Source populations are thought to occur in good-quality habitat, produce a net surplus of offspring and send out many emigrants. In contrast, sink populations are located in habitat of lower quality, have a net deficit of offspring and receive more immigrants than producing emigrants. To study source–sink dynamics, researchers must apply genetic methods that can infer the directionality of migration or gene flow. While common landscape genetic approaches measure past and symmetrical gene flow from genetic distances among populations or individuals (Storfer et al. 2010), various assignment tests not only allow to estimate contemporary or recent migration but also to infer the direction of migration (Holderegger & Wagner 2008)—exactly what is needed when studying source–sink dynamics. One such assignment approach is implemented in the software bimr (Faubet & Gaggiotti 2008). bimr uses Bayesian assignment to infer the proportion of recent immigrants in a population from their genotypes and calculates corresponding asymmetrical migration rates among pairs of populations. The estimated migration rates are based on effective gene flow, because descendants of migrants in the previous generation are determined. In contrast to other Bayesian assignment methods, bimr is able to determine recent migration even among weakly differentiated populations (i.e. FST > 0.01) with unequal sample sizes (Faubet & Gaggiotti 2008); a situation often encountered by genetic studies at the landscape level.
Andreasen et al. (2012) apply bimr to study source–sink dynamics in mountain lions (Fig. 1) in the Great Basin of Nevada and California. They use a large microsatellite sample set on 739 individuals (sampled from overlapping generations between 2004 and 2010) and objectively define populations by using the Bayesian clustering software tess (Chen et al. 2007). In this way, Andreasen et al. (2012) identify five, spatially ordered populations (South, West, North, East and Central). Subsequently, the authors apply bimr to calculate recent directional migration rates among these five populations. Their results clearly show that the Southern population forms a source population with net emigration, while the West, East and North populations are sink populations with net immigration, but less emigration. The South source area is characterized by large wildlife refuges and limited hunting. In contrast, hunting is substantially higher and/or anthropogenic disturbance more severe in the West, East and North sink areas. In two of these areas, hunting pressure thus seems to affect population dynamics in mountain lions, and sink populations are eventually only maintained by constant immigration.
Andreasen et al. (2012) also identify salt and barren desert areas, such as the Lahontan Basin, as natural landscape barriers to migration of mountain lions by simple spatial coincidences of landscape features with boundaries of genetic clusters (i.e. the overlay approach of landscape genetics; Storfer et al. 2010; Fig. 2). Yet, their study bears great potential with regard to the analysis of landscape influences on directional migration rates, particularly so, as bimr (Faubet & Gaggiotti 2008) implements a multiple regression approach of migration rates on landscape data. Contemporary or recent, and, in particular, directional estimates of migration (or gene flow) originating from assignment tests (or parentage analyses) have rarely been used in landscape genetics (Holderegger & Wagner 2008), although their application seems to be straightforward. For instance, researchers could analyse any landscape features (such as the proportion of desert area or the density of prey species) in transects or strips of particular width between pairs of populations (e.g. Emaresi et al. 2010) and relate these landscape data with information on contemporary or recent migration and gene flow. Balkenhol et al. (2009) suggest that a combination of bimr to estimate contemporary migration rates with multiple regression on distance matrices (Legendre et al. 1994) to identify those landscape features that affect migration are particularly powerful (for a conceptual outline see Fig. 2). However, some conceptual problems remain. For example, how can we incorporate directional estimates of migration or gene flow in multiple regression given that two different migration rates are related with exactly the same landscape data? bimr, for instance, just uses both directional migration rates between any given pair of populations in multiple regression (Faubet & Gaggiotti 2008). Another, still open question is whether Akaike's information criterion AIC, or any related criterion, can be used to compare the explanatory power of different regression models in landscape genetics (Goldberg & Waits 2010; Van Strien et al. 2012).
The study of Andreasen et al. (2012) shows that the incorporation of asymmetrical measurements of migration in landscape genetics adds substantially to this field's relevance for application. It would have been difficult to analyse source–sink dynamics and its potential causes at a large spatial scale in a large mammal predator using conventional methods such as population surveys, mark–recapture experiments, radio-tracking or GPS sendering of individuals. Similar insights relevant for conservation management of wild (meta) populations could be envisaged by a more widespread application of approaches measuring contemporary or recent migration in other species. We therefore advocate the use of methods to infer directional measures of contemporary or recent migration or gene flow, such as the one presented, in future landscape genetic studies.