512 Landscape Genetics HOH 2H 203 1.00 14H 0.80 0.60 L1 0.40 PJ 0.20 0.00 (a) 10 E1 247 1.5 Kilometers (b) Figure Represents a barrier detection method using a combination of Bayesian clustering analysis and overlay approach in Olympic National Park, USA (a) Bayesian clustering output using the software STRUCTURE (Pritchard et al., 2000) showing two genetic groups (red and green) identified from multiple sampling locations Each bar is an individual, and the proportion of its genotype assigned to either population is shown by the proportion of the bar that is red vs green (b) Identification of barrier using overlay approach Letter and number coded dots on map represent sampling localities Red and green circles correspond to identified populations using Bayesian clustering Black line is Hurricane Ridge, a visually identified barrier between the green and red genetic clusters compared as to which best fits the genetic distance data (Cushman et al., 2006; Cushman and Landguth, 2010) Another potential solution is to incorporate habitat suitability maps into estimation of resistance surfaces by using presence/ absence data for the study species (Wang et al., 2008) Two major modeling frameworks have been used to convert resistance values into measures of population connectivity – least-cost paths and circuit-based analyses Least-cost paths are based on the underlying assumption that costs, in terms of cumulative resistance values between populations, are minimized such that individuals tend to move along an optimal path (Figure 1(b); Adiransen et al., 2003) As an example, Epps et al (2007) used a least-cost based analysis to find that topographic distance explained a significantly higher proportion of variation in gene flow than Euclidean distance alone for desert bighorn sheep Least-cost paths have been criticized because they are based on the potentially unrealistic assumption that individuals ‘‘know’’ a priori, which is the least costly path to take during dispersal events One suggestion to overcome this problem is to consider multiple least-cost paths together in order to delineate a zone of probable movement (Rayfield et al., 2010) Circuit theory-based models simultaneously integrate all possible pathways that connect populations, while still assuming that genetic distance is positively correlated with resistance (McRae, 2006) The software CIRCUITSCAPE is used to estimate overall significance of the resistance surfaces in explaining genetic structuring among samples, along with tests of significance of individual landscape variables (Shah and McRae, 2008) The latter approach was shown to perform better, with a higher proportion of variation in gene flow explained when both least-cost path models and circuit theory-based models were applied to empirical datasets of mahogany and the wolverine (McRae and Beier, 2007) More recently, it has been suggested that least-cost paths may perform better on smaller spatial scales (Anderson et al., 2010) Nonetheless, given the assumptions inherent in least-cost path and circuit theory-based (i.e., that all possible paths are influence movement in some way) analyses, caution should be used when applying either analysis to understand population-level connectivity Spear et al (2010) provide guidelines and potential solutions for addressing these problems, including use of computer simulations or network-based approaches