Selecting the appropriate scale is a critical component of model selection, design, and application. Model scale considerations include the total areal extent and the spatial resolution or grain of modeling units. For landscape models that represent the evolution of patterns or processes with time, temporal period (extent) and time-step (resolution) must also be considered. While it is clear that inappropriate selection of model scale can limit model applicability, landscape models can also be used to identify characteristics scales of processes and patterns of interest. For example, disturbances have characteristic time and space scale, and models have been used to show that the scale oflandscape patchiness is both a response and driver of disturbance regimes. Models also range from top-down approaches, which make inferences from broad regional patterns to bottom-up approaches that attempt to re-create larger-scale patterns through the integration of point (fine-scale) processes.
A dominant feature of the current generation of landscape models is the use of maps derived from remote imagery (e.g., Landsat, SPOT, etc.) to represent the dominant features (e.g., soils, habitat types, roads, rivers and lakes, etc.) affecting ecological processes (remote sensing). Because these maps are formatted as a grid of rows and columns, the resolution and extent of the map defines the spatial dimensions of the model. The extent of the map, or total area of the landscape being considered, is usually under the control of the model user with the primary consideration being the inclusion of an area large enough to contain all important landscape attributes. The resolution or grain size is usually the more critical variable, determining the dimensions of the smallest resolvable map element (pixel). Because map resolution is usually set by the design of the remote-sensing instrument (e.g., 30 m for Landsat imagery, 10 m for SPOT), each grid cell must be regarded as a single homogeneous landscape unit. Therefore, ecological processes that depend on fine-grained interactions must use an equally fine-grained map. For example, although Landsat imagery is a common choice for model applications, there is no a priori reason why ecological phenomena should be expected to be well represented by Landsat's 30 m resolution. Those using landscape models should be familiar with the rich literature on issues of scale in ecology (scale) and should carefully consider the effect that both grain and extent may have on modeled pattern-process relationships.
Was this article helpful?