulations (e.g., [78,88,126,127,128,129,130,131,132]). Habitat heterogeneity within and among patches also needs to be sufficient in order to meet age-specific life history requirements [130,133,134,135,136]. The quality of the intervening landscape matrix, regardless of the metapopulation structure, also plays a role in the local and regional persistence of metapopulations, as a more natural and undisturbed matrix facilitates movement [137,138,139,140,141]. Akçakaya et al. [142] then suggested that metapopulations should be defined more broadly and more realistically as geographically discrete populations where movement among populations is less than movement within populations. Additionally, the capacity of landscapes to support metapopulations was addressed by Hanski and Ovaskainen [95] who introduced the idea of a “metapopulation capacity”, which ties species persistence in fragmented landscapes to threshold levels of patch numbers and habitat interconnectedness, determined by both the landscape and the properties of individual species (Figure 8).

2.3. Genetic Implications of Metapopulation Structure

How well do spatially structured populations maintain genetic diversity? The answer lies in the balance between founder effects, genetic drift and gene flow. There will not be a single answer to this question, as the interplay between life histories, landscape and habitat patch characteristics will determine movement dynamics, which in turn will determine the levels of genetic diversity [117,144,145]. Demographic stability will govern the persistence probabilities of spatially structured populations in the short term, but among-patch dynamics will determine the long-term maintenance of genetic variation and evolutionary potential [121,146,147,148]. There is often a disconnect between demography and genetic diversity, as an extinction-and-recolonization process can erode standing genetic variation through the process of drift and genetic coalescence [147,149,150]. Effective population sizes are therefore typically much lower than the census size in populations with a metapopulation dynamic [78,117,151,152,153,154].

Low levels of genetic diversity are typically found in most classic metapopulations due to the extinction-and-recolonization dynamic, which can lead to genetic coalescence [147] and reduced genetic diversity (e.g., northern pool frogs, R. lessonae [96]; Canada yew, Taxus Canadensis [155]; rock hyrax, Heterohyrax brucei and Procavia johnstoni [156]). Even in a large well-connected metapopulation of chalk-hill blue butterflies (Polyommatus coridon) native to the calcareous grasslands of Europe, fitness effects were tied to the levels of genetic diversity, and diversity was directly tied to the dispersal frequency among patches [74]. The more frequent the movement, the higher the levels of diversity. Conversely, a lack of movement led to low levels of within-patch genetic diversity, inbreeding and significant decreases in adult lifetime expectancy [74]. Low genetic diversity for species with a classical metapopulation structure is evident even in highly mobile bird species. For the lesser kestrel (Falco naumanni), individuals born in smaller and more isolated colonies were genetically less diverse, whereas their mobility would suggest greater dispersal distances and more frequent genetic mixing [157]. For the ground-dwelling black grouse (Lyrurus tetrix) found in a spatially structured habitat in the Alps, the core of its distribution in Central Europe, the dispersion of genetic variation among subpopulations, despite the classic metapopulation habitat structure, was explained by isolation by distance among individual birds and landscape-driven isolation by resistance among subpopulations [158]. The levels of genetic variation were high, and subpopulation differentiation was low, except between core and edge populations. Here, the species’ life history did not result in an extinction-and-recolonization dynamic despite the spatially structured habitat.

Large classic metapopulations can retain higher levels of genetic diversity, as seen among the patch networks of the northern Finland Glanville fritillary butterfly, which have retained neutral allelic variation over the past century [159]. In contrast, the smaller Glanville fritillary butterfly networks that had significantly lower genetic diversity have gone extinct [159]. Saccheri et al. [160] showed that the increased extirpation risk across patches in this system varied within networks and was explained by resource limitation (low flower abundance) and concomitant low levels of within-patch heterozygosity that were exacerbated in the smaller, more isolated patches. In a recent study on the Glanville fritillary butterfly that combined genomic data with demographic and dispersal data, the authors found that low genetic diversity was a strong predictor of extinction risk [161]. However, increases in heterozygosity only mattered in the large patches, while extinction probabilities did not decrease with increasing heterozygosity for small patch populations, thus highlighting the key role that demographic stochasticity plays in persistence probabilities for small patch populations.

The impact of spatially structured habitats on genetic diversity can also vary at different spatial scales. Metapopulation dynamics analyzed for six Chaoborus midge species show that species with frequent subpopulation extinctions within local metapopulations had reduced genetic variation, not surprisingly [162]. However, on a regional level, species with a highly volatile local population dynamic dominated by extinction and recolonization had greater overall genetic variation than species with permanent populations [162]. This is likely due to the drift and fixation of allelic variants among local subpopulations. High levels of regional genetic variation, however, will not ameliorate the potential lack of adaptive capacity within local metapopulations with low genetic diversity. The spread of favorable mutations and the underlying genetic architecture of adaptive traits will ultimately determine regional persistence.

Though mainland–island metapopulations harbor more genetic variation than typically seen in classic metapopulations, the tempo of patch turnover will influence the extent of drift and genetic coalescence [150,157,163]. Subpopulations in the smaller habitat patches will depend upon dispersal from the larger mainland for both recolonization and genetic rescue (dispersal among occupied patches). This dynamic is observed in large heterogeneous interconnected watersheds of multiple species of inland trout, where mainstem river migratory and tributary stream resident life histories are expressed (Figure 9) [88,127,164,165,166,167], as well as in forest- and hedgerow-dwelling fire salamanders (Salamandra salamandra) [168], riparian mainland–island root voles (Microtus oeconomus) [114] and raccoons (Procyon lotor) on Virginia barrier islands [169].

Both source–sink and rescue effects were evident for Northern Goshawks (Accipiter gentilis) on the Alexander Archipelago and in coastal British Columbia [170,171]. These birds breed in old growth and mature forests and exhibit aspects of a metapopulation system [170,171]. The directionality of gene flow provides evidence of source populations (coastal mainland British Columbia) as well as sink habitats (Revillagigedo and Vancouver islands) [171]. However, for the remainder of the island assemblage, there was no trend in the directionality of dispersal, which supports ongoing dispersal among extant subpopulations and an active genetic rescue effect.

The role of the mainland habitat in metapopulation persistence and the maintenance of genetic variation is amply illustrated by the dynamics observed in a unique American pika (Ochotona princeps) metapopulation found in the Bodie Hills, a spur of the main Sierra Nevada in California (Figure 10) [15,78,117]. Composed of ~100 habitat patches formed from hard-rock mining ore dumps established in the late 1800s [172], the habitat patch size and configuration have aspects of both classical and mainland–island spatial structures [78,79,117,173]. The southern half of the population (~40 patches) with a classical Levins-type spatial structure went extinct in the 1980s, and despite multiple natural recolonization attempts, the patch network has remained unoccupied since that time (Figure 10) [78,117]. On the other hand, the northern half of the area, dominated by a large patchy population mainland (High Peak) composed of multiple talus patches in close proximity, has remained occupied over the 20th and 21st centuries [78,117,172,173]. Population genetic analyses conducted soon after the initial population collapse provided support for a mainland–island metapopulation dynamic in the remaining extant patches along with stepping-stone dispersal, genetic rescue and heterozygosity similar to a large population found in a continuous talus habitat in the main Sierra Nevada [78,174]. However, genetic analyses of samples from the Bodie site mid-20th century to present day show a steady and significant decline in allelic richness and heterozygosity [117,172,174]. Comparisons between late 20th and early 21st century samples show a decline in effective population size as well as a steep increase in population-level relatedness despite continuous occupancy in the northern patch network [78,117]. Although climate change has been implicated in rendering the southern habitat patches inviable, mainland–island dynamics and dispersal among extant patches appear to account for the continued persistence of the northern patch network [117]. Indeed, in a range-wide comparison of nucleotide diversity, the Bodie population, in the last sampling period (2015–2017), had higher levels of nucleotide diversity than populations found within the Ruby-East Humboldt Mountains in the Great Basin physiographic province in Nevada. The Ruby-East Humboldt range is one of the largest mountain ranges with the greatest extent of montane and alpine habitat found in Nevada that still supports pikas [15]. Climate change, however, is altering the metapopulation dynamic in Bodie. The most recent genetic analysis shows fewer successful dispersal events, greater within-patch relatedness, increased population genetic structure and evidence of genetic coalescence [117]. Additionally, despite persistence in the mainland–island structured habitat, this population is becoming increasingly isolated, as neighboring populations continue to go extinct [175,176].

2.4. Limitations of a Metapopulation Approac

From a biodiversity perspective, however, the metapopulation concept is limited, as it is largely a single-species approach and fragmented habitats house entire food webs [142]. As a result, single-species studies may not provide critical information on the viability of entire biotic communities or shed light on ecosystem-level processes impacted in the face of habitat fragmentation. As global climate change exacerbates species loss in a fragmented world, the emphasis is shifting from a single-species focus to ecosystem function, trophic-level interactions and biotic community viability in relation to the fragment size, number and spatial configuration of habitat patches [177,178,179,180].

3. Food Webs, Biodiversity and Fragmented Habitats

Early discussions in the conservation literature focused on how to ameliorate the effects of fragmented habitats for species and communities and whether conservation action should secure single large patches or provide connectivity among multiple small patches (i.e., SLOSS debate) [181,182,183,184,185]. Multiple empirical studies have shown that networks of numerous small patches support greater species diversity than single large patches (reviewed in [185,186]). Riva and Fahrig [187] reviewed 32 datasets with 603 patches and 2290 taxa (FragSAD dataset compilation; [188]), which all show that biodiversity was higher in systems of multiple small patches versus single large habitats. This is likely due to risk spreading despite smaller populations being more susceptible to demographic, genetic and environmental stochasticity [63] and speaks to the importance of independent or semi-independent population dynamics across patches, an underlying assumption of metapopulation theory [59]. However, these studies do not address how species of various sizes and life histories use fragmented landscapes, which can vary, as seen in the studies on lemurs [122] and mollusks [124], as well as bird and mammal species found in the Karloof forest complex [109,111]. A recent review by Liu et al. [189] suggests that, in addition to biodiversity losses, habitat fragmentation has a non-random effect on ecosystem communities, with species that play a major role in proper ecosystem functioning disappearing first. For example, large-bodied species, species at higher trophic levels with requisite large home ranges and species with narrow niche breadth are often the first to be lost from ecosystems when a habitat is fragmented [190,191,192]. Interactions among species will also have important implications for persistence, the long-term maintenance of within-species genetic variation, food web dynamics and ecosystem function [193,194,195]. The impacts of habitat fragmentation on keystone and foundation species will be especially important to determine, as they play seminal roles in food web complexity, ecosystem integrity and ecosystem resilience.

3.1. Foundation Species

Foundation species are species with the greatest abundance and biomass within ecosystems and as such determine species community composition and biodiversity levels [196,197,198,199]. Examples include tree species such as redwoods (Sequoia sempervirens), mangroves (e.g., Avicennia germinans; Rhizophora mangle), poplars in riparian ecosystems (Populus spp.), sponges or corals, especially reef-forming corals (e.g., Acropora, Montastrea or Porites species), giant kelp (Microcystis pyrifera) and cockles (Austrovenus stutchburyi) [200,201,202,203]. Foundation species form the basis for interconnectedness within ecosystem species networks. Interactions between foundation species and other taxa are typically non-trophic or mutualistic, which includes providing structural support for other species, e.g., trees provide nest sites for birds and mammals, escape cover from predators, etc. [199]. The spatial extent of foundation species’ distributions will have profound effects on biodiversity and ecosystem function [202,204]. Ellison et al. [196,199] showed that declines in foundation tree species disrupt fundamental ecosystem processes, including the rates of decomposition, nutrient fluxes, carbon sequestration and energy flow within the ecosystem but also between connected ecosystems such as streams that flow through forests and depend upon allochthonous inputs from terrestrial environments (Figure 11). Arthropod species richness was associated with genetic diversity and trait variance in foundation poplar tree species, suggesting that habitat fragmentation and the loss of genetic diversity in foundation species will signal the onset of cascading ecological impacts [200]. The blue mussel (Mytilus edulis), also a foundation species, has declined by >60% since the 1970s in the intertidal zone of the Gulf of Maine due to rapidly warming ocean conditions [205]. As a result, the composition of the sessile benthic community was increasingly dominated by algal species with concomitant reductions in species richness in this environment [205]. The habitat fragmentation of landscapes can lead to disruptions in the interaction dynamics between foundation species and ecosystem networks, which if ecosystems decline below functional thresholds will trigger the failure of ecosystem processes and ecosystem integrity [196,199,206].

3.2. Keystone Species

Keystone species have a disproportionate effect on ecosystem function with both direct and indirect effects on the biotic and abiotic components of ecosystems. They are critical players in proper ecosystem functioning and often determine community-level diversity [207,208,209,210]. The loss of keystone species can result in the disruption of coevolved relationships, the loss of ecosystem integrity and, ultimately, ecosystem collapse [199,207,208,211,212,213,214,215,216]. As apex predators, keystone species mediate direct effects through predation [217,218,219] and indirect effects through trophic cascades [220,221,222]. Keystone species also play important roles in mutualistic relationships as plant pollinators [223,224,225], seed dispersers [226,227,228,229] and ecosystem engineers [230,231]. Found at all trophic levels, keystone species include species such as the gopher tortoise (Gopherus polyphemus) and the red-cockaded woodpecker (Leucontopicus borealis) in the longleaf pine (Pinus palustris) coastal plain ecosystem of North America [232], black-tailed prairie dogs (Cynomys ludovicianus) in North American prairie grasslands [233], African forest elephants (Loxodonta cyclotis) [234], the army ant (Eciton burchelli) in neotropical rainforests [235], the slash pine (Pinus elliottii var. densa) in the endangered pine rockland habitat [236], sea otters (Enhydra lutris) in the coastal kelp forests of the Pacific Ocean [213], the gray wolf (Canis lupus) in all ecosystems it occurs in [237,238,239], shark species in marine ecosystems worldwide [240,241,242], etc. With low functional redundancy and unique ecological niches [214], keystone species are disappearing at a faster rate due to specific habitat requirements or large home range sizes, conditions not met under habitat fragmentation, as well as the introduction of non-native species [34,235,243,244].

Habitat fragmentation and loss has led to IUCN and US ESA listings for many keystone species including jaguars (Panthera onca), African forest (L. cyclotis) and Asian elephants (Elephas maximus), sea otters (northern Enhydra lutris kenyoni and southern E. l. nereis), the red wolf (Canis rufus), the Mojave desert tortoise (Gopherus agassizii), the Florida panther (Puma concolor coryi), leopards (Panthera pardus), the San Joaquin kit fox (Vulpes macrotis mutica), ivory tree coral (Oculina varicose), the Franklin bumble bee (Bombus franklini) and multiple yellow-faced bee species (genus Hylaeus). Losses of keystone species can lead to a cascade of secondary extinctions [6,245,246], as seen with sea otter extinction on the Pacific coast of North America, which led to the collapse of the kelp forest communities [247,248]. The loss of the gray wolf led to high population densities of herbivores (moose and elk), which led to significant changes to the vegetation community in the Yellowstone ecosystem, with resulting impacts on habitats for neotropical bird migrants, small mammals and insect species [239,249]. As with other apex predator extirpations, the loss of leopards and African wild dogs from Mozambique’s Gorongosa National Park resulted in the expansion of forest-dwelling antelopes into treeless floodplains and the reduced abundance of a common forage plant utilized by other herbivores [250]. The loss of keystone species can lead to the simplification of food webs and non-equilibrium ecosystem dynamics (Figure 12) [251,252,253].

3.3. Keystone Communities

Mouquet et al. [254] introduced the idea of keystone communities, a subset of communities within larger metacommunities and metaecosystems defined as complex interaction networks at both local and regional scales that form in heterogeneous landscapes [255,256]. Leibold and Norberg [257] used linked limnetic ecosystems (e.g., lake basin wetlands consisting of many ponds, very large lakes with multiple basins and oceans) to advance the idea that metacommunities connected via dispersal act as “complex adaptive systems.” Complex adaptive systems are defined as “(1) sustained diversity and individuality of components, (2) localized interactions among those components and (3) an autonomous process that selects from among those components based upon the results of local interactions” [257]. Dispersal among these communities acts to maintain species diversity, but similar to single-species metapopulation dynamics, there is an assumption of semi-independence among spatially discrete individual communities [257]. Individual community-level species composition and the resulting interspecific interactions are influenced by local environmental conditions in addition to habitat area and proximity to other communities within the metacommunity [258,259]. Resetarits et al. [260] tested the keystone community concept with protist microcosm metacommunities. Although the removal of single patches within the network did not affect the average community richness, evenness or overall biomass, stochastic factors at larger spatial scales influenced the metacommunity assemblage [260]. In another test of the keystone community concept also using protist microcosms as the model system, Yang et al. [261] found that the removal of local communities with unique environmental conditions that supported endemic species had a significant negative effect on regional-scale biodiversity, suggesting they act as regional keystone communities.

3.4. Habitat Fragmentation and Trophic Cascades—The Black-Tailed Prairie Dog

The black-tailed prairie dog is a textbook example of a keystone species and trophic cascades. Black-tailed prairie dogs have been eliminated from much of their historic range through intensive poisoning by the agricultural and ranching communities, leaving their habitat much reduced and highly fragmented [262,263,264,265]. Prairie dogs in fragmented landscapes exhibit a metapopulation dynamic [266,267,268], where the extirpation of habitat patches is driven largely by outbreaks of sylvatic plague (Yersinia pestis) introduced into black-tailed prairie dog colonies in the 1940s [269,270]. As a result, despite its wide historical distribution, the black-tailed prairie dog is currently threatened under the Canadian Species at Risk Act (https://www.canada.ca/en/environment-climate-change/services/environmental-enforcement/acts-regulations/about-species-at-risk-act.html), a Species of Concern in Arizona (https://www.azgfd.com/wildlife-conservation/ (accessed on 8 March 2025)) and Oklahoma (https://www.wildlifedepartment.com/wildlife/threatened-and-endangered (accessed on 8 March 2025)) and endangered in Mexico (Lista de las Especies Amerzadas, https://animapedia.org/nombres-de-animales (accessed on 8 March 2025)). As a keystone species, the black-tailed prairie dog influences population- and community-level dynamics for many other organisms found within the prairie grassland ecosystem (Figure 13) [233]. Multiple studies have shown that prairie dogs, as ecosystem engineers, have both direct (birds and vegetation) and indirect (arthropods) effects on other species, which act to increase biodiversity within prairie dog colonies [271,272,273,274] by providing habitats for more rare and endangered species than comparable non-prairie dog sites [233]. Specific examples include the horned lark (Eremophila alpestris) and mountain plover (Charadrius montanus), which are found in greater abundance at sites with prairie dogs versus without in Wyoming grasslands [274]. Prairie dog colonies in New Mexico had greater abundances for four of thirty-two bird species surveyed than adjoining sites without prairie dogs (mountain plover C. montanus, ferruginous hawk Buteo regalis, burrowing owl Athene cunicularia and curve-billed thrasher Toxostoma curvirostre) [272]. Haun et al. [275] also showed a significant association between black-tailed prairie dog colonies and bird species including horned larks, marbled godwits (Limosa fedoa), Brewer’s blackbirds (Euphagus cyanocephalus), killdeer (Charadrius vociferus) and burrowing owls (A. cunicularia) at sites in the mixed-grass prairie and sagebrush steppe ecosystem of north-central Montana. Among mammalian species, the desert cottontail (Sylvilagus audubonii), the American badger (Taxidea taxus), spotted (Spilogale putorius) and striped (Mephitis mephitis) skunks and swift foxes (Vulpes velox) were also more abundant in prairie dog colonies [233,272,276,277]. Much of this increased biodiversity associated with black-tailed prairie dog colonies is due to the habitat heterogeneity created by prairie dog burrowing life history [278]. Declines in the population numbers of the American badger, an apex predator currently listed as sensitive in California and as threatened/endangered in Canada, are associated with habitat loss and declines in prairie dog numbers, which comprise a large portion of their diet [279,280,281]. As a specialized predator of black-tailed prairie dogs, the black-footed ferret (Mustela nigripes) is critically endangered largely due to the loss of prairie dogs but is also impacted by plague and canine distemper [282,283].

Black-tailed prairie dogs, swift foxes, burrowing owls and mountain plovers are likely an example of a trophic cascade (Parker et al. 2019). All are species of concern at the state and federal level [265,285,286]. Swift foxes and burrowing owls are meso-predators, with swift foxes preying upon burrowing owls and burrowing owls preying upon mountain plovers (Figure 14). Swift foxes also secondarily prey on mountain plovers and prairie dogs, but these are not their primary prey species. Population sizes ebb and flow as numbers within each species increase or decrease in relation to each other such that mountain plover numbers increase with increases in swift fox numbers, as swift foxes prey primarily upon burrowing owls [265]. Declines in swift fox numbers have led to an increase in burrowing owls and concomitant declines in mountain plovers. Hunting, trapping and predation by coyotes also influence swift fox abundances. Coyotes are themselves apex predators and with larger habitat areas required to fulfill their life history are often absent in human-modified landscapes [265,284,287,288]. The size and number of prairie dog colonies underlies the trophic dynamic among the prairie dog, swift fox and burrowing owl meso-predators and mountain plover. Small prairie dog colonies are more susceptible to extirpation by plague outbreaks, and declines in swift fox numbers are tied to the extinction and recolonization rates of colonies among colonies within prairie dog metapopulations [265,289].

In a recent study of the northernmost and isolated population of black-tailed prairie dogs in North America, the effects of isolation are apparent with low genetic diversity, a small effective population size and evidence of inbreeding and genetic bottlenecks [16]. The Canadian population in the Cullingham et al. [16] study has the lowest levels of genetic diversity recorded for this species. However, this population is found in what is considered to be the most viable habitat given climate change impacts. Genetic rescue in the short term and reestablishing connectivity with southern populations in the long term are conservation priorities [16]. As southern landscapes become unsuitable, latitudinal shifts by prairie dogs will necessitate dispersal corridors among patches of natural prairie grassland habitat [264]. Additional genetic studies show that connectivity among viable habitat patches throughout the range is a major concern, as dispersal and gene flow have been disrupted by urbanization, agriculture and the loss of intact native short-grass prairie [264,290]. The loss of genetic connectivity has and will continue to erode the available genetic resources over time [266,268,291].

4. Ecosystem Function and Biodiversity in Fragmented Landscapes

In recent years, the effect of habitat fragmentation on the relationship between biodiversity and ecosystem functioning has gained wider attention [9,10,184,292,293]. The biotic and abiotic components that define ecosystems will ultimately determine ecosystem functionality. Species community composition sets the stage for species interactions, the degree of trophic redundancy and ecosystem resilience [292]. Species within trophic levels are not always ecological equivalents, and this can have significant effects on food web dynamics such that species losses can destabilize ecosystems and decrease the ability of ecosystems to return to an equilibrium state after disturbance [292,293,294]. Examples include the effects of forest loss and fragmentation on Amazonian rainforest food web integrity, where decreased patch size resulted in steep and nonlinear decays in network complexity, the loss of interaction diversity and simplified and dysfunctional food webs (Figure 15) [293]. Small forest patches with greatly reduced food web complexity and no-longer-supported predator species led to the erosion of ecosystem functionality and higher extinction risk [293]. Another example from aquatic ecosystems shows that specific types of physically distinct habitat patches, i.e., mesohabitats—pool, riffle, run and glide—varied across sampled sites; riffle mesohabitats had greater habitat diversity and more fish species. The authors concluded that riffles, which comprised a small percentage of river habitat diversity in their study, act as keystone habitats [295]. Mesohabitat diversity is also important for the support of all age classes in stream-dwelling fish. A population viability analysis of Lahontan cutthroat trout (Oncorhynchus clarki henshawi), listed as threatened under the US ESA and currently found in <90% of its historical habit and now largely isolated in headwater reaches of once large interconnected and heterogeneous watersheds, showed that habitat heterogeneity was directly tied to the effective population size (Peacock and Dochtermann 2012). The predominance of a pool mesotype within fragmented streams reduced the habitat diversity key to younger age classes’ survival, which negatively impacted recruitment [130]. Regolin et al. [296] examined environmental heterogeneity in a spatial context of landscape composition, landscape configuration and habitat spatial structure as determinants of species richness. The results showed that landscape configuration largely explained species richness levels, whereas nestedness within species communities was explained primarily by spatial heterogeneity [296].

Although there is ample evidence that increased environmental heterogeneity has positive effects on biodiversity within ecosystems (e.g., [297,298,299,300]), the universality of this relationship continues to be debated, especially in the context of habitat fragmentation. Multiple reviews of the extensive literature on this topic and meta-analysis shows that intermediate levels of environmental heterogeneity for natural and undisturbed ecosystems resulted in the highest levels of biodiversity, whereas semi-natural environments and environments heavily impacted by anthropogenic disturbance showed a mix of positive and negative relationships between heterogeneity and biodiversity [301,302,303,304]. A spatially explicit model, which incorporated environmental heterogeneity to study the effects of habitat fragmentation on species diversity, showed the same result in that fragmentation had little effect on biodiversity levels when environmental heterogeneity was at intermediate levels [305]. Similarly, Wan et al. [306] found that the strength of environmental heterogeneity effects varied among terrestrial taxa across a wide variety of ecosystems, although the overall relationship between increased environmental heterogeneity and increased levels of biodiversity was positive, especially for tree species, signifying the importance of trees as foundation species. Albrecht et al. [307] showed that increased environmental heterogeneity led to increased biodiversity across multiple ecosystem types on Mt Kilimanjaro, Tanzania, and that increased biodiversity was positively associated with increased ecosystem functioning across multiple measures of ecosystem functionality. The results of such studies covering a wide array of ecosystems and taxa found in natural and relatively undisturbed landscapes support the role of biodiversity in maintaining proper ecosystem functioning, albeit at different spatial scales and different levels of environmental heterogeneity [308,309,310]. Habitat fragments of different sizes will differ in environmental heterogeneity, and different levels of heterogeneity will affect individual species to different degrees, such that community composition within metacommunities will vary across fragmented landscapes and ultimately determine ecosystem function [311,312].

5. Biodiversity and Metapopulation Capacity

Organisms in fragmented landscapes will need to function as metapopulations in metacommunities within metaecosystems in order to persist in an increasingly fragmented world. The diversity of organismal life histories coupled with habitat type, the extent of fragmentation and the composition of the surrounding matrix that organisms have to navigate in order to move among fragments will determine whether organismal communities can sustain themselves in fragmented landscapes. However, to date, there are few studies on landscape metapopulation capacity in the context of entire metacommunities and metaecosystems [313,314].

Examining the metacommunity dynamics of rainforest mammals in fragmented landscapes in Borneo, Brodie et al. [315] showed that landscape links among habitat patches of various sizes and distances defined how connected patch networks were for a diversity of species with different dispersal capabilities within interacting species communities. Large habitat patches that were well connected (spatially tractable) were particularly important, especially for poorly dispersing species. Additionally, well-connected large patches could form a mainland habitat for larger species with greater dispersal ability, thereby reducing metapopulation extinction risk and enhancing metacommunity stability [315]. Further, the study showed that landscape connectedness was important to metacommunity stability across both heterogeneous and homogeneous landscapes, but not surprisingly, the level of landscape connectivity that was important in determining metapopulation capacity varied significantly among species [315].

The metaecosystem concept expands upon the metacommunity concept in that it includes not only biotic but also abiotic components critical to ecosystem functioning that flow across ecosystem boundaries [316]. Metaecosystems involve large landscape level processes and, as such, the movement of organic and inorganic nutrients among ecosystems will play a critical role in metacommunity persistence and ecosystem functionality [316]. The complexity of the interacting pieces of ecosystems—species communities, food web dynamics, habitat diversity, nutrient availability and nutrient cycling—makes it difficult to predict metapopulation capacity and the sustainability of biodiversity across the diverse ecosystems globally given the range of landscape fragmentation scenarios.

6. Landscape Genetics/Genomics and Biodiversity in a Fragmented World: Integration

Landscape ecology is the science of landscape diversity, diversity that forms from the intersection of biodiversity and geodiversity [317]. Landscape genetics combines population genetics and landscape ecology [318,319,320]. Here, population genetic data used in conjunction with data on landscape features can elucidate the determinants of organismal movement and help elucidate impacts on species interaction dynamics in diverse landscapes. A landscape genetics approach combined with demographic and ecological data can give us further insight into how habitat fragmentation affects ecological processes and ecosystem function. This can be accomplished by overlaying patterns of genetic variation onto fragmented habitats, providing insights into impacts on individual species and also revealing the genetic effects of altered interactions among species and ultimately food web dynamics due to species losses [321,322]. Patterns of genetic variation across diverse species and the links among those species will provide information on the viability of ecosystems in fragmented landscapes [323,324].

Now, in the era of genomic sequencing, we have tremendous statistical power to quantify both selectively neutral and adaptive trait loci across entire genomes within and among species [204,325,326]. The large datasets afforded by sequencing technologies allow us a finer-grained resolution of habitat fragmentation impacts, variation in spatial structure and habitat extent on movement dynamics and gene flow, maintenance of genetic variation and effective population sizes for individual species and potentially whole communities [327,328]. Methodological innovations include a shift from population-based to individual-based analyses, analyses based on pairwise relatedness, Bayesian methods and inference from landscape resistance and genotype-by-environment interactions [329,330]. That global biodiversity is now constrained by habitat loss, degradation and fragmentation is unquestioned, but of more importance is the determination of what constitutes the threshold levels of ecosystem functionality. What species, what level of genetic resources and what landscape characteristics inform the persistence probabilities and evolutionary potential for diverse biota across global ecosystems in a fragmented world? The use of genomic tools has the promise to bridge the gap between pattern and process.

7. Conclusions

The burgeoning human population is reshaping the landforms we live on, the atmosphere we breathe and the ecosystems and ecosystem services we depend upon [331,332,333,334,335]. Habitat fragmentation has occurred on a massive and global scale, which continues unabated into the 21st century [9,11]. Now, global climate change is exacerbating the impact of loss, fragmentation and habitat degradation on biodiversity and ecosystem function [336,337]. No corner of Earth is immune to the synergistic effects of these perturbations [338]. We are in a race against time to preclude reaching the threshold at which fundamental ecosystem processes break down and the extinction vortex cascades beyond the point of recovery. Metapopulation, metacommunity and metaecosystem approaches together with landscape genetics/genomics will play key roles going forward in determining at what spatial and temporal scales, across the diversity of global habitats, we can preserve ecosystem functioning together with maximal biodiversity.