Focused Assessment of Scale-Dependent Vegetation Pattern

Ecological processes frequently occur at multiple spatial scales simultaneously. For example, fires imprint the landscape at a variety of spatial scales, from small areas of high burn intensity due to patchy surface fuels, to large stands within fires that escape conflagration entirely (Fig. 7.1). These types of complex disturbances can increase environmental heterogeneity and thus species diversity by creating a variety of microhabitats and by increasing patch diversity (Romme and Knight 1982; Christensen 1985; Denslow 1985; Pickett and White 1985; Turner et al. 1998). The flow of organisms, genes, and populations provides another excellent example of scale-dependent ecological processes (see Chap. 8). Because ecological data are scale-specific, any model based on these data will provide inferences at a specific spatial scale. The scale of inference of statistical models is a critical consideration in predictive modeling (see Chap. 4). Some confusion can arise if the variables used to build a model are themselves scaled differently. For example, soils variables and terrain features have different characteristic scales (Urban et al. 2000). As a consequence a regression based on these data would be multi-scaled (Lookingbill and Urban 2004). Peters et al. (2004) refer to this approach to landscape modeling as spatially implicit, a nonspatial model built from geospatial data. By far, most models of species distribution are spatially implicit. By contrast, a spatially explicit model would include predictive variables that account for spatial processes explicitly. From a modeling perspective, few approaches are currently available to ecologists to account for scale explicitly. Multi-level modeling (Gelman and Hill 2007) offers a framework for multi-scale