Expert-based versus habitat-suitability models to develop resistance surfaces in landscape genetics

Landscape genetics aims to investigate functional connectivity among wild populations by evaluating the impact of landscape features on gene flow. Genetic distances among populations or individuals are generally better explained by least-cost path (LCP) distances derived from resistance surfaces than by simple Euclidean distances. Resistance surfaces reflect the cost for an organism to move through particular landscape elements. However, determining the effects of landscape types on movements is challenging. Because of a general lack of empirical data on movements, resistance surfaces mostly rely on expert knowledge. Habitat-suitability models potentially provide a more objective method to estimate resistance surfaces than expert opinions, but they have rarely been applied in landscape genetics so far. We compared LCP distances based on expert knowledge with LCP distances derived from habitat-suitability models to evaluate their performance in landscape genetics. We related all LCP distances to genetic distances in linear mixed effect models on an empirical data set of wolves (Canis lupus) from Italy. All LCP distances showed highly significant (P ≤ 0.0001) standardized β coefficients and R ² values, but LCPs from habitat-suitability models generally showed higher values than those resulting from expert knowledge. Moreover, all LCP distances better explained genetic distances than Euclidean distances, irrespective of the approaches used. Considering our results, we encourage researchers in landscape genetics to use resistance surfaces based on habitat suitability which performed better than expert-based LCPs in explaining patterns of gene flow and functional connectivity.     

The effect of cost surface parameterization on landscape resistance estimates

A cost or resistance surface is a representation of a landscape's permeability to animal movement or gene flow and is a tool for measuring functional connectivity in landscape ecology and genetics studies. Parameterizing cost surfaces by assigning weights to different landscape elements has been challenging however, because true costs are rarely known; thus, expert opinion is often used to derive relative weights. Assigning weights would be made easier if the sensitivity of different landscape resistance estimates to relative costs was known. We carried out a sensitivity analysis of three methods to parameterize a cost surface and two models of landscape permeability: least cost path and effective resistance. We found two qualitatively different responses to varying cost weights: linear and asymptotic changes. The most sensitive models (i.e. those leading to linear change) were accumulated least cost and effective resistance estimates on a surface coded as resistance (i.e. where high-quality elements were held constant at a low-value, and low-quality elements were varied at higher values). All other cost surface scenarios led to asymptotic change. Developing a cost surface that produces a linear response of landscape resistance estimates to cost weight variation will improve the accuracy of functional connectivity estimates, especially when cost weights are selected through statistical model fitting procedures. On the other hand, for studies where cost weights are unknown and model selection is not being used, methods where resistance estimates vary asymptotically with cost weights may be more appropriate, because of their relative insensitivity to parameterization.     

Ecological resistance surfaces predict fine‐scale genetic differentiation in a terrestrial woodland salamander

Landscape genetics has seen tremendous advances since its introduction, but parameterization and optimization of resistance surfaces still poses significant challenges. Despite increased availability and resolution of spatial data, few studies have integrated empirical data to directly represent ecological processes as genetic resistance surfaces. In our study, we determine the landscape and ecological factors affecting gene flow in the western slimy salamander (Plethodon albagula). We used field data to derive resistance surfaces representing salamander abundance and rate of water loss through combinations of canopy cover, topographic wetness, topographic position, solar exposure and distance from ravine. These ecologically explicit composite surfaces directly represent an ecological process or physiological limitation of our organism. Using generalized linear mixed‐effects models, we optimized resistance surfaces using a nonlinear optimization algorithm to minimize model AIC. We found clear support for the resistance surface representing the rate of water loss experienced by adult salamanders in the summer. Resistance was lowest at intermediate levels of water loss and higher when the rate of water loss was predicted to be low or high. This pattern may arise from the compensatory movement behaviour of salamanders through suboptimal habitat, but also reflects the physiological limitations of salamanders and their sensitivity to extreme environmental conditions. Our study demonstrates that composite representations of ecologically explicit processes can provide novel insight and can better explain genetic differentiation than ecologically implicit landscape resistance surfaces. Additionally, our study underscores the fact that spatial estimates of habitat suitability or abundance may not serve as adequate proxies for describing gene flow, as predicted abundance was a poor predictor of genetic differentiation.     

Landscape characteristics influencing the genetic structure of greater sage‐grouse within the stronghold of their range: a holistic modeling approach

Given the significance of animal dispersal to population dynamics and geographic variability, understanding how dispersal is impacted by landscape patterns has major ecological and conservation importance. Speaking to the importance of dispersal, the use of linear mixed models to compare genetic differentiation with pairwise resistance derived from landscape resistance surfaces has presented new opportunities to disentangle the menagerie of factors behind effective dispersal across a given landscape. Here, we combine these approaches with novel resistance surface parameterization to determine how the distribution of high‐ and low‐quality seasonal habitat and individual landscape components shape patterns of gene flow for the greater sage‐grouse (Centrocercus urophasianus) across Wyoming. We found that pairwise resistance derived from the distribution of low‐quality nesting and winter, but not summer, seasonal habitat had the strongest correlation with genetic differentiation. Although the patterns were not as strong as with habitat distribution, multivariate models with sagebrush cover and landscape ruggedness or forest cover and ruggedness similarly had a much stronger fit with genetic differentiation than an undifferentiated landscape. In most cases, landscape resistance surfaces transformed with 17.33‐km‐diameter moving windows were preferred, suggesting small‐scale differences in habitat were unimportant at this large spatial extent. Despite the emergence of these overall patterns, there were differences in the selection of top models depending on the model selection criteria, suggesting research into the most appropriate criteria for landscape genetics is required. Overall, our results highlight the importance of differences in seasonal habitat preferences to patterns of gene flow and suggest the combination of habitat suitability modeling and linear mixed models with our resistance parameterization is a powerful approach to discerning the effects of landscape on gene flow.     

Landscape genetics of a pollinator longhorn beetle [Typocerus v. velutinus (Olivier)] on a continuous habitat surface

Landscape connectivity, the degree to which the landscape structure facilitates or impedes organismal movement and gene flow, is increasingly important to conservationists and land managers. Metrics for describing the undulating shape of continuous habitat surfaces can expand the usefulness of continuous gradient surfaces that describe habitat and predict the flow of organisms and genes. We adopted a landscape gradient model of habitat and used surface metrics of connectivity to model the genetic continuity between populations of the banded longhorn beetle [Typocerus v. velutinus (Olivier)] collected at 17 sites across a fragmentation gradient in Indiana, USA. We tested the hypothesis that greater habitat connectivity facilitates gene flow between beetle populations against a null model of isolation by distance (IBD). We used next‐generation sequencing to develop 10 polymorphic microsatellite loci and genotype the individual beetles to assess the population genetic structure. Isolation by distance did not explain the population genetic structure. The surface metrics model of habitat connectivity explained the variance in genetic dissimilarities 30 times better than the IBD model. We conclude that surface metrology of habitat maps is a powerful extension of landscape genetics in heterogeneous landscapes.     

Are populations like a circuit? Comparing isolation by resistance to a new coalescent‐based method

A number of methods commonly used in landscape genetics use an analogy to electrical resistance on a network to describe and fit barriers to movement across the landscape using genetic distance data. These are motivated by a mathematical equivalence between electrical resistance between two nodes of a network and the ‘commute time’, which is the mean time for a random walk on that network to leave one node, visit the other, and return. However, genetic data are more accurately modelled by a different quantity, the coalescence time. Here, we describe the differences between resistance distance and coalescence time, and explore the consequences for inference. We implemented a Bayesian method to infer effective movement rates and population sizes under both these models, and found that inference using commute times could produce misleading results in the presence of biased gene flow. We then used forwards‐time simulation with continuous geography to demonstrate that coalescence‐based inference remains more accurate than resistance‐based methods on realistic data, but difficulties highlight the need for methods that explicitly model continuous, heterogeneous geography.     

Density-dependent dispersal suggests a genetic measure of habitat suitability

Recent research shows that density dependence should result in predictable movements between habitats of different suitability, depending on whether population densities are increasing or decreasing. When population densities are increasing, habitats become filled in order of their suitability, resulting in a net flow from high suitability to low suitability. When populations decrease in density, the reverse can happen. These patterns suggest that genetic information can be used to infer habitat suitability since individual-based genetic assignment tests permit high resolution assessments of migration. We used replicated landscapes to study fishers ( Martes pennanti ) during a population increase and predicted that there should be a net flow of individuals from areas of shallow to deep snow, since snow depth has previously been linked to fisher fitness. A total of 769 fishers were sampled from 35 different landscapes and profiled at 16 microsatellite loci. From assignment tests, we inferred five genetic populations. By assigning each of the 35 landscapes to one of these five populations, we were able to determine the proportion of immigrants to each. Consistent with our prediction, there was a positive relationship between the proportion of immigrants and snow depth. The best model of fisher habitat suitability was one with both snow depth and the proportion of coniferous forest in landscapes. Our findings suggest that where population trend is known, genetic information can be used to measure habitat suitability.     

Landscape resistance and habitat combine to provide an optimal model of genetic structure and connectivity at the range margin of a small mammal

We evaluated the effect of habitat and landscape characteristics on the population genetic structure of the white‐footed mouse. We develop a new approach that uses numerical optimization to define a model that combines site differences and landscape resistance to explain the genetic differentiation between mouse populations inhabiting forest patches in southern Québec. We used ecological distance computed from resistance surfaces with Circuitscape to infer the effect of the landscape matrix on gene flow. We calculated site differences using a site index of habitat characteristics. A model that combined site differences and resistance distances explained a high proportion of the variance in genetic differentiation and outperformed models that used geographical distance alone. Urban and agriculture‐related land uses were, respectively, the most and the least resistant landscape features influencing gene flow. Our method detected the effect of rivers and highways as highly resistant linear barriers. The density of grass and shrubs on the ground best explained the variation in the site index of habitat characteristics. Our model indicates that movement of white‐footed mouse in this region is constrained along routes of low resistance. Our approach can generate models that may improve predictions of future northward range expansion of this small mammal.     

Small geographic range but not panmictic: how forests structure the endangered Point Arena mountain beaver (Aplodontia rufa nigra)

The landscape genetics framework is typically applied to broad regions that occupy only small portions of a species’ range. Rarely is the entire range of a taxon the subject of study. We examined the landscape genetic structure of the endangered Point Arena mountain beaver (Aplodontia rufa nigra), whose isolated geographic range is found in a restricted (85 km²) but heterogenous region in California. Based on its diminutive range we may predict widespread gene flow and a relatively weak role for landscape variation in defining genetic structure. We used skin, bone, tissue and noninvasively collected hair samples to describe genetic substructure and model gene flow. We examined spatial partitioning of multilocus DNA genotypes and mitochondrial DNA haplotypes. We identified 3 groups from microsatellite data, all of which had low estimates of effective population size consistent with significant tests for historical bottlenecks. We used least-cost-path analysis with the microsatellites to examine how vegetation type affects gene flow in a landscape genetics framework. Gene flow was best predicted by a model with “Forest” as the most permeable, followed by “Riparian”. Agricultural lands demonstrated the highest resistance. MtDNA data revealed only two haplotypes: one was represented in all 57 individuals that occurred north of the east–west flowing Garcia River. South of the river, however, both haplotypes occurred, often at the same site suggesting that the river may have affected historical patterns of genetic divergence.     

Grains of connectivity: analysis at multiple spatial scales in landscape genetics

Landscape genetic analyses are typically conducted at one spatial scale. Considering multiple scales may be essential for identifying landscape features influencing gene flow. We examined landscape connectivity for woodland caribou (Rangifer tarandus caribou) at multiple spatial scales using a new approach based on landscape graphs that creates a Voronoi tessellation of the landscape. To illustrate the potential of the method, we generated five resistance surfaces to explain how landscape pattern may influence gene flow across the range of this population. We tested each resistance surface using a raster at the spatial grain of available landscape data (200 m grid squares). We then used our method to produce up to 127 additional grains for each resistance surface. We applied a causal modelling framework with partial Mantel tests, where evidence of landscape resistance is tested against an alternative hypothesis of isolation-by-distance, and found statistically significant support for landscape resistance to gene flow in 89 of the 507 spatial grains examined. We found evidence that major roads as well as the cumulative effects of natural and anthropogenic disturbance may be contributing to the genetic structure. Using only the original grid surface yielded no evidence for landscape resistance to gene flow. Our results show that using multiple spatial grains can reveal landscape influences on genetic structure that may be overlooked with a single grain, and suggest that coarsening the grain of landcover data may be appropriate for highly mobile species. We discuss how grains of connectivity and related analyses have potential landscape genetic applications in a broad range of systems.     

Simulating natural selection in landscape genetics

Linking landscape effects to key evolutionary processes through individual organism movement and natural selection is essential to provide a foundation for evolutionary landscape genetics. Of particular importance is determining how spatially-explicit, individual-based models differ from classic population genetics and evolutionary ecology models based on ideal panmictic populations in an allopatric setting in their predictions of population structure and frequency of fixation of adaptive alleles. We explore initial applications of a spatially-explicit, individual-based evolutionary landscape genetics program that incorporates all factors--mutation, gene flow, genetic drift and selection--that affect the frequency of an allele in a population. We incorporate natural selection by imposing differential survival rates defined by local relative fitness values on a landscape. Selection coefficients thus can vary not only for genotypes, but also in space as functions of local environmental variability. This simulator enables coupling of gene flow (governed by resistance surfaces), with natural selection (governed by selection surfaces). We validate the individual-based simulations under Wright-Fisher assumptions. We show that under isolation-by-distance processes, there are deviations in the rate of change and equilibrium values of allele frequency. The program provides a valuable tool (cdpop v1.0; http://cel.dbs.umt.edu/software/CDPOP/) for the study of evolutionary landscape genetics that allows explicit evaluation of the interactions between gene flow and selection in complex landscapes.     

Landscape genetics of the blotched tiger salamander (Ambystoma tigrinum melanostictum)

The field of landscape genetics has great potential to identify habitat features that influence population genetic structure. To identify landscape correlates of genetic differentiation in a quantitative fashion, we developed a novel approach using geographical information systems analysis. We present data on blotched tiger salamanders (Ambystoma tigrinum melanostictum) from 10 sites across the northern range of Yellowstone National Park in Montana and Wyoming, USA. We used eight microsatellite loci to analyse population genetic structure. We tested whether landscape variables, including topographical distance, elevation, wetland likelihood, cover type and number of river and stream crossings, were correlated with genetic subdivision (F(ST)). We then compared five hypothetical dispersal routes with a straight-line distance model using two approaches: (i) partial Mantel tests using Akaike's information criterion scores to evaluate model robustness and (ii) the BIOENV procedure, which uses a Spearman rank correlation to determine the combination of environmental variables that best fits the genetic data. Overall, gene flow appears highly restricted among sites, with a global F(ST) of 0.24. While there is a significant isolation-by-distance pattern, incorporating landscape variables substantially improved the fit of the model (from an r2 of 0.3 to 0.8) explaining genetic differentiation. It appears that gene flow follows a straight-line topographic route, with river crossings and open shrub habitat correlated with lower F(ST) and thus, decreased differentiation, while distance and elevation difference appear to increase differentiation. This study demonstrates a general approach that can be used to determine the influence of landscape variables on population genetic structure.     

Landscape genetic analyses reveal fine‐scale effects of forest fragmentation in an insular tropical bird

Within the framework of landscape genetics, resistance surface modelling is particularly relevant to explicitly test competing hypotheses about landscape effects on gene flow. To investigate how fragmentation of tropical forest affects population connectivity in a forest specialist bird species, we optimized resistance surfaces without a priori specification, using least‐cost (LCP) or resistance (IBR) distances. We implemented a two‐step procedure in order (i) to objectively define the landscape thematic resolution (level of detail in classification scheme to describe landscape variables) and spatial extent (area within the landscape boundaries) and then (ii) to test the relative role of several landscape features (elevation, roads, land cover) in genetic differentiation in the Plumbeous Warbler (Setophaga plumbea). We detected a small‐scale reduction of gene flow mainly driven by land cover, with a negative impact of the nonforest matrix on landscape functional connectivity. However, matrix components did not equally constrain gene flow, as their conductivity increased with increasing structural similarity with forest habitat: urban areas and meadows had the highest resistance values whereas agricultural areas had intermediate resistance values. Our results revealed a higher performance of IBR compared to LCP in explaining gene flow, reflecting suboptimal movements across this human‐modified landscape, challenging the common use of LCP to design habitat corridors and advocating for a broader use of circuit theory modelling. Finally, our results emphasize the need for an objective definition of landscape scales (landscape extent and thematic resolution) and highlight potential pitfalls associated with parameterization of resistance surfaces.     

Predicting landscape-genetic consequences of habitat loss, fragmentation and mobility for multiple species of woodland birds

Inference concerning the impact of habitat fragmentation on dispersal and gene flow is a key theme in landscape genetics. Recently, the ability of established approaches to identify reliably the differential effects of landscape structure (e.g. land-cover composition, remnant vegetation configuration and extent) on the mobility of organisms has been questioned. More explicit methods of predicting and testing for such effects must move beyond post hoc explanations for single landscapes and species. Here, we document a process for making a priori predictions, using existing spatial and ecological data and expert opinion, of the effects of landscape structure on genetic structure of multiple species across replicated landscape blocks. We compare the results of two common methods for estimating the influence of landscape structure on effective distance: least-cost path analysis and isolation-by-resistance. We present a series of alternative models of genetic connectivity in the study area, represented by different landscape resistance surfaces for calculating effective distance, and identify appropriate null models. The process is applied to ten species of sympatric woodland-dependant birds. For each species, we rank a priori the expectation of fit of genetic response to the models according to the expected response of birds to loss of structural connectivity and landscape-scale tree-cover. These rankings (our hypotheses) are presented for testing with empirical genetic data in a subsequent contribution. We propose that this replicated landscape, multi-species approach offers a robust method for identifying the likely effects of landscape fragmentation on dispersal.     

The use and misuse of regression models in landscape genetic analyses

The field of landscape genetics has been rapidly evolving, adopting and adapting analytical frameworks to address research questions. Current studies are increasingly using regression‐based frameworks to infer the individual contributions of landscape and habitat variables on genetic differentiation. This paper outlines appropriate and inappropriate uses of multiple regression for these purposes, and demonstrates through simulation the limitations of different analytical frameworks for making correct inference. Of particular concern are recent studies seeking to explain genetic differences by fitting regression models with effective distance variables calculated independently on separate landscape resistance surfaces. When moving across the landscape, organisms cannot respond independently and uniquely to habitat and landscape features. Analyses seeking to understand how landscape features affect gene flow should model a single conductance or resistance surface as a parameterized function of relevant spatial covariates, and estimate the values of these parameters by linking a single set of resistance distances to observed genetic dissimilarity via a loss function. While this loss function may involve a regression‐like step, the associated nuisance parameters are not interpretable in terms of organismal movement and should not be conflated with what is actually of interest: the mapping between spatial covariates and conductance/resistance. The growth and evolution of landscape genetics as a field has been rapid and exciting. It is the goal of this paper to highlight past missteps and demonstrate limitations of current approaches to ensure that future use of regression models will appropriately consider the process being modeled, which will provide clarity to model interpretation.     

A new analytical approach to landscape genetic modelling: least-cost transect analysis and linear mixed models

Landscape genetics aims to assess the effect of the landscape on intraspecific genetic structure. To quantify interdeme landscape structure, landscape genetics primarily uses landscape resistance surfaces (RSs) and least-cost paths or straight-line transects. However, both approaches have drawbacks. Parameterization of RSs is a subjective process, and least-cost paths represent a single migration route. A transect-based approach might oversimplify migration patterns by assuming rectilinear migration. To overcome these limitations, we combined these two methods in a new landscape genetic approach: least-cost transect analysis (LCTA). Habitat-matrix RSs were used to create least-cost paths, which were subsequently buffered to form transects in which the abundance of several landscape elements was quantified. To maintain objectivity, this analysis was repeated so that each landscape element was in turn regarded as migration habitat. The relationship between explanatory variables and genetic distances was then assessed following a mixed modelling approach to account for the nonindependence of values in distance matrices. Subsequently, the best fitting model was selected using the statistic. We applied LCTA and the mixed modelling approach to an empirical genetic dataset on the endangered damselfly, Coenagrion mercuriale. We compared the results to those obtained from traditional least-cost, effective and resistance distance analysis. We showed that LCTA is an objective approach that identifies both the most probable migration habitat and landscape elements that either inhibit or facilitate gene flow. Although we believe the statistical approach to be an improvement for the analysis of distance matrices in landscape genetics, more stringent testing is needed.     

Watershed‐level brook trout genetic structuring: Evaluation and application of riverscape genetics models

相似文献   

Spatial genetic structure and landscape connectivity in black bears: Investigating the significance of using different land cover datasets and classifications in landscape genetics analyses

Landscape genetic analyses allow detection of fine‐scale spatial genetic structure (SGS) and quantification of effects of landscape features on gene flow and connectivity. Typically, analyses require generation of resistance surfaces. These surfaces characteristically take the form of a grid with cells that are coded to represent the degree to which landscape or environmental features promote or inhibit animal movement. How accurately resistance surfaces predict association between the landscape and movement is determined in large part by (a) the landscape features used, (b) the resistance values assigned to features, and (c) how accurately resistance surfaces represent landscape permeability. Our objective was to evaluate the performance of resistance surfaces generated using two publicly available land cover datasets that varied in how accurately they represent the actual landscape. We genotyped 365 individuals from a large black bear population (Ursus americanus) in the Northern Lower Peninsula (NLP) of Michigan, USA at 12 microsatellite loci, and evaluated the relationship between gene flow and landscape features using two different land cover datasets. We investigated the relative importance of land cover classification and accuracy on landscape resistance model performance. We detected local spatial genetic structure in Michigan''s NLP black bears and found roads and land cover were significantly correlated with genetic distance. We observed similarities in model performance when different land cover datasets were used despite 21% dissimilarity in classification between the two land cover datasets. However, we did find the performance of land cover models to predict genetic distance was dependent on the way the land cover was defined. Models in which land cover was finely defined (i.e., eight land cover classes) outperformed models where land cover was defined more coarsely (i.e., habitat/non‐habitat or forest/non‐forest). Our results show that landscape genetic researchers should carefully consider how land cover classification changes inference in landscape genetic studies.     

Landscape genetics and the spatial distribution of chronic wasting disease

Predicting the spread of wildlife disease is critical for identifying populations at risk, targeting surveillance and designing proactive management programmes. We used a landscape genetics approach to identify landscape features that influenced gene flow and the distribution of chronic wasting disease (CWD) in Wisconsin white-tailed deer. CWD prevalence was negatively correlated with genetic differentiation of study area deer from deer in the area of disease origin (core-area). Genetic differentiation was greatest, and CWD prevalence lowest, in areas separated from the core-area by the Wisconsin River, indicating that this river reduced deer gene flow and probably disease spread. Features of the landscape that influence host dispersal and spatial patterns of disease can be identified based on host spatial genetic structure. Landscape genetics may be used to predict high-risk populations based on their genetic connection to infected populations and to target disease surveillance, control and preventative activities.     

Permeability of the landscape matrix between amphibian breeding sites

For organisms that reproduce in discrete habitat patches, land cover between patches (known as the matrix) is important for dispersal among breeding sites. Models of patchy populations often incorporate information on the permeability of the matrix to dispersal, sometimes based on expert opinion. I estimated the relative resistance to gene flow of land cover types and barriers using F_ST calculated from microsatellite markers in two amphibians, within an 800‐km² area in northern Switzerland. The species included a frog (Rana temporaria: 996 individuals, 48 populations, seven markers) and a newt (Triturus alpestris: 816 individuals, 41 populations, seven markers). Open fields and urban areas were more resistant to gene flow than forested land; roads and highways also reduced permeability. Results were similar for the two species. However, differences in resistance among matrix elements were relatively low: gene flow through urban areas was reduced by only 24–42% relative to forest; a divided highway reduced gene flow by 11–40% and was 7–8 times more resistant than a secondary road. These data offer an empirically based alternative to expert opinion for setting relative resistance values in landscape models.