Landscape Modeling for Forest Restoration Planning and Assessment: Lessons from the Southern Appalachian Mountains

Forest restoration is affected by two compounding problems. First, the goals of restoration often are unclear and involve the simultaneous consideration of a number of diverse criteria including aesthetics, biodiversity, recreation, and economic cost. Second, restoration management is a long-term process such that effects are difficult to measure at the time of implementation. Moreover, restoration goals and strategies may vary considerably according to region. Efficient and effective methods are needed to analyze the potential outcomes and impacts of multiple restoration goals and strategies for different forest types. Spatially explicit landscape models are promising tools for evaluating management alternatives, including restoration strategies in damaged forest areas (Gustafson et al. 2000, Urban 2006). Spatial issues are important for forest management and restoration (Bettinger et al. 2005). Spatially explicit landscape models are increasingly being used to contribute to the evaluation of management and monitoring in strategic forest planning (Mladenoff 2004, Perry and Enright 2006, Scheller et al. 2007). Recent examples of modeling in forestry include linking optimization models (linear programming) with spatial simulation models (Bettinger 2001, Gustafson et al. 2003), using spatially explicit models in assessing forest timber harvesting strategies (Gustafson et al. 2000), and strategic management planning alternatives (Mehtaa et al. 2004). We recently used LANDIS (landscape disturbance and succession; He et al. 2002, Mladenoff 2004, Scheller et al. 2007) to examine forest landscape dynamics and to evaluate restoration strategies for forests damaged by SPBs in the southern Appalachian Mountains (Waldron et al. 2005, 2007, Lafon et al. 2007, Xi et al. 2007, Cairns et al. 2008a). This region represents an important case for restoration strategies because biodiversity is high and multiple disturbances, including invasive pests, are reducing the existence of yellow pine forests (Southern Appalachian Man and the Biosphere 1996). The yellow pines here are four members of the genus Pinus (subsection Australes Loud.): Table Mountain pine (Pinus pungens), pitch pine (Pinus rigida), shortleaf pine (Pinus echinata), plus Virginia pine (Pinus virginiana). In this article, we provide a framework for using this approach to assist strategic forest management and restoration planning. We describe the methods we used to assess the succession of forests given a number of biotic, abiotic, and management scenarios relevant to the landscape. We present a case study in western North Carolina. Finally, we discuss the usefulness of this approach including both its strengths and its limitations. A Modeling Framework Forest Landscape Models. Forest landscape models simulate vegetation change through time using spatially referenced data across a broad spatial scale generally larger than a single stand ( 100– 1,000,000 ha). Spatial interactions between stands are a key component of such models. Therefore, these models can be used to simulate vegetation change with regards to multiple stand management within a given area (i.e., a watershed or a Ranger District of National Forest; Turner et al. 1994, Boston and Bettinger 2001). Such models have also been used to simulate the effects of a variety of timber harvesting practices (Gustafson et al. 2000), to project habitat loss and alteration resulting in impacts on biodiversity (Li et al. 2000, Akcakaya 2001, Shifley et al. 2006), and to assess the impact of land-use activities on ecological resources (e.g., Dale 2003). Restoration Goals. The goals of forest restoration may vary greatly depending on the temporal and spatial scales considered. In some cases, the goal of restoration may be known and well defined a priori. For example, the restoration goals for national forests often are clear and defined in land and resource management plans (e.g., Gustafson et al. 2000). Where the goal of a restoration strategy is known and well defined, a modeling approach can be used to determine an effective management strategy that leads to the landscape structure that best fits this goal. For example, the purpose of restoration may be to recreate a historical forest condition (supported by scientific data) or a forest structure crucial for the rehabilitation of an endangered species. Using spatially explicit landscape models, outcomes of various forest management and restoration practices can be evaluated according to various criteria, including the cost of the management strategy, economic gains, biodiversity protection, and resistance to pests and disease. However, the limitations of historical data (e.g., a full picture of a historic forest may not be known) and the fact that effective management involves optimal allocation of limited economic resources often makes determining a priori restoration goals difficult. In these cases, spatially explicit landscape models can help transform general or conceptual goals into more quantitative and realistic goals. An iterative modeling approach may allow the exploration of various management strategies and their consequences. This approach may facilitate the identification of appropriate, practical, and pragmatic restoration goals. Clearly, the success of this approach is dependant on a model that is flexible enough to represent a variety of disturbances and management practices, that is simple to use, and that accommodates available data. The Modeling Framework. Our approach is to integrate natural disturbance agents and restoration/management alternatives into a spatially explicit landscape model capable of simulating forest vegetation dynamics through space and time. We used LANDIS as our primary modeling environment because it provides a general framework for determining the combined effect of major natural and anthropogenic disturbances (e.g., fire regimes, insects, harvesting, and planting) that are capable of driving changes in forest structure (Figure 1). LANDIS is a stochastic simulation model that allows forest succession and disturbances to operate on cellular landscapes comprising heterogeneous patterns of tree species, age class, and seed dispersal (He et al. 1999, Gustafson et al. 2003, Mladenoff 2004). It has been used in North America as well in some areas of Europe and Asia to investigate forest dynamics and fire management strategies across landscapes. Examples of applications relevant to forest restoration include harvesting and fires (He et al. 2002, Shang et al. 2004, Scheller et al. 2007), pests and disease (Sturtevant et al. 2004), risk assessment and landscape habitat models (Larson et al. 2003, Shifley et al. 2006), landscape change and management practices (Shifley et al. 2006), and succession and harvesting (He et al. 2002, Gustafson et al.