Landscape genetics and phylogeography are complementary fields with the primary goals of understanding how environmental and geographic factors influence and have influenced population-level evolutionary processes (Avise et al. 1987; Manel et al. 2003; Storfer et al. 2007; Knowles 2009; Wang 2010). Both disciplines continue to expand rapidly, presenting researchers with numerous options for data collection and analysis. The volumes of genetic and GIS data that can now be collected and the range and number of methods with which they can be analysed provide myriad opportunities for investigating population structure, gene flow, demography and evolutionary history at various spatial and temporal scales (Holderegger & Wagner 2008; Knowles 2009; Storfer et al. 2010). Nevertheless, a certain subset of data types and analytical tools exists, which is optimally suited for investigation into each of the specific goals of landscape genetics and phylogeography (Sunnucks 2000; Selkoe & Toonen 2006; Scoble & Lowe 2010). Recognizing these benefits and limitations, and choosing the best combination of data and methods for examining a specific hypothesis, gives researchers the best opportunity to arrive at a biologically accurate and meaningful answer to their question of interest.
Despite their similarities and shared general objectives, landscape genetics and phylogeography do possess some important differences (Wang 2010). In particular, according to their traditional definitions, phylogeography investigates the historical processes generating patterns of genetic variation (Avise et al. 1987; Knowles 2009), while landscape genetics examines the contemporary processes affecting genetic variation (Storfer et al. 2007; Scoble & Lowe 2010). Last year, I wrote an article outlining the important reasons for this difference (Wang 2010), to which Bohonak & Vandergast (2011) have provided a counter-argument. Here, I clarify a number of points in my original article (Wang 2010) and argue further that different genetic markers are appropriate for drawing inferences at different temporal scales. Specifically, I will summarize the commonly held viewpoint that markers with higher mutation rates, like microsatellites, are better for inferring recent and contemporary (ongoing to a few tens of generations) evolutionary processes, while those with lower mutation rates, including typical cp/mtDNA and nDNA sequences, are better for inferring historical (many hundreds to thousands of generations) processes and that there are indeed important differences between landscape genetics and phylogeography.
At the centre of Bohonak & Vandergast’s (2011) argument is that I have overlooked the value of mtDNA for landscape genetics. However, in my original article, I actually wrote that ‘there are good reasons for incorporating cp/mtDNA into landscape genetic studies’ (Wang 2010, p. 2606), but whereas Bohonak & Vandergast (2011) feel that it can be utilized to investigate contemporary processes, I believe it should be used for examining historical processes and highlighted several studies that have utilized it to this effect (e.g. Braaker & Heckel 2009; Pease et al. 2009). Researchers have long recognized that significant limitations exist for the use of mtDNA data in the study of recent and ongoing processes. For instance, Moritz (1994, p. 401) wrote that ‘estimating demographic parameters, e.g. migration rate and population size, from patterns of mtDNA diversity is fraught with difficulty, particularly where populations are fluctuating, and is unlikely to produce quantitative estimates sufficiently accurate’, and more recently, Pease et al. (2009, p. 1849) pointed out that ‘the use of DNA sequence data alone has made it difficult to detect more recent, environmentally mediated divergence’ among populations. The reality is that while mtDNA and microsatellites have been proven powerful for many applications, ‘both have important and unavoidable limitations’ (Zhang & Hewitt 2003, p. 563).
I am not the first or only author to take the viewpoint that different genetic markers are appropriate for inference at different temporal scales because of their different mutation rates. Sunnucks (2000, p. 199) pointed out that ‘by examining genetic markers with appropriate rates of change, and, therefore, suitable signals, information can be obtained about almost any population and evolutionary process through the hierarchy of life’. The illumination of evolutionary processes at different times through the use of genetic markers with different mutation rates is possible because ‘a slow mutational process allows the signature of events in the distant past to persist longer’ (Selkoe & Toonen 2006, p. 618), and concordantly, a faster mutational process allows the signature of recent events to be captured, which is why ‘microsatellite loci are especially useful for the inference of recent demographic events’ (Pearse & Crandall 2004, p. 595). Thus, mutational rates are important for choosing a proper marker, and ‘if one is interested in a potential historical barrier to gene flow or tracing the recolonization of territory since the last ice age, markers with lower mutational rates are likely to be the most informative. In contrast, if one is interested in present day demography or connectivity patterns… microsatellites with higher mutational rates are preferable’ (Selkoe & Toonen 2006, p. 618).
Thus, we can see how genetic markers with different mutational rates, such as mtDNA and microsatellites, can by utilized together to look at multiple temporal scales. As Zhang & Hewitt (2003, p. 579) wrote, ‘mitochondrial and microsatellite markers are complementary in the sense that they reveal different aspects of a complex story at different depth of perception’. This has led some authors ‘to separately examine the effects of historical versus contemporary fragmentation on genetic structure by using different molecular markers that, because of their different rates of mutation, presumably reflect different temporal scales of response’ (Keyghobadi 2007, p. 1053), an approach I advocated in my original article (Wang 2010). This approach has been embraced (prior to and after the publication of my article) by many studies that have successfully used mtDNA and microsatellite data together to disentangle historical and contemporary evolutionary processes (e.g. Johnson et al. 2003; Joseph et al. 2008; Lada et al. 2008; Braaker & Heckel 2009; Pease et al. 2009; Cook et al. in press). Some of these studies provide empirical evidence for how microevolutionary processes affect mtDNA and microsatellites differently. For instance, Johnson et al. (2003, p. 3343) found that ‘recent fragmentation and isolation of greater prairie-chicken populations has had a stronger effect on microsatellite than mtDNA population structure’. Thus, both theoretical reasoning and empirical evidence suggest that microsatellites are preferable to mtDNA for the study of contemporary microevolutionary processes. The general acceptance of this idea is reflected in the overwhelming use of microsatellites compared with mtDNA in landscape genetics. Microsatellites were used in 70% of landscape genetic studies of animals from 1998 to 2008, seven times as often as mtDNA (Storfer et al. 2010).
Bohonak and Vandergast defend their viewpoint using one of their studies (Vandergast et al. 2007). Essentially, they argue that mtDNA sequence data can be used to accurately infer both historical and contemporary processes and that this viewpoint is supported by their study (Vandergast et al. 2007), which assumes that historical and contemporary processes can be accurately inferred from mtDNA sequence data. This logic is circular; while Vandergast et al. (2007) described a clever and innovative approach that may prove to be valuable, their method of inferring contemporary microevolutionary processes from mtDNA has not yet been externally validated. Although Vandergast et al. (2007) used a multitude of creative approaches, none of these independently affirms the accuracy of their approach to inferring contemporary evolutionary processes from mtDNA sequences. Spurious signals of genetic structure can be detected from mtDNA under some circumstances (Irwin 2002), making it important that studies use other markers or alternative, previously validated methods of detecting contemporary genetic structure to confirm the efficacy of a novel method.
Additionally, Bohonak & Vandergast (2011) claim that I did not consider frequency-based and nonequilibrium methods in landscape genetics and ‘implicitly focus’ only on methods requiring fixed differences between populations. However, in my article, I pointed to several examples of methods useful for detecting ongoing microevolutionary processes and others better at identifying historical processes; some of these included commonly used frequency-based methods and I wrote explicitly about analyses ‘based on allele frequency variation’ (Wang 2010, p. 2606). The primary goal of my discussion of different classes of methods was to emphasize that ‘understanding the timescale over which particular analytical methods are useful is critical to applying appropriate methods to questions in phylogeography and landscape genetics’ (Wang 2010, p. 2606). The issue is that some data and analyses can target the deep historical processes and some the very recent and ongoing processes, but an intermediate period exists that current methods and data will have trouble examining. It is this intermediate region, between the very recent and the historical, that is difficult to investigate.
Of course, there is clearly some debate about how well this intermediate period can be investigated and how much overlap there is between landscape genetics and phylogeography. Ideally, future methods will shrink the gap between landscape genetics and phylogeography, making it possible to construct models that show how genetic structure, gene flow and demography have changed over time with shifting environmental conditions. However, this is not presently the state of the field, and understanding current methodological limitations is important for future developments in landscape genetics. Realistically, additional studies are necessary to fully evaluate the efficacy of current analytical methods for investigating different periods of time. Simulation studies may be particularly useful for uncovering potential deficits (Epperson et al. 2010), and these should become one focus of the ongoing development of methods for landscape genetics and phylogeography.
Finally, innovations in next generation sequencing technologies are making the collection of large genetic data sets increasingly feasible in nonmodel organisms (Thomson et al. 2010; Wang 2010). The opportunity to incorporate many markers from across the nuclear and organellar genomes will likely provide significant advantages over the use of either microsatellites or mtDNA sequence data alone. Methods for analysing genomic data from many individuals are still only beginning to be developed, but these hold great promise for a variety of applications. For both landscape genetics and phylogeography, the development of spatially explicit methods that can integrate the massive volumes of genetic and GIS data becoming available are key to the advancement of these fields. These methods have the potential to greatly expand the scopes of landscape genetics and phylogeography and to enable the thorough investigation of all temporal and spatial scales.