Chapter 14 CURRENT STATUS, FUTURE OPPORTUNITIES, AND REMAINING CHALLENGES IN LANDSCAPE GENETICS 1 2 3 4 Niko Balkenhol, Samuel A. Cushman, Lisette P. Waits, and Andrew Storfer 1 Department of Wildlife Sciences, University of Göttingen, Germany 2 Forest and Woodlands Ecosystems Program, Rocky Mountain Research Station, United States Forest Service, USA 3 Fish and Wildlife Sciences, University of Idaho, USA 4 School of Biological Sciences, Washington State University, USA 14.1 INTRODUCTION essentially replaced the isolation-by-distance paradigm with spatially explicit tests of effects of landscape varia- Just over a decade ago, advances in geographic informa- bles on genetic population structure. More generally, tion systems and genetic methodology helped usher in landscape genetics has advanced the field of evolution- the field of landscape genetics, an amalgamation of ary ecology by providing a direct focus on relationships landscape ecology, population genetics, and spatial sta- between landscape patterns and population processes, tistics aimed at testing how landscape heterogeneity such as gene flow, selection, and genetic drift. shapes patterns of spatial genetic variation. It is clear Since the inception of landscape genetics, there have that the tools developed in landscape genetics have been many calls for: (1) better communication among shaped a new paradigm for conducting empirical popu- spatial statisticians, landscape ecologists, and population lation genetics studies; although population genetics geneticists (Storfer et al. 2007; Balkenhol et al. 2009a); theory itself remains largely unchanged, the majority (2) more rigorous hypothesis testing (Storfer et al. 2007, of empirical studies of spatial genetic structure now 2010; Cushman & Landguth 2010; Segelbacher et al. include spatially explicit tests of landscape influences 2010; Manel & Holderegger 2013); (3) increased con- on gene flow. As a result, landscape genetics has sideration of proper sampling design (Storfer et al. 2007; Landscape Genetics: Concepts, Methods, Applications, First Edition. Edited by Niko Balkenhol, Samuel A. Cushman, Andrew T. Storfer, and Lisette P. Waits. © 2016 John Wiley & Sons, Ltd. Published 2016 by John Wiley & Sons, Ltd. 248 Current status, future opportunities, and remaining challenges Holderegger & Wagner 2008; Anderson et al. 2010; 2012; Keller et al. 2013). While there have been a Spear et al. 2010; Landguth et al. 2012; see Chapter 4); few empirical and simulation studies on the effects of (4) developing predictive models for conservation and scale and landscape definition in the field (e.g. Cush- management, particularly in the face of climate change man & Landguth 2010; Galpern et al. 2012), the (Landguth & Cushman 2010; Segelbacher et al. 2010; vast majority of studies published in landscape genet­ Manel et al. 2010; Bollinger et al. 2014); (5) selecting ics have not addressed these issues at all. Indeed, appropriate analytical methods and understanding their many past studies have naively sought correlations underlying assumptions (e.g., Balkenhol et al. 2009b; between the genetic structure of a population and a Wagner & Fortin 2013; Chapters 3 to 5): and (6) under­ putative barrier feature, such as a road or river, standing the interactions of scale and landscape defini­ without considering that other landscape features tion with the heterogeneity of the environment in might also be important. Similarly, too few studies affecting the strength and detectability of landscape quantitatively assess the influences of alternative effects on genetic variation (e.g., Cushman & Landguth scales or landscape definitions. While some researchers 2010; Cushman et al. 2013a; Chapter 2). have begun to address these challenges by different In this book, it was our goal to summarize the current optimization procedures (e.g. Shirk et al. 2012; Gal­ state of knowledge with respect to these topics and pern & Manseau 2013; Castillo et al. 2014), this topic foreshadow a number of emerging opportunities and has not been thoroughly explored. Given the funda­ challenges the field will face in the coming decades. mental importance of scale optimization and correct Landscape genetics is a field in its infancy and is landscape definition in quantifying any pattern– characterized by rapid growth, multiple lines of parallel process relationship, we strongly feel that much and sometimes contradictory work, and an apparent more attention should be given to investigating the cultural and conceptual divide between practitioners relationships between landscape definition, spatial coming from genetic versus landscape ecological back­ scale, and the accurate quantification of landscape– grounds. Given this, the field appears to be producing a genetic relationships. confusing thicket of ideas. We are confident that over time the competition among ideas and approaches will lead to increasing clarity and, we hope, a true synthesis 14.3 CONCLUSION 2: SAMPLING of population and evolutionary genetic theory with NEEDS TO SPECIFICALLY TARGET spatial ecology. For now, we will offer our own view LANDSCAPE GENETIC QUESTIONS of some of the current and emerging challenges and opportunities, which we hope may focus and facilitate One of the largest limitations of most landscape genetics future progress in the field. Based on the previous book research conducted to date is that sampling is often chapters and other published literature, we believe that done without a priori consideration of expected land­ at least ten conclusions can be drawn about the current scape effects on genetic variation and underlying pro­ state-of-the-art in landscape genetics. cesses. However, sampling for landscape genetics can only be effective if it is based on these expectations, which should be stated as testable hypotheses (see 14.2 CONCLUSION 1: ISSUES below and Chapter 4). To derive such hypotheses, OF SCALENEEDTOBECONSIDERED we need to develop a theory that includes the multi­ faceted influences of landscape heterogeneity on Most landscape genetic studies have not carefully genetic variation (see conclusion 10). Nevertheless, considered the effects of spatial scale and the defini­ even if hypotheses are clearly formulated at the begin­ tion of the landscape, even though these aspects can ning of a study, the complexity of landscape genetic substantially affect our ability to detect relationships research will always be a challenge for optimal study between population genetic structure and landscape design and sampling. Thus, we strongly advocate sim­ features. As explained in Chapter 2, landscape defi­ ulations as a means to test different sampling options nitions that differ in thematic content, thematic before beginning a study, and to conduct a power resolution, and spatial scale may dramatically alter analysis to determine whether certain landscape– the statistical relationships between genetic varia­ genetic relationships can actually be detected within tion and landscape structure (e.g., Blair & Melnick a given study (Cushman 2014). Conclusion 5 249 14.4 CONCLUSION 3: CHOICE 14.5 CONCLUSION 4: SIMULATIONS OF APPROPRIATE STATISTICAL PLAY A KEY ROLE IN LANDSCAPE METHODS REMAINS CHALLENGING GENETICS We have not yet reached a true consensus on which Simulation modeling will play a vital role for the future analytical methods to use for the three analytical steps development of landscape genetics. Analysis of data is of landscape genetics described in Chapter 1. Reflec­ the foundation of empirical science, and is critical to tive of the rapid growth of a new field, there are a advance landscape genetics. However, when analyzing number of alternative methods currently in use to empirical data a researcher never knows the true process analyze landscape genetic data in node-, neighbor­ that governs the observed response. One can only infer hood-, and link-based frameworks (see Wagner & the process from the pattern of response and its associa­ Fortin 2013; Chapter 5). Few of these methods tion with one or multiple hypotheses. The great power of have been rigorously compared, and there is no gen­ simulation is that it allows this inferential pathway to be eral agreement as to which methods are best for a inverted. That is, in simulation modeling the researcher specific question. This is an area of utmost importance stipulates and controls the process being modeled and to advance the field and should receive very high then can generate the patterns of genetic structure that priority for future research. Given the complexity of would result from that process. This provides critical the task, we will probably never have a single method control over the pattern–process relationship that is that fits all research questions and data, but the essential to reliably evaluate such things as effect of conceptual framework suggested by Wagner and For- different landscape definitions, spatial scale, effectiveness tin (2013; see Chapter 5) helps to guide future efforts of alternative sampling schemes, and power of different for developing appropriate methods. To address some statistical methods (Cushman 2014). While there is of the challenges associated with existing methods clearly much more work to be done in developing used in landscape genetics, increasingly complex sta­ realistic and efficient simulation models for landscape tistical approaches are suggested. Unfortunately, some genetics, much progress has already been made of the more complex analytical approaches seem to be (Chapter 6). Importantly, simulations will also