Boundary-fitted grid in landscape modeling

Abstract

Landscape modeling requires the delineation of system boundaries and interior features. Quite often, these components are complex and difficult to accurately represent. A rectangular grid is used to represent the study and adjacent non-study areas in most cases. When the non-study area occupies a large portion of the grid, computer memory is wasted, and computational time increases. An elliptical grid generator for non-orthogonal curvilinear coordinates is used to generate a boundary-fitted grid for a landscape model. In a boundary-fitted grid coordinate system, one coordinate axis follows the landscape domain boundary and is non-orthogonal to the second axis. The boundary-fitted grid uses elliptic partial differential equations to distribute grid points inside the landscape domain. Although the boundary-fitted grid follows the domain boundary, the grid pattern and point allocation remain structured. Thus, a landscape model can use a boundary-fitted grid without changing the model’s data structure or the computational scheme. In this study, a boundary-fitted grid and a raster-based grid were applied to the Everglades Landscape Fire Model. Use of the boundary-fitted grid decreased model simulation time by about one fifth and computer storage by 58% relative to the raster-based grid. Also, the linear characteristics of interior geographical features such as rivers and airboat trails were preserved by the boundary-fitted grid, but not by the raster-based grid. This preservation provided a more reasonable base map for simulating ecological processes, such as fire across heterogenous landscapes.