Landscape genetic studies offer a fine-scale understanding of how habitat heterogeneity influences population genetic structure. We examined population genetic structure and conducted a landscape genetic analysis for the endangered Central American Squirrel Monkey (Saimiri oerstedii) that lives in the fragmented, human-modified habitats of the Central Pacific region of Costa Rica. We analyzed non-invasively collected fecal samples from 244 individuals from 14 groups for 16 microsatellite markers. We found two geographically separate genetic clusters in the Central Pacific region with evidence of recent gene flow among them. We also found significant differentiation among groups of S. o. citrinellus using pairwise FST comparisons. These groups are in fragments of secondary forest separated by unsuitable “matrix” habitats such as cattle pasture, commercial African oil palm plantations, and human residential areas. We used an individual-based landscape genetic approach to measure spatial patterns of genetic variance while taking into account landscape heterogeneity. We found that large, commercial oil palm plantations represent moderate barriers to gene flow between populations, but cattle pastures, rivers, and residential areas do not. However, the influence of oil palm plantations on genetic variance was diminished when we restricted analyses to within population pairs, suggesting that their effect is scale-dependent and manifests during longer dispersal events among populations. We show that when landscape genetic methods are applied rigorously and at the right scale, they are sensitive enough to track population processes even in species with long, overlapping generations such as primates. Thus landscape genetic approaches are extremely valuable for the conservation management of a diverse array of endangered species in heterogeneous, human-modified habitats. Our results also stress the importance of explicitly considering the heterogeneity of matrix habitats in landscape genetic studies, instead of assuming that all matrix habitats have a uniform effect on population genetic processes.
Citation: Blair ME, Melnick DJ (2012) Scale-Dependent Effects of a Heterogeneous Landscape on Genetic Differentiation in the Central American Squirrel Monkey (Saimiri oerstedii). PLoS ONE 7(8): e43027. https://doi.org/10.1371/journal.pone.0043027
Editor: João Pinto, Instituto de Higiene e Medicina Tropical, Portugal
Received: April 25, 2012; Accepted: July 16, 2012; Published: August 15, 2012
Copyright: © Blair, Melnick. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This research was generously funded by National Science Foundation (NSF) Award No. BCS-0847912, the American Association of University Women, the Margot Marsh Biodiversity Foundation, Columbia University, the International Primatological Society, the American Society of Primatologists, and the Northwest Primate Conservation Society. MEB was supported during her graduate training by a NSF Integrative Graduate Education and Research Traineeship program, the New York Consortium in Evolutionary Primatology, Award No. DGE-0333415. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Many species exist in spatially structured populations linked by dispersal and gene flow, which can influence evolutionary, demographic, and ecological processes. Studies of population genetic structure are important for species living in complex, heterogeneous landscapes, which may affect that structure . Understanding population genetic structure is also critical to informing conservation management ; conservation managers need to measure the extent and distribution of genetic diversity to accurately predict population persistence, especially for small, fragmented populations .
Landscape genetic approaches are increasingly used to understand the influence of landscape characteristics on population genetic structure and dispersal patterns , , , . These emerging approaches combine population genetics, spatial statistics, and landscape ecology to measure the effects of landscape features on gene flow , , . Landscape genetic approaches attempt to detect genetic discontinuities among individuals and then correlate those discontinuities with landscape features . For example, many recent landscape genetic studies of birds, herpetofauna, terrestrial mammals, and primates found that geographic distances incorporating a cost to particular landscape features showed a stronger correlation to genetic distances than straight-line Euclidean distances between sampled individuals , , , , , , .
One particular challenge in landscape genetics has been to quantify the relative effects of various landscape parameters on gene flow , , . Recent studies in landscape ecology reject simplistic models where the matrix, or the unsuitable habitat between patches of suitable habitat for a given species, uniformly inhibits movement among patches. Instead, the matrix is dynamic, heterogeneous, and can have both positive and negative effects on dispersal and thus the long-term persistence of a species , , , , , , , . Responses to matrix quality and heterogeneity are species-specific, often correlating with body size, degree of arboreality, dietary specialization, and habitat breadth , , . Thus, landscape genetic studies should compare different classes of matrix habitats at multiple scales to understand their relative effects on genetic variation and better predict processes of population divergence in modified landscapes.
The endangered Central American Squirrel Monkey (Saimiri oerstedii, Primates: Cebidae) provides an ideal opportunity to investigate population genetic structure in a human modified, heterogeneous landscape. S. oerstedii live in groups of 18 or more individuals, which have home ranges of approximately 200 ha. Their diet includes arthropods, flowers, fruits, and small vertebrates , , and they are restricted to the Pacific moist forests of Costa Rica and northern Panama below ∼500 m asl , , , . This range area is characterized by frequent landscape disturbance from high rainfall, wind, hurricanes, and rugged topography , . The subspecies S. o. citrinellus inhabits a particularly heterogeneous landscape in the Central Pacific region of Costa Rica, where fruit and rice plantations and cattle pasture replaced approximately 80% of the natural forest in the early 1900s .
Despite this heterogeneity, little work has been done to determine the effects of such drastic natural and anthropogenic landscape change on the genetic structure of S. o. citrinellus populations. To date the only published study of genetic diversity in S. o. citrinellus was based on a small sample (N = 8) from Manuel Antonio National Park (MANP) , which is the smallest national park in Costa Rica and the only protected area within the range of S. o. citrinellus. The study concluded that S. oerstedii have moderate to high genetic diversity compared to the three other primate species in Costa Rica (Alouatta palliata, Ateles geoffroyi, and Cebus capucinus). A much larger sample collected across a broader spatial scale and wider range of habitats is necessary to determine the effects of landscape heterogeneity on S. o. citrinellus population genetic structure. In this study, we analyzed a large number of non-invasively collected molecular samples of S. o. citrinellus and characterized overall population genetic structure using Bayesian clustering algorithms and F-statistics. We also used fine-scale landscape data within a least-cost distance framework to determine whether there is a relationship between landscape heterogeneity and population genetic structure and which, if any, matrix habitats might be related to genetic structure in S. o. citrinellus. Several aspects of S. oerstedii behavioral ecology suggest that some types of matrix habitat will affect patterns of gene flow more than others. For example, S. oerstedii are known to traverse small fruit plantations and live fences around residential areas, while they likely do not traverse large commercial oil palm plantations and rice plantations , . If some matrix habitats represent barriers to gene flow while others do not, geographic distances that weight barrier matrix habitats with high costs should correlate more strongly with genetic distance than geographic distances that weight passable matrix habitats with high costs. By contrast, if all matrix habitats prevent gene flow, different least-cost measures of distance through matrix habitat should not differ in the strength of their associations with genetic distance.
Permits to collect and import S. o. citrinellus fecal samples included Costa Rican Ministry of Energy and the Environment permit ACOPAC-INVN-14-08 and Centers for Disease Control and Prevention permit 2008-08-151. Also, an IACUC animal care protocol was approved by Columbia University for this research (AC-AAAA5583).
Sampling and DNA Extraction
Fecal samples were collected from S. o. citrinellus in the Central Pacific region of Costa Rica from September 2008 – April 2009 (Figure 1). S. o. citrinellus is not continuously distributed between sampling locations as they are restricted to larger secondary forest fragments , ; two groups (PD and O) were sampled in oil palm plantations but adjacent to a forest fragment or riparian forest. From 304 fecal samples, we verified genotypes for 233 S. o. citrinellus individuals, comprised of 10 to 20 adult individuals from each of 14 groups. The average size of sampled groups was 39 individuals (range 18–67). Whenever possible (N = 13 groups), more than 10 individuals per group were sampled to increase the precision of genetic analyses in detecting dispersal and migration . Samples were stored in 8 ml plastic tubes with RNAlater buffer (Ambion) at −4°C in the field and −20°C in the laboratory. Eleven additional DNA samples from individuals of the southern subspecies S. o. oerstedii, used as an outgroup, were contributed by G. Gutierrez (University of Costa Rica) for a total of 244 samples.
Sampled S. o. citrinellus groups in the Central Pacific region of Costa Rica, showing the limit of 500 m asl to their distribution and different classes of matrix habitat as defined by a manual land cover classification.
We extracted DNA using QIAamp DNA Stool Minikits (Qiagen) with small modifications to the “Isolation of DNA from Stool for Human DNA Analysis” protocol (see , ). We used real-time quantitative PCR after extraction to quantify the amount of nuclear DNA in each sample with iQ SYBR Green Supermix (Bio-Rad) and universal primate primers amplifying approximately 200 bp of nuclear DNA . We included samples with greater than 0.5 ng/µl DNA concentrations (averaged over two replicate runs) in our genotyping analyses.
We PCR amplified 17 autosomal microsatellite markers on multiplex panels of three or four markers (CJ7 ; D17s804, D3s1210, D3s1229, D3s1776, D4s111, D5s111, D8s165, D8s260 ; Leon 15, Leon 21 ; LL118, LL157, LL311 ; Locus5 ; SB38 ; Table S1) using Multiplex PCR Kits (Qiagen; for reaction concentrations and conditions see ). PCR products were electrophoresed on an ABI 3730 DNA Analysis System with GENESCAN 500 ROX size standard, and genotypes were called using GeneMapper software (ABI). We confirmed heterozygous genotypes by scoring alleles at least four times and homozygous genotypes at least seven times since allelic dropout is often a problem when amplifying microsatellite markers from fecal samples , , . We used the program MICROCHECKER  to test for null alleles, and removed one of the 17 microsatellite markers because it was found to possibly contain null alleles (D13s160). We tested for linkage disequilibrium and deviations from Hardy-Weinberg equilibrium (HWE) across markers in ARLEQUIN v 3.1  and found no evidence for linkage disequilibrium. We did find violations of HWE when analyzing samples at the species and population levels, consistent with a Wahlund effect , . At the group level, three groups had one marker that was significantly out of HWE (D3s1766 at Gamalotillo, Leon21 at Chirraca, and Leon15 at MANP), likely due to the presence of related individuals in the sample , , .
Analysis of Population Genetic Structure
We ran our multilocus genotypes in STRUCTURE v 2.2  and BAPS v 2  to infer the number of genetic clusters in our dataset. We ran 10 independent iterations of K = 1−16 in STRUCTURE for 2,000,000 Markov Chain Monte Carlo (MCMC) generations with a 200,000 burn-in period, assuming correlated allele frequencies and admixture. We inferred K using ln P(X | K) and the ΔK method , where optimum K has the highest ΔK value, or rate of change in the log probability of the data between successive K-values. We ran BAPS under the individual clustering module and the default settings (stochastic optimization) also for 10 separate iterations for a maximum K of 1–21. Both programs were run without spatial information.
We also examined genetic structure in S. o. citrinellus microsatellite data with pairwise tests for differentiation among groups and populations using F-statistics  calculated with Weir and Cockerham’s  estimators in FSTAT v 2.9 , for 10,000 randomizations not assuming HWE. Bonferroni corrections were used throughout when conducting multiple comparisons.
We detected first generation migrants and admixed individuals using STRUCTURE and GENECLASS v 2.0 , . Migrants were defined as individuals that were assigned to one population but geographically sampled from another population. Admixed individuals were defined as those individuals that could not be confidently assigned to either population following the ranking and plotting approach of Beaumont et al. . To detect first generation migrants in STRUCTURE, we ran the program using the cluster memberships inferred as described above using the ΔK method as prior population information. We conducted several runs using a range of values for MIGRPRIOR (0.001−0.1) following Pritchard et al. . Because choice of MIGRPRIOR did not significantly affect program outputs, we present results from MIGRPRIOR = 0.09, the average migration rate between populations of S. o. citrinellus found using the software BAYESASS , following Liu et al. . Burn-in and run length were the same as earlier runs of STRUCTURE without prior population information. We also performed an exclusion test and used the ‘Detect first generation migrants’ option in GENECLASS , , using both Lh and Lh/Lmax, which represent, respectively, the most appropriate statistic when all potential source populations have not been sampled and when they have . The probability of individual genotypes coming from each population was calculated by comparing individual genotypes to 10,000 simulated individuals per population .
We tested for genetic signatures of a recent population bottleneck using BOTTLENECK . We tested our data under the Infinite Alleles Model (IAM), the Stepwise Mutation Model (SSM), and the Two Phase Model (TPM) with 10,000 replications using a sign test. The sign test compares observed and expected heterozygosity excess. If excess is higher than expected (based on equilibrium) for a large majority of markers in a population, the population may have recently experienced a genetic bottleneck , .
Land Cover Classification
Using 32 geo-referenced aerial photographs (taken with a DCS camera, each 10×10.5 km), a 90×7 km multispectral MASTER line image taken in the year 2005 obtained from the Centro Nacional de Alta Technología (CENAT) in Costa Rica, and a forest cover dataset generated by EOSL, CCT and FONAFIFO  with Landsat 7 TM satellite imagery from the year 2000, we delineated five habitat classes (forest, cattle pasture, rivers, oil palm plantations, and residential areas; Figure 1) manually in ArcGIS v 9.3 (ESRI) at a 20×20 m resolution for the 1800 km2 study area. We defined residential areas as clusters of human residences less than 100 m apart and consisting of a total area greater than 3 hectares. Our classification also included an “other non-forest” category, which included shrimp farms, rice plantations, abandoned lots, and residences that were not concentrated enough to fit our definition of a residential area. Ninety-eight percent of 131 ground reference points were accurately reflected in the classification.
Distribution of three genetic clusters estimated in STRUCTURE (A) and BAPS (B). Vertical lines are broken into colored segments showing the proportion of each individual assigned to each K (within S. o. citrinellus: western cluster – black, eastern cluster 1– light grey, eastern cluster 2– grey; S. o. oerstedii – white). Sample locations are listed at the bottom of the figure, and are arrayed in west to east order (left to right).
Landscape Genetic Analyses
We estimated pairwise genetic relationships between individuals using Rousset’s â  and Moran’s I ,  to measure both distance (â) and similarity (I) . We measured two types of geographic distances: Euclidean distances, straight-line distances on a map, and least-cost geographic distances, where the costs of dispersing across different habitat classes were incorporated into the measure of geographic distance. We calculated Euclidean geographic distances in ArcMap v 9.3 (ESRI) using the sampled coordinates and we calculated least-cost geographic distances for the five habitat classes identified above (forests, oil palm plantations, cattle pastures, residential areas, and rivers) using the COSTDISTANCE function in ArcGIS v 9.3 (ESRI). We varied the cost of one class while keeping all others at an equal, low cost (1) and then repeated this process for each habitat class. We assessed least-cost distances for a range of 6 arbitrary cost values (10, 50, 100, 1000, 5000, 10000) to account for sensitivity , .
We performed Mantel tests of matrix correspondence  between genetic distances and geographic distances in ZT . Also, because least-cost distances and Euclidean distances are not independent, we performed partial Mantel tests  in ZT to test the strength of relationships between genetic and least-cost distance matrices while controlling for the effect of Euclidean distance , . Partial correlations show high power and accuracy in their ability to infer the effect of landscapes on dispersal when there is a strong contrast between the permeability of different landscape elements . Significance was assessed with 10,000 permutations.
Recent studies have suggested that the spatial scale of landscape genetic analyses can have important effects on results, especially inferences about which landscape features affect gene flow , , . To test the stability of our inferences across spatial scales, we repeated the above analyses including only pairs of samples within genetic clusters of S. o. citrinellus as inferred by analyses of population genetic structure, excluding between population pairs.
We used the results of our least-cost distance analyses to generate a resistance surface characterizing the cumulative effects of landscape heterogeneity on gene flow in S. o. citrinellus, implemented in the software CIRCUITSCAPE 3.5 . Instead of calculating a single least-cost path, CIRCUITSCAPE incorporates aspects of electronic circuit theory (i.e. electronic resistance) with a random walk approach to visualize resistance patterns across the landscape , , . We tested for relationships between pairwise resistance distances (generated using the “pairwise” mode in CIRCUITSCAPE) and genetic distances using simple and partial Mantel tests as described above. Also, we produced cumulative current flow maps in the “all to one” mode, with focal points as sampled groups and an 8-neighbor connection scheme, and the source current for each group scaled to group size. The habitat grid encompassed a 5–25 km buffer around peripheral focal points, with forests, rivers, residential areas and cattle pastures at very low costs (<10) and oil palm plantations at a moderate cost of 20, following recommendations from the CIRCUITSCAPE manual (resistance values above 20 are considered moderate, while values above 200 are considered high). We ran the program under several other parameterizations with similar results to those presented.
Analysis of Population Genetic Structure
Bayesian clustering analysis revealed at least two genetically distinct populations within S. o. citrinellus in the Central Pacific of Costa Rica. The most likely number of clusters across the whole sample, which included outgroup S. o. oerstedii individuals, was four in both STRUCTURE and BAPS (Figure 2,S1), although K = 2 or 3 were only slightly less likely (Figure S1). For K = 4, one cluster represents samples from the subspecies S. o. oerstedii while the other three clusters are within S. o. citrinellus. The first cluster within S. o. citrinellus includes almost all individuals from western groups. The other two clusters do not seem to be geographically separated and include mostly members from eastern groups (Figure 2). The two eastern clusters likely represent ancestral polymorphism that has not been sorted out in this population. We defined a western population and an eastern population of S. o. citrinellus based on the strong geographic clustering inferred using STRUCTURE and BAPS. We also found significant population differentiation using F-statistics between the defined eastern and western populations of S. o. citrinellus (FST = 0.0903, P = 0.05). Microsatellite allelic diversity in populations of S. o. citrinellus ranged from 3 to 15 alleles (mean = 8.1) in the western population and from 5 to 27 (mean = 10.6) in the eastern population. Pairwise FST values among groups of S. o. citrinellus ranged from 0.016–0.19, with a mean of 0.103 (Table 1). Pairwise FST values among groups from the same population (mean = 0.06, range 0.016–0.11) were less than pairwise FST values among groups from different populations (mean = 0.14, range 0.070–0.19, Table 1).
Analyses using STRUCTURE and GENECLASS estimated 7 likely migrants and 10 potentially admixed individuals between the western and eastern populations (Table 2). STRUCTURE estimated 1 potential migrant (individual G1, P = 0.019), while GENECLASS estimated the same individual as a migrant in addition to 6 others (P<0.01) using both likelihood methods (Lh and Lh/Lmax). These 7 migrants were assigned to a cluster where they were not geographically sampled in GENECLASS and also had lower probabilities of belonging to their geographic origin cluster compared to other individuals in STRUCTURE (Table 2). Ten to 40 km separated sampled sites and inferred populations of origin.
In STRUCTURE, we found breaks in mean Q-values at Q = 0.2 and 0.8 and therefore defined individuals with mean Q-values from 0.2 to 0.8 as potentially admixed , , , , . We found 20 individuals with mean Q values between 0.2 and 0.8 (Figure S2), and 10 were also assigned in GENECLASS to >1 cluster with a high probability (>0.2) of assignment to a cluster other than the origin (Table 2). In one case (individual K1, from Chirraca), an individual had low probability of being in either cluster as estimated by GENECLASS, but STRUCTURE identified it as a potential migrant (P = 0.046). We interpreted this individual as potentially admixed, or a migrant from an unsampled ‘ghost’ population.
We found no evidence of a bottleneck from the microsatellite data. None of the sign or Wilcoxon tests across any models suggested heterozygosity excess. Similarly, all mode-shift tests showed normal L-shaped distributions.
Landscape Genetic Analyses
Genetic distances were correlated with Euclidean geographic distances both within populations (eastern population, â: r = 0.14, P = 0.0021; I: r = −0.10, P<0.0001; western population, â: r = 0.17, P<0.0001; I: r = −0.32, P<0.0001) and when the entire sample was considered (â: r = 0.21, P<0.0001; I: r = −0.18, P<0.0001). Across the entire sample, oil palm plantations at a cost of 10 were the only habitat class for which Mantel’s r-values between least-cost and genetic distance matrices were consistently larger than Mantel’s r-values between genetic and Euclidean distance matrices; these results held for both Moran’s I and Rousset’s â and in both simple and partial Mantel tests (Figure 3, Table S2). However, when only within population pairs were considered, Mantel’s r-values between least-cost and genetic distance matrices for oil palm plantations did not differ greatly from Mantel’s r-values between genetic and Euclidean distance matrices (Figure 4, Table S3,S4). For both the eastern and western within population pairs, there were large, significant partial Mantel’s r-values for Rousset’s â when costs of 5,000 and 10,000 were given to oil palm plantations, but they were not greater than the Mantel’s r-value between genetic and Euclidean distance matrices, and Moran’s I did not show the same pattern (Figure 4, Table S3,S4).
Results of Mantel tests of least-cost distances against genetic relationships, including all pairs of individuals. Negative Mantel’s r-values are given (left) for Morans’ I for easier comparison with trends in Rousset’s â (right). Filled symbols (diamonds for simple Mantel tests and circles for partial Mantel tests) represent statistically significant Mantel’s r-values in the expected direction (positive for â and negative for I). Dotted lines represent the Mantel’s r-value for Euclidean distance against genetic distance.
Least-cost distances for forests showed the expected relationship for non-barrier habitat classes when all sample pairs were considered, with absolute Mantel’s r-values almost consistently decreasing with increasing cost in both simple and partial Mantel tests (Figure 3, Table S2). This pattern was less clear when only within population pairs were considered (Figure 4, Table S3,S4).
Results of Mantel tests of least-cost distances against genetic relationships, including only pairs within the eastern population (black) or western population (grey). Least-cost distances are for oil palm plantations, cattle pastures, forests, rivers, and residential areas (top to bottom). Negative Mantel’s r-values are given (left) for Morans’ I for easier comparison with trends in Rousset’s â (right). Filled symbols represent statistically significant Mantel’s r-values in the expected direction (positive for â and negative for I).
Pairwise resistance distances calculated in CIRCUITSCAPE showed strong relationships with genetic distance when all samples were considered (simple Mantel, â: r = 0.29, P<0.0001; I: r = −0.25, P<0.0001; partial Mantel controlling for Euclidean distance, â: r = 0.24, P<0.0001; I: r = −0.21, P<0.0001). Relationships were not as strong when only within population pairs were considered (eastern population, simple Mantel â: r = 0.17, P = 0.001; I: r = −0.11, P<0.0001; partial Mantel â: r = 0.09, P>0.05; I: r = −0.04, P = 0.042; western population, simple Mantel â: r = 0.14, P = 0.02; I: r = −0.23, P<0.0001; partial Mantel â: r = 0.07, P>0.05; I: r = −0.08, P>0.05). When controlling for the effect of one another, resistance distances and least-cost distances for oil palm plantations at a cost of 10 both showed significant relationships with genetic distances (Table 3). However, least-cost distances showed stronger relationships with genetic distances when controlling for resistance distances than vice versa, at least when all pairs or only western pairs were considered (Table 3).
The resistance surface output from CIRCUITSCAPE showed that even with a moderate cost, oil palm plantations cause an extensive area of low current flow in the middle of the landscape due to the cumulative costs of traversing through this expansive matrix habitat type (Figure 5). The resistance surface also shows that current flow is stronger among sites in the eastern population than in the western population (Figure 5), which is consistent with the consistently larger Mantel’s r-values between genetic and geographic distance matrices in the western population (Figure 4).
Population Genetic Structure
Clustering analysis and F-statistics revealed that S. o. citrinellus are structured into two genetically distinct populations, an eastern and a western population. These populations have also been supported by AMOVAs as well as mtDNA haplogroups , . Genetic differentiation between populations was significant but weak, which is consistent with recent separation or a high level of gene flow between them. There was a high level of agreement between clustering algorithms in terms of assignment of individuals to the two subpopulations, including the inference of recent migrants, despite the different algorithms that each program uses to describe population membership (GENECLASS calculates the probability that an individual belongs to a population, whereas STRUCTURE calculates the proportion of an individual’s genome that is characteristic of a population; Table 2). These results suggest the two populations are connected by gene flow despite their genetic distinctiveness. Importantly, individuals inferred to be migrants or of mixed or ambiguous ancestry were scattered throughout the sampled area, indicating weak differentiation across the landscape (Figure 1, Table 2).
We note that our sample contains more individuals from the eastern population, which contains more contiguous forest patches, and fewer individuals from the western population, where forest patches are less contiguous and more isolated. STRUCTURE and BAPS have been shown to produce spurious results if samples are obtained unevenly from a continuous population . Also, previous publications based on simulation results have suggested that clustering analyses will group all individuals from the largest, most continuous region together, while grouping all other individuals in a second, less well resolved cluster . However, this was not the case in our results; the western population, which was less continuously sampled due to the natural distribution of S. o. citrinellus in the area, showed a more resolved cluster than the eastern population, which showed a less well resolved cluster with evidence of ancestral polymorphism.
Landscape Genetic Analysis
In addition to significant differentiation among populations, we also found significant genetic differentiation among several groups of S. o. citrinellus (but not all) using pairwise FST comparisons. These groups are in fragments of secondary forest separated by varying types of unsuitable habitats such as cattle pasture, African oil palm plantations, and rice plantations. While F-statistics effectively measure spatial variance in gene frequencies, we used an individual-based landscape genetic approach to instead measure aspects of spatial patterns of gene frequencies while taking into account landscape heterogeneity to better understand the forces behind the patterns of population genetic structure shown here.
Our results suggest that landscape heterogeneity affects genetic relationships in S. o. citrinellus and that different matrix habitat classes have different effects on dispersal in the studied landscape. Least-cost distances for oil palm plantations at a cost of 10 had stronger relationships with genetic distances than Euclidean distances in both simple and, more importantly, partial Mantel tests, suggesting that these least-cost distances contributed different information from Euclidean distances and that oil palm plantations represent moderate barriers to gene flow. When we restricted the analyses to within population pairs, none of the least-cost distances had stronger effects than Eucidean distances, suggesting that the effect of oil palm plantations on gene flow is largely scale-dependent and manifests during longer dispersal events between populations.
Similarly, the resistance surface showed how oil palm plantations, even when given only moderate resistance values, impede current flow because they dominate the landscape and the cost of crossing them accumulates over large distances. The strongest current flow in the resistance surface was in the eastern population near Manuel Antonio National Park (MANP), where there is not only the highest density of natural forest and monkey groups in the region, but also a break in the oil palm plantations due to complex topography. We found weaker isolation by distance in the eastern population as compared to the western population, also likely due to this break in the oil palm plantations.
By contrast, we found that cattle pastures, rivers, and residential areas do not differ greatly from Euclidean distance in their effects on genetic distance. Cattle pastures, rivers, and residential areas in this region are often surrounded by live fences of fruiting trees, which might explain why they did not show strong negative effects on S. o. citrinellus gene flow in this landscape. Alternatively, the landscape composition and configuration of these habitat types is quite different from that of oil palm plantations, which occur as large, contiguous expanses. Cattle pastures, rivers, and residential areas, by contrast, are smaller and more isolated from one another. In this particular landscape, these features did not influence gene flow, but if we tested multiple landscapes with a range of variability in the composition and configuration of landscape features, we might find different results , , .
We can lend further support to this idea by examining differences in landscape configuration between the eastern and western populations in our dataset. For example, in the western population, increasing cost for residential areas did not change Mantel’s r-values between least-cost and genetic distances, but in the eastern population, residential areas showed a non-barrier pattern, with increasing costs resulting in decreasing Mantel’s r-values. It is likely that residential areas are non-barriers in both populations, but this analysis was not sensitive enough to pick up the signal in the western population, where residential areas are smaller in number and more spread out in comparison to the eastern population.
A disadvantage to the Mantel test framework is that it is difficult to choose among closely related models or models with only slightly different Mantel’s r-values . Here, pairwise resistance distances created in CIRCUITSCAPE produced similar Mantel’s r-values to least-cost distances for oil palm plantations at a cost of 10. Ours is one of a small number of empirical studies that has used CIRCUITSCAPE to characterize landscape connectivity in human-dominated environments, and the other studies have found that resistance distances outperform least-cost distances in how well they characterize gene flow , . Here, partial Mantel tests controlling for the effect of least-cost distances on resistance and vice versa suggested that oil palm plantations at a cost of 10 had the stronger relationship with genetic distance when all pairs were considered and when only western pairs were considered (Table 3). However, the Mantel’s r-values were within a reasonable margin of error of one another. Both distances assigned moderate costs to oil palm plantations and had stronger relationships with genetic distances than Euclidean distance matrices when between population pairs were included in the analysis, showing a consistent trend.
It is possible that the temporal scale of our landscape data has influenced our results , . Long, overlapping generations affect the power of landscape genetic approaches to detect the effects of current or even historical landscape patterns on genetic structure, and this issue is particularly important for long-lived primate taxa. A recent simulation study showed that it would take 1–15 generations to detect barriers to gene flow using Mantel’s r . Saimiri have an average generation time of 3–6 years , meaning that at least 12 and up to 25 generations have likely transpired since the major transformation of the Central Pacific landscape in the early 1900s (although at that time the plantations were of banana, and later replaced with African oil palm, they were of the same configuration and area) . Thus, we should have been able to detect barriers to gene flow caused by landscape features in this study. However, genetic distance measures such as Rousset’s â are based on FST and as such may reflect processes that are more likely to be apparent in historical landscape data . Although we have shown some effect of landscape heterogeneity on gene flow using current landscape data, historical data from the early 1900s may show a stronger relationship. Unfortunately, such data not readily available.
Although we tested for a population contraction and found no evidence for a bottleneck, sequencing additional loci, in particular nuclear introns and coding mtDNA loci, would allow for a more robust analysis of alternative hypotheses through Approximate Bayesian Computation (ABC) . ABC would allow us to test various alternative scenarios of population expansion and/or contraction and also to test for a temporal correlation between increasing genetic structure and the expansion of banana and oil palm plantations in the Central Pacific of Costa Rica.
Another issue that is common in many landscape genetic studies is that the different focal matrix habitats used in our analysis are likely of different ages. Rivers are older than the other four habitat classes we considered, and some residential areas are likely younger than the cattle pastures and oil palm plantations, most of which were established in the early 1900s. Future analyses might address this issue by including different sets of models and molecular markers that pinpoint different temporal scales in order to distinguish between the effects of historical and recent landscape changes on population genetic structure .
Implications for Conservation Management
This study exemplifies how important it is to conduct landscape genetic analyses in species that show evidence of population genetic structure in heterogeneous landscapes. Furthermore, we highlight the importance of quantifying the relative effects of different matrix habitat classes in landscape genetic analyses , , , instead of assuming that all non-suitable habitats have a uniform effect on dispersal and gene flow. Because we distinguished among matrix habitat classes, we have a finer understanding of what does and does not constitute a barrier to S. o. citrinellus gene flow in the Central Pacific Costa Rican landscape. We are also able to make more detailed recommendations to conservation managers regarding the types of matrix habitat that S. o. citrinellus may or may not use to disperse among patches of forest in the Central Pacific. In a concurrent study, we used resistance surfaces to test different biological corridor configurations for their potential ability to augment gene flow through oil palm plantations . Another strategy to augment gene flow through oil palm plantations might be to plant understory vegetation, which has been shown to increase bird richness in oil palm plantations in eastern Guatemala .
However, we must recognize that our results are specific to the Central Pacific landscape, and conservation managers should be careful not to apply our results in other landscapes or for other populations of S. oerstedii. For example, in a landscape where cattle pastures dominate instead of oil palm plantations, a separate landscape genetic study would be necessary to measure the relative effects of each matrix habitat to determine whether it might be more important for conservation managers to augment gene flow through cattle pastures rather than oil palm plantations.
When attempting to translate the results of any landscape genetic analysis to patterns of functional connectivity, we must also acknowledge that measures of genetic distance do not equate to animal movement patterns. Simulations that model the sums of individual behavioral decisions are likely necessary to best inform conservation management of taxa in heterogeneous landscapes , , , . A possible next step would be to incorporate a recently published new form of population viability analysis that uses individual-based models (simulations) that incorporate behavioral decisions alongside models of landscape change over time .
Inference of the number of genetic clusters (K) estimated using STRUCTURE.
Ranked mean Q (proportional membership in each cluster) for each individual.
Microsatellite markers amplified in 244 Saimiri oerstedii samples.
Results of simple and partial Mantel tests between genetic distances (Moran’s I and Rousset’s â ) and cost distances among all individuals.
Results of simple and partial Mantel tests between genetic distances (Moran’s I and Rousset’s â ) and cost distances, including only sample pairs within the eastern population.
We thank A. Goncalves da Silva, J. Pinto, and three anonymous reviewers for comments that helped improve this manuscript. We also thank the other members of M. Blair’s dissertation advisory committee M. Cords, R. DeSalle, T. Disotell, and A. Di Fiore, who provided valuable comments throughout this research project. We thank collaborators in Costa Rica including G. Gutierrez, G. Wong, J. Aguero, M. Cook, L. Leon, J. Bustamante, O. Masis, L. Rubí, M. Schulte, F. Villanea, and H. Abarca. Thanks especially to J. Jimenez (IGN) and C. Varagas (PRIAS/CENAT) for their help in procuring aerial photographs for the Central Pacific of Costa Rica. Field assistants F. Rutka, D. Lake, R. Leon, and W. Chacon helped collect and catalogue samples in the field. Thanks to K. Chiou, M. Montague, A. Morales-Jimenez, C. Bergey, A. Burrell, L. Pozzi, J. Corush, and S. Pickett for their help in the laboratory and to J. Trinidad-Christensen and others at Columbia University’s Electronic Data Service for help creating the cost-distance python script. And finally, many thanks to M. Brown, L. Douglas, W. Erb, C. Schmitt, J. Hodgson, Z. Liu, K. Guschanski, O. Pineda, B. McRae, R. Raaum, and J. Munshi-South for helpful advice on data analyses.
Conceived and designed the experiments: MEB DJM. Performed the experiments: MEB. Analyzed the data: MEB. Contributed reagents/materials/analysis tools: MEB. Wrote the paper: MEB DJM.
- 1. Keyghobadi N, Roland J, Strobeck C (1999) Influence of landscape on the population genetic structure of the alpine butterfly Parnassius smintheus (Papilionidae). Mol Ecol 8: 1481–1495.
- 2. Simberloff D (1988) The contribution of population and community biology to conservation science. Annu Rev Ecol Syst 19: 473–511.
- 3. DeSalle R, Amato G (2004) The expansion of conservation genetics. Nat Rev Genet 5: 702–712.
- 4. Balkenhol N, Gugerli F, Cushman SA, Waits LP, Coulon A, et al. (2009) Identifying future research needs in landscape genetics: where to from here? Landscape Ecol 24: 455–463.
- 5. Segelbacher G, Cushman SA, Epperson BK, Fortin M, Francois O, et al. (2010) Applications of landscape genetics in conservation biology: concepts and challenges. Conserv Genet 11: 375–385.
- 6. Sork VL, Waits LP (2010) Contributions of landscape genetics - approaches, insights, and future potential. Mol Ecol 19: 3489–3495.
- 7. Storfer A, Murphy MA, Spear SF, Holderegger R, Waits LP (2010) Landscape genetics: where are we now? Mol Ecol 19: 3496–3514.
- 8. Holderegger R, Wagner R (2006) A brief guide to landscape genetics. Landscape Ecol 21: 793–796.
- 9. Manel S, Schwarts MK, Luikart G, Taberlet P (2003) Landscape genetics: Combining landscape ecology and population genetics. Trends in Ecol Evol 18: 189–197.
- 10. Storfer A, Murphy MA, Evans JS, Goldberg CS, Robinson S, et al. (2007) Putting the ‘landscape’ in landscape genetics. Heredity 98: 128–142.
- 11. Blumenthal JM, Abreu-Grobois FA, Austin TJ, Broderick AC, Bruford MW, et al. (2009) Turtle groups or turtle soup: dispersal patterns of hawksbill turtles in the Caribbean. Mol Ecol 18: 4841–4853.
- 12. Frantz AC, Pope LC, Etherington TR, Wilson GJ, Burke T (2010) Using isolation-by-distance-based approaches to assess the barrier effect of linear landscape elements on badger (Meles meles) dispersal. Mol Ecol 19: 1663–1674.
- 13. Greenwald KR, Purrenhage JL, Savage WK (2009) Landcover predicts isolation in Ambystoma salamanders across region and species. Biol Conserv 142: 2493–2500.
- 14. Hokit DG, Ascunce MS, Ernst J, Branch LC, Clark AM (2010) Ecological metrics predict connectivity better than geographic distance. Conserv Genet 11: 149–159.
- 15. Lada H, Thomson JR, MacNally R, Taylor AC (2008) Impacts of massive landscape change on a carnivorous marsupial in south-eastern Australia: inferences from landscape genetics analysis. J Appl Ecol 45: 1732–1741.
- 16. Liu Z, Ren B, Wu R, Zhao L, Hao Y, et al. (2009) The effect of landscape features on population genetic structure in Yunnan snub-nosed monkeys (Rhinopithecus bieti) implies an anthropogenic genetic discontinuity. Mol Ecol 18: 3831–3846.
- 17. Quemere E, Crouau-Roy B, Rabarivola C, Louis EE, Chikhi L (2010) Landscape genetics of an endangered lemur (Propithecus tattersalli) within its entire fragmented range. Mol Ecol 19: 1606–1621.
- 18. Cushman SA, McKelvey KS, Hayden J, Schwartz MK (2006) Gene flow in complex landscapes: testing multiple hypotheses with causal modeling. Am Nat 168: 486–499.
- 19. Braunisch V, Segelbacher G, Hirzel AH (2010) Modelling functional landscape connectivity from genetic population structure: a new spatially explicit approach. Mol Ecol 17: 3664–3678.
- 20. Baum KA, Haynes KJ, Dillemuth FP, Cronin JT (2004) The matrix enhances the effectiveness of corridors and stepping stones. Ecology 85: 2671–2676.
- 21. Crooks KR, Sanjayan M (2006) Connectivity conservation. Cambridge, UK: Cambridge University Press.
- 22. Dunford W, Freemark W (2005) Matrix matters: Effects of surrounding land uses on forest birds near Ottawa, Canada. Landscape Ecol 20: 497–511.
- 23. Fischer J, Lindenmayer DB (2007) Landscape modification and habitat fragmentation: a synthesis. Global Ecol Biogeogr 16: 265–280.
- 24. Hilty J, Lidicker W, Merenlender A (2006) Corridor ecology: The science and practice of linking landscapes for biodiversity conservation. Washington DC: Island Press.
- 25. Jules ES, Shahani P (2003) A broader ecological context to habitat fragmentation: Why matrix habitat is more important than we thought. J Veg Sci 14: 459–464.
- 26. Kindlmann P, Aviron S, Burel F (2005) When is landscape matrix important for determining animal fluxes between resource patches? Ecol Complex 2: 150–158.
- 27. Kupfer JA, Malanson GP, Franklin SB (2006) Not seeing the ocean for the islands: the mediating influence of matrix-based processes on forest fragmentation effects. Global Ecol Biogeogr 15: 8–20.
- 28. Beier P, Noss RF (1998) Do habitat corridors provide connectivity? Conserv Biol 12: 1241–1252.
- 29. Gehring TM, Swihart RK (2003) Body size, niche breadth, and ecologically scaled responses to habitat fragmentation: mammalian predators in an agricultural landscape. Biol Conserv 109: 283–295.
- 30. Goncalves da Silva A (2007) Causes of spatial genetic structure in mammals: A case study in the Atlantic Forest, Brazil. New York: Columbia University. 182 p.
- 31. Boinski S (1999) The social organizations of squirrel monkeys: Implications for ecological models of social evolution. Evol Anthropol 8: 101–112.
- 32. Wong G (1990) Uso del hábitat, estimación de la composición y densidad poblacional del mono tití (Saimiri oerstedii citrinellus) en la zona de Manuel Antonio, Quepos, Costa Rica. Heredia, Costa Rica: Universidad Nacional.
- 33. Arauz J (1993) Estado de conservación del mono tití (Saimiri oerstedii citrinellus) en su área de distribución original, Costa Rica. Heredia, Costa Rica: Universidad Nacional.
- 34. Boinski S, Jack K, Lamarsh C, Coltrane JA (1998) Squirrel monkeys in Costa Rica: drifting to extinction. Oryx 32: 45–58.
- 35. Boinski S, Sirot L (1997) Uncertain conservation status of Squirrel monkeys in Costa Rica. Folia Primatol 68: 181–193.
- 36. Wallace DR (1997) Central American landscapes. In: Coates AG, editor. Central America: A natural and cultural history. New Haven: Yale University Press. 72–96.
- 37. Boinski S, Ehmke E, Kauffman L, Schet S, Vreedzaam A (2005) Dispersal patterns among three species of squirrel monkeys (Saimiri oerstedii, S. boliviensis and S. sciureus): II. Within-species and local variation. Behaviour 142: 633–677.
- 38. Mattey J (1992) Colonización espontánea, uso y capacidad de uso de suelos. Quepos, Costa Rica: Ministerio de Agricultura y Ganadería.
- 39. Zaldivar ME, Rocha O, Glander KE, Aguilar G, Huertas AS, et al. (2004) Distribution, ecology, life history, genetic variation, and risk of extinction of nonhuman primates from Costa Rica. Rev Biol Trop 52: 679–693.
- 40. Sierra C, Jimenez I, Altricher M, Fernandez M, Gomez G, et al. (2003) New data on the distribution and abundance of Saimiri oerstedii citrinellus. Primate Conserv 19: 5–9.
- 41. Goudet J, Perrin N, Waser P (2002) Tests for sex-biased dispersal using bi-parentally inherited genetic markers. Mol Ecol 11: 1103–1114.
- 42. Blair ME (2011) Habitat modification and gene flow in Saimiri oerstedii: Landscape genetics, intraspecific molecular phylogenetics, and conservation. Ph.D. Dissertation. New York, NY: Columbia University. 205 p.
- 43. Blair ME, Melnick DJ (2012) Genetic evidence for dispersal by both sexes in the Central American Squirrel Monkey, Saimiri oerstedii citrinellus. Am J Primatol 74: 37–47.
- 44. Morin PA, Chambers KE, Boesch C, Vigilant L (2001) Quantitative polymerase chain reaction analysis of DNA from noninvasive samples for accurate microsatellite genotyping of wild chimpanzees (Pan troglodytes verus). Mol Ecol 10: 1835–1844.
- 45. Nievergelt CM, Mundy NI, Woodruff DS (1998) Microsatellite primers for genotyping common marmosets (Callithrix jacchus) and other callitrichids. Mol Ecol 7: 1432–1434.
- 46. Invitrogen ResGen Human MapPair. Carlsbad, California.
- 47. Perez-Sweeney BM, Valladares-Padua C, Burrell AS, Di Fiore A, Satkoski JA, et al. (2005) Dinucleotide microsatellite primers designed for a critically endangered primate, the black lion tamarin (Leontopithecus chrysopygus). Mol Ecol Notes 5: 198–201.
- 48. Di Fiore A, Fleischer RC (2004) Microsatellite markers for woolly monkeys (Lagothrix lagotricha) and their amplification in other New World primates (Primates: Platyrrhini). Mol Ecol Notes 4: 246–249.
- 49. Grativol AD, Ballou JD, Fleischer RC (2001) Microsatellite variation within and among recently fragmented populations of the golden lion tamarin (Leontopithecus rosalia). Conserv Genet 2: 1–9.
- 50. Bohle UR, Zischler H (2002) Polymorphic microsatellite loci for the mustached tamarin (Saguinus mystax) and their cross-species amplification in other New World monkeys. Mol Ecol Notes 2: 1–3.
- 51. Broquet T, Menard N, Petit E (2007) Noninvasive population genetics: a review of sample source, diet, fragment length and microsatellite motif effects on amplification success and genotyping error rates. Conserv Genet 8: 249–260.
- 52. Roon DA, Thomas ME, Kendall KC, Waits LP (2005) Evaluating mixed samples as a source of error in non-invasive genetic studies using microsatellites. Mol Ecol 14: 195–201.
- 53. Vigilant L (2002) Technical challenges in the microsatellite genotyping of a wild chimpanzee population using feces. Evol Anthropol 11: 162–165.
- 54. van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) Micro-checker: software for identifying and correcting genotyping errors in microsatellite data. Mol Ecol Notes 4: 535–538.
- 55. Excoffier L, Laval G, Schneider S (2005) Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evol Bioinform Online 1: 47–50.
- 56. Goossens B, Chikhi L, Jalil MF, Ancrenaz M, Lackman-Ancrenaz I, et al. (2005) Patterns of genetic diversity and migration in increasingly fragmented and declining orang-utan (Pongo pygmaeus) populations from Sabah, Malaysia. Mol Ecol 14: 441–456.
- 57. Wahlund S (1928) Zusammensetzung von populationen nd korrelationserscheinungen von standpunkt der vererbungslehre aus betrachtet. Hereditas 11: 65–106.
- 58. Bergl R, Vigilant L (2007) Genetic analysis reveals population structure and recent migration within the highly fragmented range of the cross-river gorilla (Gorilla gorilla diehli). Mol Ecol 16: 501–516.
- 59. Bourgain C, Abney M, Schneider D, Ober C, McPeek MS (2004) Testing for Hardy-Weinberg equilibrium in samples with related individuals. Genetics 168: 2349–2361.
- 60. Lukas D, Bradley BJ, Nsubuga AM, Doran-Sheehy D, Robbins MM, et al. (2004) Major histocompatibility complex and microsatellite variation in two populations of wild gorillas. Mol Ecol 13: 3389–3402.
- 61. Falush D, Stephens M, Pritchard JK (2003) Inference of population structure using multilocus genotype data: Linked loci and correlated allele frequencies. Genetics 164: 1567–1587.
- 62. Corander J, Waldmann P, Marttinen P, Sillanpaa MJ (2004) BAPS 2: enhanced possibilities for the analysis of genetic population structure. Bioinformatics 20: 2363–2369.
- 63. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 14: 2611–2620.
- 64. Wright S (1978) Evolution and the genetics of populations, Vol. 4. Variability within and among natural populations. Chicago: University of Chicago Press.
- 65. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38: 1358–1370.
- 66. Goudet J (2001) FSTAT, a program to estimate gene diversity and fixation indices. Institute for Ecology, Laboratory for Zoology, University of Laussane.
- 67. Cornuet JM, Piry S, Luikart G, Estoup A, Solignac M (1999) New methods employing multilocus genotypes to select or exclude populations as origins of individuals. Genetics 153: 1989–2000.
- 68. Piry S, Alapetite A, Cornuet JM, Paetkau D, Baudouin L, et al. (2004) GeneClass2: a software for genetic assignment and first-generation migrant detection. J Hered 95: 536–539.
- 69. Beaumont MA, Barratt EM, Gotelli D, Kitchener AC, Daniels MJ, et al. (2001) Genetic diversity and introgression in the Scottish wildcat. Mol Ecol 10: 319–336.
- 70. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945–959.
- 71. Wilson GA, Rannala B (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genetics 163: 1177–1191.
- 72. Paetkau D, Slade R, Burden M, Estoup A (2004) Genetic assignment methods for the direct, real-time estimation of migration rate: a simulation-based exploration of accuracy and power. Mol Ecol 13: 55–65.
- 73. Piry S, Luikart G, Cornuet JM (1999) Bottleneck: a computer program for detecting recent reductions in the effective size using allele frequency data. J Hered 90: 502–503.
- 74. Cornuet JM, Luikart G (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics 144: 2001–2014.
- 75. EOSL CCT, FONAFIFO (2002) Estudio de cobertura forestal de Costa Rica con imágenes Landsat TM 7 para el año 2000. San Jose, Costa Rica: Laboratorio de Sistemas de Observación Terrestre (EOSL), Departamento de Ciencias de la Tierra y la Atmósfera, Universidad de Alberta, Centro Científico Tropical (CCT), and Fondo Nacional de Financiamiento Forestal (FONAFIFO). 12 p.
- 76. Rousset F (2000) Genetic differentiation between individuals. J Evol Biol 13: 58–62.
- 77. Epperson BK (2003) Geographic Genetics. Princeton, N.J.: Princeton University Press.
- 78. Moran PAP (1950) Notes on continuous stochastic phenomena. Biometrika 37: 17–23.
- 79. Perez-Espona S, Perez-Barberia FJ, McLeod JE, Jiggins CD, Gordon IJ, et al. (2008) Landscape features affect gene flow of Scottish Highland red deer (Cervus elaphus). Mol Ecol 17: 981–996.
- 80. Rayfield B, Fortin M, Fall A (2010) The sensitivity of least-cost habitat graphs to relative cost surface values. Landscape Ecol 25: 519–532.
- 81. Mantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Res 27: 209–220.
- 82. Bonnet E, Van de Peer Y (2002) zt: a software tool for simple and partial Mantel tests. J Stat Softw 7: 1–12.
- 83. Smouse PE, Long JC, Sokal RR (1986) Multiple regression and correlation extensions of the Mantel test of matrix correspondence. Syst Zool 35: 627–632.
- 84. Broquet T, Ray N, Petit E, Fryxell JM, Burel F (2006) Genetic isolation by distance and landscape connectivity in the American marten (Martes americana). Landscape Ecol 21: 877–889.
- 85. Cushman SA, Landguth EL (2010) Spurious correlations and inference in landscape genetics. Mol Ecol 19: 3592–3602.
- 86. Jaquiery J, Broquet T, Hirzel AH, Yearsley J, Perrin N (2011) Inferring landscape effects on dispersal from genetic distances: how far can we go? Mol Ecol 20: 691–705.
- 87. Anderson CD, Epperson BK, Fortin M, Holderegger R, James PMA, et al. (2010) Considering spatial and temporal scale in landscape-genetic studies of gene flow. Mol Ecol 19: 3565–3575.
- 88. Short Bull RA, Cushman SA, Mace R, Chilton T, Kendall KC, et al. (2011) Why replication is important in landscape genetics: American black bear in the Rocky Mountains. Mol Ecol 20: 1092–1107.
- 89. Latch EK, Boarman WI, Walde A, Fleischer RC (2011) Fine-Scale Analysis Reveals Cryptic Landscape Genetic Structure in Desert Tortoises. PLoS ONE 6: e27794.
- 90. McRae BH, Shah VB (2009) CIRCUITSCAPE User Guide. ONLINE. Available at http://www.circuitscape.org: The University of California Santa Barbara.
- 91. McRae BH (2006) Isolation by resistance. Evolution 60: 1551–1561.
- 92. McRae BH, Beier P (2007) Circuit theory predicts gene flow in plant and animal populations. Proc Natl Acad Sci USA 104: 19885–19890.
- 93. McRae BH, Dickson BG, Keitt TH, Shah VB (2008) Using circuit theory to model connectivity in ecology, evolution and conservation. Ecology 89: 2712–2724.
- 94. Latch EK, Dharmarajan G, Glaubitz JC, Rhodes OE (2006) Relative performance of Bayesian clustering software for inferring population substructure and individual assignment at low levels of population differentiation. Conserv Genet 7: 295–302.
- 95. Fogelqvist J, Niittyvuopio A, Agren J, Savolainen O, Lascoux M (2010) Cryptic population genetic structure: the number of inferred clusters depends on sample size. Mol Ecol Resources 10: 314–323.
- 96. Guillot G, Leblois R, Coulon A, Frantz AC (2009) Statistical methods in spatial genetics. Mol Ecol 18: 4734–4756.
- 97. Munshi-South J (2012) Urban landscape genetics: Canopy cover predicts gene flow between white-footed mouse (Peromyscus leocupus) populations in New York City. Molecular Ecology 21: 1360–1378.
- 98. Brooks TM, Pimm SL, Oyugi JO (1999) Time lag between deforestation and bird extinction in tropical forest fragments. Conserv Biol 13: 1140–1150.
- 99. Landguth EL, Cushman SA, Schwartz MK, McKelvey KS, Murphy MA, et al. (2010) Quantifying the lag time to detect barriers in landscape genetics. Mol Ecol 19: 4179–4191.
- 100. Jack K (2007) The Cebines. In: Campbell C, Fuentes A, MacKinnon K, Panger M, Bearder S, editors. Primates in Perspective. New York: Oxford University Press. 107–120.
- 101. Hickerson MJ, Carstens BC, Cavender-Bares J, Crandall KA, Graham CH, et al. (2010) Phylogeography’s past, present, and future: 10 years after Avise, 2000. Molecular Phylogenetics and Evolution 54: 291–301.
- 102. Chiucchi JE, Gibbs HL (2010) Similarity of contemporary and historical gene flow among highly fragmented populations of an endangered rattlesnake. Mol Ecol 19: 5345–5358.
- 103. Watling JI, Nowakowski AJ, Donnelly MA, Orrock JL (2011) Meta-analysis reveals the importance of matrix composition for animals in fragmented habitat. Global Ecol Biogeogr 20: 209–217.
- 104. Nájera A, Simonetti JA (2010) Can oil palm plantations become bird friendly? Agroforest Syst 80: 203–209.
- 105. Tracey JA (2006) Individual-based modeling as a tool for conserving connectivity. In: Crooks KR, Sanjayan M, editors. Connectivity Conservation. New York: Cambridge University Press. 343–368.
- 106. Bowler DE, Benton TG (2005) Causes and consequences of animal dispersal strategies: relating individual behaviour to spatial dynamics. Biol Rev 80: 205–225.
- 107. Knowlton JL, Graham CH (2010) Using behavioral landscape ecology to predict species’ responses to land-use and climate change. Biol Conserv 143: 1342–1354.
- 108. Lowe WH, Allendorf FW (2010) What can genetics tell us about population connectivity? Mol Ecol 19: 3038–3051.
- 109. Nabe-Nielsen J, Sibly RM, Forchhammer MC, Forbes VE, Topping CJ (2010) The effects of landscape modifications on the long-term persistence of animal populations. PLoS ONE 5: e8932.