One of the underlying premises of ecosystem management (Samson and Knopf 1996, Boyce and Haney 1997, Kohm and Franklin 1997, Vogt et al. 1997) is that ecosystems and landscapes are dynamic; that disturbance (both natural and anthropogenic) and succession processes create patterns in the distribution of resources that feed back to affect these and other ecological processes and maintain ecosystems and landscapes in a constant state of flux (Turner 1989a). Adopting this ‘dynamic view' and incorporating it into land management strategies has become a major challenge for land management agencies (Franklin 1997). Recently, in fact, it was proposed as a mandate for the U.S. Forest Service (Federal Register, "National Forest System Land and Resource Management Planning. Final Rule 36 CFR Parts 217 and 219" Federal Register 65: 65714-65781). Although this ‘dynamic view' has become the operating paradigm in landscape ecology and land management (Turner et al. 2001), it is largely understood in concept only; a comprehensive quantitative framework still eludes scientists and managers, primarily due to a lack of quantitative studies (both empirical and theoretical).

       Landscape ecology is based on the premise that landscape patterns influence ecosystem processes over a wide range of spatial and temporal scales (Urban et al. 1987, Turner 1989, 1990, Forman 1995, Turner et al. 2001). Processes such as disturbance and succession play a central role in structuring ecological systems by producing a spatiotemporal mosaic of vegetation patches in different successional states (Paine and Levin 1981, Sousa 1984, Krummel et al 1987). This shifting mosaic of vegetation patches has important implications for landscape function, including the ability to support viable populations of plant and animal species (Forman 1995). The distribution of species within this shifting mosaic depends upon an interaction between species' life history traits and the spatial and temporal structure of the environment. Species with different life history traits and habitat associations may respond very differently to landscape changes (e.g., Hansen and Urban 1992). Landscape changes may cause some populations to decline or be subdivided into disjunct populations due to direct loss of suitable habitat and/or loss of connectivity among habitat patches, as in metapopulations (Levin and Paine 1974, Gilpin and Hanski 1991, Weins 1997) and source-sink populations (Pulliam 1988). Conversely, the same landscape changes may cause some populations to increase in size and/or viability due to an increase in the amount and connectivity of suitable habitats. These varied responses to landscape change make it exceedingly difficult for land managers to understand the ecological consequences of proposed management actions.

      Our study was motivated by the need to provide a better quantitative understanding of landscape dynamics and was stimulated by four basic information needs.

      First, the concept of natural variability has emerged as a paradigm for ecosystem management, species conservation, and landscape restoration in western North America (Cissel et al. 1994, Morgan et al 1994, Swanson et al. 1994), and it is increasingly believed that maintaining landscapes within their "natural" range of variability is an effective “coarse-filter” approach to maintaining ecological integrity (Landres et al 1999, Romme et al. 2000, Buse and Perara 2002). This approach is based on the premise that ecosystems are naturally dynamic and that native species have adapted to disturbance-driven fluctuations in their habitats (Bunnell 1995). Therefore, the potential for survival of any given species may diminish if temporal and spatial patterns of species’ habitats shift outside their natural range of variation (Hessburg et al. 1999). An estimated range of natural variability can provide a frame of reference for assessing landscape patterns and processes over an extended period of time (Swetnam et al. 1999, Wimberley et al. 1999). One advantage of this approach is that it provides a basis for understanding ecological systems and the changes occurring in these systems, as well as for evaluating the long-term consequences of proposed management actions (Landres et al. 1999, Wimberley et al. 1999). Although the notion of "natural" in ecological systems is equivocal (Sprugel 1991), it is common to define an historical reference period during which human activities had relatively minor effects on overall landscape structure and function, and use this as a benchmark for comparison with contemporary or potential future conditions (Landres et al 1999, Romme et al. 2000, Buse and Perara 2002). A prerequisite to this management strategy is knowledge of the range of variation in key landscape attributes during the reference period, hereafter referred to as the historic range of variability (HRV). Moreover, a quantitative understanding of HRV is essential if we are to know whether recent human activities have caused landscapes to move outside their HRV (Landres et al.1999; Swetnam et al. 1999). See Romme et al. (2003) for a detailed discussion of the concept of reference conditions.

      Second, the notion of quantifying landscape dynamics immediately invokes issues of scale. Indeed, it is well known that observed ecological patterns depend largely upon the scale that is chosen for observation (Wiens 1989). Thus, it seems intuitively obvious that in a given landscape subject to a specific disturbance regime (i.e., the spatial and temporal distribution of disturbances), the measured range of variability will depend on the size of the landscape relative to the size of the disturbances and the length of the time period over which landscape behavior is examined. Turner and colleagues addressed this issue theoretically using a simple landscape simulation model to address the question of how we might expect systems to behave over time, given a specific disturbance regime and a particular reference area (e.g., study area) (Turner et al. 1993). They noted three things: (1) that the characteristic dynamics can be predicted from the relative scaling of the disturbance regime, (2) that disturbance-driven landscapes might be equilibrium, quasi-equilibrium, or inherently nonequilibrium; and (3) that anthropogenic influences may rescale these and change the qualitative dynamics of systems (e.g., fire suppression rescales a fire regime). Despite the clarity of this theoretical framework, few studies have explored this relationship empirically (e.g., Wimberly et al. 2000, McGarigal et al. 2001).

      Third, myriad evidence exists that landscapes are naturally dynamic, fluctuating in structure and function over time in response to the interplay of disturbance and succession processes (e.g., Romme 1982, Romme and Despain 1989, Baker 1992, Wallin et al. 1994 and 1996), yet there is a paucity of evidence that dynamism itself is essential to the maintenance of ecological integrity. It is well known that landscapes and the populations they contain are dynamic over time (Forman 1995, Turner et al. 2001). Yet, many landscapes contain areas that continually escape some disturbance processes, while other areas are disturbed at a much higher rate. Such a wide range of temporal dynamics may have profound effects on population persistence (Fahrig 1992, Reeves et al. 1995). For example, during times of low habitat availability, areas that continually escape disturbance may function as refugia. When habitat area increases, these refugia may function as source areas for population expansion. Therefore, certain populations may be as dependent on the disturbance patterns determining the temporal structure of habitats as they are on the spatial distribution of habitats. Unfortunately, while many studies have addressed the spatial structure of habitats (Kareiva and Wennergren 1995, Tilman and Kareiva 1997), few studies have addressed the temporal structure of habitats (Fahrig 1992, Johnson 2000, Keymer et al. 2000) or the importance of these temporal dynamics to population persistence (Fahrig 1992, Reeves et al. 1995).

      Fourth, only by understanding and quantifying the long-term dynamics of landscapes can we provide the proper perspective for interpreting current and future landscape patterns and make informed land use decisions consistent with the goals of ecosystem management (Gustafson 1998, McGarigal 2002). Due in part to widely available software programs like FRAGSTATS (McGarigal and Marks 1995, McGarigal et al. 2002), it is now quite easy to compute a wide variety of landscape metrics for categorical maps. This has helped advance the field of quantitative landscape ecology and has provided land managers with a means of quantitatively evaluating the impacts of alternative management scenarios on landscape structure. Unfortunately, quantifying landscape structure is fraught with many difficulties which are often overlooked or not given serious consideration. Perhaps the single greatest difficulty in interpreting landscape metrics is the lack of a good reference framework. Often we wish to interpret the meaning of a landscape metric computed for the current condition of a landscape, yet that interpretation is inextricably linked to the historic pattern of variability in that metric. Developing a quantitative understanding of the metric behavior under historic reference period conditions is prerequisite to an informed ecological interpretation of the metric and of the current and future state of the landscape.

Literature Cited