A case for incorporating phylogeography and landscape genetics into species distribution modelling approaches to improve climate adaptation and conservation planning


  • Jolene Scoble,

    1. Australian Centre for Evolutionary Biology and Biodiversity, School of Earth and Environmental Science University of Adelaide, North Terrace, SA 5005, Australia
    2. CSIRO, Climate Adaptation Flagship, Glen Osmond, SA 5064, Australia
    Search for more papers by this author
  • Andrew John Lowe

    Corresponding author
    1. Australian Centre for Evolutionary Biology and Biodiversity, School of Earth and Environmental Science University of Adelaide, North Terrace, SA 5005, Australia
    2. State Herbarium of South Australia, Science Resources Centre, Department for Environment and Heritage, Adelaide, SA 5005, Australia
      Correspondence: Andrew John Lowe, Australian Centre for Evolutionary Biology and Biodiversity, School of Earth and Environmental Science University of Adelaide, North Terrace, SA 5005, Australia.
      E-mail: andrew.lowe@adelaide.edu.au
    Search for more papers by this author

Correspondence: Andrew John Lowe, Australian Centre for Evolutionary Biology and Biodiversity, School of Earth and Environmental Science University of Adelaide, North Terrace, SA 5005, Australia.
E-mail: andrew.lowe@adelaide.edu.au


Aim  We seek to demonstrate that whilst information derived from phylogeographic and landscape genetic approaches has been used broadly to further ecological and evolutionary hypothesis testing, it can also be used to further species modelling approaches, particularly where bioclimatic and demographic methodologies are to be combined to tackle climate change adaptation and conservation planning.

Location  General application.

Methods  We start with a review of species distribution modelling studies that have used data derived from molecular marker studies, and identify which parameters can realistically be derived from molecular marker studies for inclusion in species and ecosystem distribution prediction and conservation planning.

Results  We find that the uptake of phylogeographic and landscape genetic methods to inform species distribution modelling studies has to date been limited (particularly the latter approaches), despite offering clear potential to improve species modelling approaches that aim to combine climatic envelope and demographic parameters. Using a series of cases studies, we demonstrate that phylogeographic approaches can be particularly useful for identifying biogeographic barriers and refugia, testing alternative demographic models, identifying concordant demographic patterns between species within a single ecosystem and testing temporal niche conservatism. We also find that landscape genetic approaches are particularly useful for quantifying landscape permeability and source/sink dynamics of meta-populations and identifying adaptive variation in the landscape. A summary of parameters that are derivable from such studies for modelling and conservation applications is provided.

Main conclusions  Molecular marker methods have much to offer species distribution modelling, particularly in the field of climate adaptation. Molecular information can inform on species historical dynamics and contemporary demography necessary to advance species modelling paradigms that seek to integrate climatic and demographic drivers. Furthermore, recognizing diversity below species level and incorporating this information into modelling frameworks will enable conservation managers to plan for the capture of areas of evolutionary potential.


Recent and future predictions of climatic changes have important ramifications for the abundance, range, phenology and physiology of a substantial number of species (Hughes, 2000; Root et al., 2003; IPCC, 2007). In recognition of these issues, biodiversity conservation planning has shifted direction to align with new knowledge and community awareness about anthropogenically forced climate change (Rouget et al., 2006; Dunlop & Brown, 2008; Vandergast et al., 2008; Vos et al., 2008). However, central to such a shift in ethos is the safeguarding of species distributions and the evolutionary processes maintaining species abundances and adaptational capacity (Klein et al., 2009), the uptake of which has, to date, been much slower for conservation planning and on-ground adaptation (Dunlop & Brown, 2008).

It also has been a difficult challenge for biologists to develop robust tools to develop biodiversity conservation planning for climate adaptation. Species distribution modelling, based predominantly on climate parameters, continues to lead the way (Beaumont & Hughes, 2002; Pearson & Dawson, 2003; Franklin, 2010). New advances in the field of species modelling and prediction have refined these tools by incorporating more sophisticated species parameters, such as meta-population demography and landscape interactions (Opdam & Wascher, 2004; Keith et al., 2008; Vos et al., 2008), species life history traits (Poyry et al., 2008), species interactions (Araujo & Luoto, 2007) and a consideration of evolutionary history (Rouget et al., 2003; Byrne, 2008; Byrne et al., 2008).

In parallel, but to date in a largely separate literature, there have been significant advances in the application of molecular markers to understand population dynamics and historical demography (e.g. review by Sunnucks & Taylor, 2008). We believe that the lack of integration between these fields is potentially hindering progress to develop robust species distribution modelling approaches, particularly those which wish to progress beyond simple climatic envelope approaches and incorporate demographic processes.

In this article, we review recent literature to determine whether developments in molecular marker tools have been used frequently in species distribution modelling. We then outline, with case studies, the range of species and population parameters that can be derived from different molecular marker studies (mainly in the fields of phylogeography and landscape genetics, but also in the recently expanding area of adaptive variation screening). Finally, we attempt to give clear guidance on the most appropriate parameters, derivable from molecular marker approaches, for species distribution modelling and conservation planning, with a particular focus on climate adaptation.

The use of molecular marker studies in species modelling approaches

A species’ genome cannot only be used to inform on the impact of historical and contemporary demography and connectivity, but it also contains the raw material for species to adapt to future challenges (Lacy, 1987; Riddle et al., 2008). Species persistence under a changing climate is integrally tied to meta-population cohesion and adaptational capacity, which has also been shaped by historic environmental change and impacts on demography (Waits et al., 1998; Riddle et al., 2008). A number of molecular marker methods are available for such studies, and their appropriate application depends on the temporal and spatial scale of the study. A number of reviews have outlined in detail the appropriate application of a range of different marker types to phylogeographic, landscape genetic and adaptive gene studies (e.g. Lowe et al., 2004; Sunnucks & Taylor, 2008), which are summarized briefly in Table 1 and covered in more detail in the subsequent sections later.

Table 1.   Overview of demographic parameters that can be derived from molecular marker approaches. For each parameter, the molecular marker approach, most suitable molecular marker method/required information, modelling application and relevance to conservation is summarized.
Demographic parameterMolecular approach to inform actionsMarkers/information requiredScaleModelling application?Relevance to conservation
Location of long-term barriers to movementPhylogeographyOrganelle and nuclear DNA, appropriate paleoclimateRegional/ContinentalDistributional modellingWill mediate a species’ ability to track a shifting climatic niche or reach refugia
Location of former barriers to movement/hybrid zonesPhylogeographyOrganelle and nuclear DNA, appropriate paleoclimateRegional/ContinentalDistributional modellingHybrid zones are often locations of novel genetic combinations that may facilitate evolution
Location of long-term refugiaPhylogeographyOrganelle and nuclear DNA, appropriate paleoclimateRegional/ContinentalDistributional modellingHarbour high levels of genetic diversity that help prevent extinction and may facilitate evolution
Confirmation of niche conservatismPhylogeographyOrganelle and nuclear DNA, distribution of species and associated climatic variablesRegional/ContinentalDistributional modellingAn important assumption to test when modelling species distribution
Which parts of the landscape matrix facilitate or prevent gene flowLandscape geneticsMicrosatellites or AFLPs, landscape and climatic dataLandscapePopulation demographic modellingGene flow is essential to maintaining strong meta-population structure
Habitat characteristics associated with source and sink populationsSource–sink frameworkMicrosatellites or AFLPs, landscape and climatic dataLandscape/LocalPopulation demographic modellingIdentify what habitat characteristics required for self-sustaining populations
Location of areas with heightened adaptive potentialPhylogeography/Landscape geneticsOrganelle and nuclear DNA, microsatellites and AFLPsRegional/ContinentalPopulation demographic modellingPromote in situ adaptation to changing conditions
Location of adaptive loci in a species meta-populationLandscape geneticsMicrosatellites, AFLPs or candidate gene markersLandscapePopulation demographic modellingIdentify adaptive potential of different populations within a given species

Whilst recent developments in molecular marker approaches could have tremendous application for species modelling studies, in our experience, there has not been broad integration across these fields. To test this notion, we reviewed modelling studies, published between 2008 and July 2009, to investigate whether molecular marker data had been incorporated into a range of species distribution and demographic modelling studies (Fig. 1 and Appendix S1 in Supporting Information). It is clear from this literature survey that where they have been used, molecular marker methods have mainly been applied to understand and model species distributions under past climates (sometimes in lieu of a fossil record; e.g. Byrne, 2008). However, studies considering more contemporary distributions rarely utilized DNA information, and never did so when employing a bioclimatic model (Fig. 1; e.g. Keith et al., 2008).

Figure 1.

 Modelled species distributions under past and future climates. Publications (2008–July 2009 only) were sourced within the Zoological Records’ database using the topic search: ‘climate change AND model*’ and ‘range OR distribution’. Future distributions are divided into those generated using bioclimatic models (‘Modelled’) and those inferred on the basis of empirical studies and/or expert knowledge (‘Inferred’). Past distributions were all bioclimatically modelled and are divided on the basis of temporal scale; those considering distributions <500 years ago (Recent) and those considering distributions >500 years ago (but generally investigating the last glacial maxima; Ancient). Black bars indicate the proportion of studies (number of studies indicated within the bars) that did not utilize information derived form molecular marker studies and white bars those that did. For a full listing of references used and their classification under this system, see Appendix S1 in Supporting Information.

It appears therefore that whilst phylogeographic parameters have been partially recognized for their potential to offer an independent data source for modelling historical species distributions, the use of contemporary population dynamic data derived from molecular marker studies has been much more limited.

Recent modelling studies have questioned the relative importance of climate in relation to species demography (Mustin et al., 2007; Davis et al., 1998; Pearson & Dawson, 2003; Araujo & Luoto, 2007) and have highlighted the critical importance of incorporating life history attributes into modelling approaches to produce more realistic estimates of changed species distributions (Araujo & Luoto, 2007; Keith et al., 2008). Thus, we strongly support the combined use of historical and contemporary population dynamic information derived from molecular marker studies in species distribution modelling approaches to help advance this developing field.

In the rest of this article we outline, with case studies, how approaches in the fields of phylogeography and landscape genetics can be applied to species modelling and planning for climate adaptation and conservation. In particular, we consider the application of phylogeographic approaches to the identification of biogeographic barriers and refugia, the testing of alternative demographic models, identification of concordant demographic patterns between species of a single ecosystem and the testing of temporal niche conservatism. For landscape genetic approaches, we consider the quantification of landscape permeability and source/sink dynamics of meta-populations and the identification of adaptive variation in the landscape. A summary of parameters that are derivable from such studies for modelling and conservation applications is provided in Table 1.

Lessons from the past: consulting the temporal record to predict the future

Phylogeography seeks to interpret patterns of inter and intra-specific genetic diversity in a combined phylogenetic and geographical framework, to understand historic population demography and structure (Avise et al., 1987). Phylogeographic analyses enable the identification of historic barriers to dispersal and refugia (locations that maintain environmental suitability for species during environmental change; Avise et al., 1987; Moritz et al., 1987; Zink & Barrowclough, 2008; Joseph & Omland, 2009). Such analyses generally focus on organellar variation but more recently nuclear-encoded variation has been used to examine timings of coalescence, gene flow and changes in population size across multiple loci (Edwards & Beerli, 2000; Zink & Barrowclough, 2008; Table 1).

Phylogeography has most often been applied to explore the impact of Quaternary climate oscillations on the distribution and resilience of species (Martinez-Meyer et al., 2004; Carstens & Richards, 2007; Bhagwat & Willis, 2008; Kearns et al., 2009). Of particular interest is the impact of recent glacial cycles, where global ecosystems experienced a substantial reduction in temperature and widespread aridity, resulting in desertification in some areas (White, 2006) and the formation of land ice in others (Bhagwat & Willis, 2008; Stewart et al., 2010). Such major historical ecosystem upheavals have had a lasting impact on the genetic variation resident within species, and phylogeographic methods can be used to identify signatures of historical genetic discontinuities or location of species refugia within contemporary populations.

Identifying barriers from a long-term, evolutionary standpoint

The Eyrean barrier in southern Australia is an example of important Quaternary aridity barrier, which in some cases still acts as a barrier to migration today. Such barriers are commonly believed to be implicit in biogeographic patterning (Ford, 1974, 1987), although in some cases molecular evidence is wanting (Joseph & Omland, 2009). By identifying barriers such as these in the landscape, two objectives are possible. First, by identifying long-term barriers to migration, it is possible to more realistically model species response to contemporary climate change. Suitable climatic space will be irrelevant if barriers to migration prevent individuals from reaching such areas. Secondly, by identifying suture zones across former barriers (areas of hybridization between formally separated populations), it is possible to identify likely sites of future evolutionary potential. This application was considered by Vandergast et al. (2008), who observed ‘divergence hotspots’ for 21 southern Californian taxa were most often associated with suture zones.

Taking shelter from the storm: locating refugia

Refugia are locations where species may persist through periods of stress, including climatic extremes (Mackey et al., 2002). Some refugia are important over long time scales and may be identified through phylogeographic studies. Long-term refugia may be defined as those that species populate for at least an entire glacial-interglacial cycle (Stewart & Dalen, 2008; Stewart et al., 2010) and are generally locations, which harbour high levels of allelic diversity.

One of the first spatially explicit approaches to identifying refugia was used by Hugall et al. (2002), who combined molecular phylogeography and paleoclimatological modelling to locate Pleistocene refugia by projecting a species’ environmental envelope onto paleoclimatic surfaces. In this study, paleoclimatological modelling predictions of refugial location and temporal change in the distribution of suitable habitat were found to be consistent with phylogeographic analyses for the land snail, Gnarosophia bellendenkerensis.

Statistical phylogeography has recently been developed to facilitate the identification of causal processes in biogeography and offers a robust method to identify both barriers and refugia. The method involves examination of competing phylogeographic hypotheses of a species’ history generated from a priori models of population structure and statistically assessing the stochastic expectations of each against patterns of observed genetic variation (Knowles & Maddison, 2002; Knowles, 2004). Richards et al. (2007) used this method to assess whether contemporary, isolated ‘sky island’ populations of the flightless montane grasshopper (Melanoplus marshalli) were colonized from proximate refugia as opposed to a single ancestral population (the null hypothesis) (Knowles et al., 2007). When compared to model outputs of the two hypotheses, genetic data most closely fitted a multiple refugial scenario.

Highlighting community patterns

Testing alternative refugial scenarios and assessing what processes have shaped community structure under historical climates can be explored through comparative phylogeographic analysis of multiple species (Bermingham & Moritz, 1998). Comparative phylogeography is a particularly valuable approach because it enables the identification of long-term barriers and refugia common to groups of species and is consequently highly relevant to conservation planning. Carstens & Richards (2007) considered alternative refugial scenarios for four contemporarily co-distributed species in the Pacific Northwest mesic forest ecosystem of North America. Whilst the contemporary distribution of all species is similar, paleodistributions were markedly different and were then used for generating phylogeographic hypotheses. A model of population history was defined for each species, considering the order and timing of population divergence. Results indicated that the amphibian lineages (Ascaphus montanus/Ascaphus truei and Plethodon idahoensis/Plethodon vandykei) had responded as a cohort and have been restricted to two refugia, within the Cascades and northern Rocky Mountains. However, the water vole (Microtus richardsoni) and willow, (Salix melanopsis), which exhibit similar genetic patterns, had very different refugial patterns; the vole restricted to the Cascades, and the willow to the northern Rocky Mountains (Carstens & Richards, 2007).

The ability to test niche conservatism, a fundamental assumption of climate-niche modelling

The output of bioclimatic models has been called into questions by some because of concerns over the validity of underlying assumptions (e.g. Araujo et al., 2005a,b). One fundamental assumption is that a species’ niche remains constant over time. By characterizing the ecological niche of contemporary populations using known distribution points and correlated ecological data, there is an assumption that the species in question will inhabit the same ‘ecological space’ for the projected time period. Hence, an important step in assessing the validity of bioclimatic modelling approaches should be to gain an estimate of the accuracy of predictions of spatial distribution under a changed climatic scenario (Pearson & Dawson, 2003). Evaluation can be achieved if long-term temporal datasets exist (Araujo et al., 2005a,b), or by looking back in time using the fossil record.

In a study of 23 extant North American mammal species Martinez-Meyer et al. (2004) employed hindcasting approaches, verified by a fossil record, to test the accuracy of ecological niche conservatism. Ecological niche envelopes were developed for all species using the Genetic Algorithm for Rule-set Prediction (GARP). Model predictions (based on contemporary occurrence data) were tested by projecting the modelled ecological niche onto Pleistocene paleoclimates, and statistically assessing the location of fossil remains in relation to the niche predictions (Martinez-Meyer et al., 2004). Likewise, an ecological niche based on species distribution from the Pleistocene (using fossil data) was used to project present-day distributions and compared to contemporary occurrence points. Of the 23 species considered, predictions for nine species corresponded significantly in both temporal directions, a further three species were significant using a Pleistocene-derived niche to model contemporary distributions, and five species significant when hindcasting. Whilst this is strong evidence for temporal niche conservatism in North American mammals, the study also demonstrated the necessity of testing for niche conservatism, since five species demonstrated no predictivity of modelled niche in either direction.

The use of fossils to test ecological niche conservatism is a powerful approach; unfortunately, the fossil record may be rare or non-existent for some species groups or locations. In such cases, niche conservatism can be tested using the distribution of molecular variation in contemporary populations as an indicator of historical range. Peterson & Nyari (2008) used GARP to test for ecological niche conservatism in the thrush-like mourner (Schiffornis turdina) complex in the neotropics. The modelled niche of seven genetic clusters (or phylogroups) identified within the species complex was initially tested for predictivity both across space and by lineage membership. Once predictivity within contemporary climatic conditions was demonstrated, the ‘best-subsets models’ were projected onto paleoclimate characteristic of the last glacial maximum (LGM), to test the ability of climatic refugia to explain the distribution of molecular phylogroups. Using this approach, climatic refugia were able to predict the geographic structure of phylogroups better than by chance alone, confirming ecological niche conservatism for this species complex.

The integration of a statistical framework into a phylogeographic approach allows a rigorous assessment of alternative distributional scenarios during different time periods and offers the ability to identify the biogeographical processes operating both at the level of individual species and comparatively across a given community. Understanding refugial patterns and how species adapted and evolved in response previous climatic extremes provides an important insight for future changes in climate.

My, how things have changed: human influence on the landscape and biota

Expanding urban centres and agricultural practices are decreasing the quality, connectivity and amount of native habitat available for many species (Hannah et al., 2002; Opdam et al., 2003; Opdam & Wascher, 2004; Foley et al., 2005; Vos et al., 2008). Recent landscape ecological approaches (Hanski, 2004) no longer recognize elements of the landscape outside native habitat patches as uniform, but instead as a heterogeneous matrix in which some elements can enhance connectivity and gene flow between increasingly fragmented populations (Ricketts, 2001). Empirical studies (e.g. Brooker & Brooker, 2002) have demonstrated that it is the landscape matrix that most impacts dispersal and persistence of spatially structured meta-populations. Consequently, the whole landscape may be considered a ‘functional template for biodiversity’ (i.e. its environment), and it is at the spatial scale of the landscape that anthropogenic land-use changes occur and dominate (Opdam et al., 2003).

The synergistic effect of habitat fragmentation and climate change puts increasing pressure on species persistence, but maintenance of strong meta-population structure, via gene flow, will enable species to recolonize following disturbance events (Opdam & Wascher, 2004). The value of landscape connectivity in enabling species to track changes to their environmental envelope (Hannah, 2008) is now being increasingly recognized (Bradshaw & Holzapfel, 2006). Yet modelling studies assessing a species distribution under climate change often assume that the habitat is homogeneous (Opdam & Wascher, 2004). Furthermore, species characteristics that dictate its ability to respond to changes in habitat distribution, are rarely considered, and include; mobility, dispersal characteristics and source/sink dynamics within a meta-population structure.

A new approach to understanding the relative effect of mosaic attributes on meta-population dynamics comes from the integration of landscape ecology and population genetics, landscape genetics (Manel et al., 2003). Landscape genetics seeks to understand how microevolutionary processes, such as gene flow, genetic drift and selection operate within the landscape mosaic to structure populations, without the a priori identification of populations (Manel et al., 2003; Storfer et al., 2007). In particular, landscape genetics seeks to understand how landscape features have shaped genetic variation, but focuses on contemporary events and at a smaller geographic scale than phylogeography (Joseph & Omland, 2009). A range of molecular marker types can be used for such studies (e.g. Sunnucks & Taylor, 2008; Lowe et al., 2004; Table 1).

Whilst comprehensive research agendas (Storfer et al., 2007) and potential future directions (Holderegger & Wagner, 2008; Balkenhol et al., 2009) have been published for landscape genetics, the impressive number of studies assessing genetic connectivity using this framework (Balkenhol et al., 2009) is rarely used to inform species modelling approaches. We outline here the potential utility of these approaches for informing species demographic planning decisions under future climate change scenarios.

Identifying aspects of the landscape matrix that facilitate and disrupt gene flow

Assessing the permeability of the matrix between habitat remnants is probably one of the most common applications within the field of landscape genetics. Amongst a vast array of options (Storfer et al., 2007), the most popular approach has been to use least-cost path (LCP) modelling. LCPs differ from Euclidian geographic distances by calculating effective distances between habitat patches based on the species-specific cost of moving through particular landscape features, as well as considering behavioural characteristics of the species in question (Adriaensen et al., 2003). In a study of the American marten (Martes americana), least-cost distances, based on individuals avoiding known dispersal barriers such as logging sites and undeveloped forest stands, better explained pairwise-genetic distances than did Euclidean geographic distances (Broquet et al., 2006). Likewise, studies on the European roe deer found that pairwise population genetic distances were best explained by least-cost distance (Coulon et al., 2004). Furthermore, it was shown that two populations had likely diverged because of several high resistance barriers (e.g. roads) that reduced roe deer movements [and hence gene flow (Coulon et al., 2006)].

Barriers to connectivity and gene flow may not always be as obvious as roads or areas of deforestation. Environmental gradients, such as temperature gradients (Dionne et al., 2008) or snow melt gradients (Hirao & Kudo, 2004), can restrict gene flow similar to more obvious barriers but are of particular interest when likely to alter under climate change. Information generated by LCP on likely barriers and dispersal routes can also be used to inform the design of corridors enabling short and long distance movement (Epps et al., 2007).

Novel methods to assess connectivity

LCP is somewhat limited in its ability to accurately model gene flow in real populations (McRae & Beier, 2007; Murphy et al., 2008), rather gene flow is likely to occur via numerous direct pathways through the landscape and using indirect stepping stone movements over more than one generation (Slatkin, 1993; McRae & Beier, 2007). Therefore, understanding dispersal pathways is better served by methods that consider multiple pathways, such as isolation by resistance (IBR). Distance metrics based in electrical circuit theory have provided the basis for IBR, based on analogous characteristics between conductivity and gene flow (McRae, 2006). Increasing the number and width of conductive pathways increase conductivity much in the same way that gene flow might increase should the same principles be applied to landscape connectivity (McRae, 2006; McRae & Beier, 2007). IBR represents an improvement on LCP methods because of its strong theoretical foundation described by the connection between the timing of gene coalescence in a given meta-population and measures of effective resistance in a landscape matrix (McRae, 2006). The effectiveness of this approach was assessed for central American populations of the big-leaf mahogany (Swientenia macrophylla) and North American populations of the wolverine (Gulo gulo) (McRae & Beier, 2007). IBR analysis outperformed both LCP and isolation by distance approaches in the ability to explain gene flow within each species meta-population because of a capacity to consider range shape, as well as multiple pathways and individual widths simultaneously.

The potential of a landscape genetic approach could also be greatly enhanced with the development of analyses that can investigate the effect of multiple environmental variables and interactions within a single population genetic model (Storfer et al., 2007). To date, analyses have most often relied on simple Mantel tests to integrate landscape information with population genetics (Storfer et al., 2007; Pavlacky et al., 2009), despite the likely violation of the stringent assumptions of the test (Bossart & Prowell, 1998), and non-independence of genetic similarity measures between populations (Yang, 2004). In an attempt to develop a more robust landscape genetic framework, Pavlacky et al. (2009) employed a linear mixed model (Yang, 2004), to estimate appropriate sampling variance and distinguish between contemporary and historic landscape factors influencing genetic connectivity of the logrunner (Orthonyx temminckii). This approach demonstrated that whilst spatial heterogeneity for rainforest and sclerophyll woodland elements enhanced dispersal historically, contemporary migration rates were strongly disrupted by anthropogenic deforestation (Pavlacky et al., 2009). The linear mixed model provides a mechanistic measurement of logrunner dispersal through different landscape features, allowing an improvement in precision compared with LCP, e.g. (Council, 2004; Broquet et al., 2006; Coulon et al., 2006) that rely on hypothetical migration routes (Pavlacky et al., 2009).

Characterizing meta-population source–sink dynamics

Within a spatially structured population, migration through the landscape matrix facilitates a balance of local births, deaths, immigration and emigration that influence population density at a given point in time (Hanski, 2004). Such populations can be characterized within a source–sink framework (Pulliam, 1988), where populations are defined on the basis of net emigration (sources) or net immigration (sinks). This framework allows the demographic and habitat features associated with robust or alternatively, unstable populations, to be identified. This approach was employed to understand the impact of old-growth forest harvesting on the population viability of a threatened seabird, the Marbled Murrelet (Brachyramphus marmoratus) (Peery et al., 2008). The smallest Murrelet population is located in central California, United States of America, and is considered a sink because of its isolation from other populations, low birth rate and immigration estimates from mark-recapture studies (Peery et al., 2006). To test this hypothesis, the number of parent–offspring dyads expected under both a closed (self-sustaining) and sink population was modelled given a suite of expected demographic characteristics. Demographic models were compared to estimates derived from a random sample of individuals from the central Californian population, identifying the number of parent–offspring dyads using multilocus genetic profiles. The estimated number of parent–offspring dyads was best explained by modelled sink populations experiencing c. 2–6% immigration annually, and hence the null hypothesis of a self-sustaining population could be rejected. Interestingly, an earlier mark-recapture study reported immigration rates of 16% (Peery et al., 2006), suggesting that in some situations migrants are selected against, or do not permanently relocate to the central Californian population (Peery et al., 2008).

Investigating a meta-population under a source–sink framework can also extend to characterizing the effect of the intervening matrix on dispersal between populations, as well as consideration of different spatial scales. Historical population demography and contemporary dispersal and structuring were examined using molecular marker methods on the common European toad, Bufo bufo (Martinez-Solano & Gonzalez, 2008). Separate phylogeographic clades were identified in Morocco and the Iberian Peninsula, but neither exhibited within region geographical structuring, indicating gene flow throughout the two regions. Analysis of heterozygosity values indicated that whilst most populations are self-sustaining, some exhibited low levels of heterozygote excess, indicative of ‘sinks’, and assignment tests could be used to identify the probable sources. The influence of landscape variables on dispersal permeability was assessed within the central Spanish populations. Results indicated that gene flow both within and between ecosystems was similar, demonstrating that riverine habitats in this region do not play an important role in facilitating dispersal. Conversely, higher altitude populations exhibited increased levels of genetic differentiation indicating that altitude might impede dispersal in this species.

Beyond neutral gene flow: identifying the potential for evolution and adaptation

Whilst many species may undergo changes in their distributional range in response to environmental change, others will adapt to changing conditions. In particular, steep environmental gradients are often associated with adaptive variation, reflected in high levels of genetic divergence across the gradient (Moritz, 2002; Vandergast et al., 2008), and are increasingly recognized as relevant to preserving evolutionary potential. For example, in the Subtropical Thicket Biome of South Africa (Rouget et al., 2006), regional conservation corridors have been designed around major rivers to capture environmental gradients. Amongst the biodiversity features considered were biome interfaces, as well as topographical and macroclimatic gradients. Klein et al. (2009) likewise considered both ecological and evolutionary refugia in their recent conservation plan for Australia. They sought to identify ‘evolutionary refugia’; places in the environment that facilitated species persistence and radiation during periods of high climatic and/or environmental stress (Morton et al., 1995).

When considering individual species, capturing recent, local adaptation has been greatly enhanced by the recent increased availability and use of adaptive gene makers (Gebremedhin et al., 2009). Bonin et al. (2007) developed a population adaptive index (PAI) that identifies outlier loci, which are candidates for selective action and exhibit increased differentiation compared to neutral population parameters such as gene flow (Neilsen, 2005; Storz, 2005). Bonin et al. (2007) applied this concept to the common frog (Rana temporaria) and the Austrian dragonhead (Dracocephalum austriacum). For both species, measures of neutral and adaptive variation did not coincide, indicating that protecting neutral genetic diversity does not necessarily mean that current adaptive potential is also preserved. However, it is worth noting that adaptive loci identified by methods such as PAI may not necessarily represent adaptive potential in the future under novel threats (Bonin et al., 2007; Gebremedhin et al., 2009). It is likely that adaptive loci will complement neutral genetic diversity that currently represents our best measurable way of identifying evolutionary potential at the intra-specific level. However, the techniques to screen such adaptive variation are still developing. At present, the incorporation of adaptive gene information into modelling studies may allow a useful understanding of the distribution of adaptive variation associated with a single strong selective force. However, to progress modelling that combines a broader range of active and latent adaptation will require much more work both from the genomics and modelling fields.


Molecular marker methods offer a unique chance to step back in time to assess historical rangewide and demographic changes and understand links with past environmental change. By incorporating such information into species distribution modelling approaches, we should be able to improve our future predictions of species distribution and demographic trajectories.

Of particular relevance to species distribution modelling approaches, molecular marker studies can identify biogeographic barriers that may still limit species movement, and hence impact on future responses to climate change. Alternatively, former barriers that are now hybrid zones between previously isolated populations may harbour important evolutionary potential and warrant protection. Historical refugia can likewise be identified, the location of which can be used to independently verify hindcasting species distribution modelling. Statistical phylogeography provides a rigorous framework for testing alternative hypotheses of species and community response to biogeographical processes associated with historical climatic extremes. Furthermore, by examining the concept of niche conservatism, it is possible to test a fundamental assumption of bioclimatic models.

In addition to advancing understanding of historical range change dynamics, molecular marker approaches can also further understanding of contemporary population demography and its relationship to recent land-use changes. Human activities, through habitat fragmentation, introduction of invasives and climate forcing are altering the composition and distribution of natural ecosystems. It is therefore imperative that species modelling approaches consider the influence of the contemporary landscape matrix on evolutionary processes and consequent meta-population structure. Molecular approaches such as landscape genetics, which can identify contemporary gene flow barriers and source–sink dynamics, are well placed to robustly inform modelled species distributions under future climate change.


We thank Leo Joseph and Anita Smyth at CSIRO Sustainable Ecosystems and Climate Adaptation Flagship, Peter Cale and Jaco Le Roux for comments on the manuscript, and J.S. acknowledges the support of a CSIRO Climate Change Flagship postgraduate top up grant for this work.


Jolene Scoble is currently completing her PhD in genetic analysis of the southern scrub-robin to elucidate metapopulation structure, and understand historical and contemporary dispersal across ecotones between intact and relictual mallee vegetation in South Australia. Andrew Lowe leads a research group that apply molecular markers, landscape analysis and genomic assessments of adaptive genes, to demonstrate gene flow and selection pressure changes across a range of landscapes; contemporary, historical, fragmented and exploited.

Author contributions: A.J.L. and J.S. conceived the ideas; J.S. searched and collated the literature; and J.S. and A.J.L. shared the writing.

Editor: David Richardson