Basic Design

The landscape object described in this chapter was conceived from a single fundamental idea: any given portion of the Earth's surface is a unique location, which reacts to localized processes with no notion of the global behavior of the landscape. For a given portion of land, the processes that govern such things as whether vegetation will grow on it, soil will erode, or material will burn is largely a function of factors that have to do with localized conditions on that portion of land and its immediate neighborhood. It is safe to say with some generality that vegetation growth in a particular piece of land is more a function of the local conditions of soil type and moisture in than the fact that the majority of the landscape is populated with similar vegetation. Groundwater flow through a (small) unit area of land is ultimately a function of the permeability and saturation of that parcel of land, and the hydrolic head of the parcels of land immediately adjoining it. While some global factors (say, windspeed, air pressure, or rainfall) may matter, much of the behavior of a unit area of land is going to be decided at the most local level.

Given this basic idea, the idea was to construct a landscape whose basic design would be like a raster image. In a raster image, the entire ``world'' or picture is described by the coordinates of a particular cell (or pixel) and the value (color) of that cell. The cell represents a unit of surface area, and for the purposes of the model, the properties of the landscape are homogeneous within the boundary defined by that cell. Landscape variability is represented by having a sufficiently large number of cells, much the same way as subtleties of color in an image can be captured with a large numbers of pixels.

  

Figure 1: Cellsurface object with models in cells

In the landscape object described here, each raster cell would be replaced by an object that contains a number of state variables that describe the qualities of that cell that one wishes to model, as well as whatever functions are required for the cell to act upon those variables. In short, each cell would become an independent ``model'' of the processes that one wishes to study that exact location (figure 1). The model within the cell itself may be aspatial in nature, and described as a flow chart much as the FORTRAN-based Forrester models ²; as with the raster image, spatial variability would be captured in the large number of cells running concurrent models over the landscape. In this conceptualization, each cell is a virtual processor capable of executing a model relevant to its (potentially unique) situation; the landscape then is a collection of thousands of cells acting in parallel, each one accepting conditions of its neighbors as inputs into its own execution, but executing its functions according to its internal rules. If the conditions and behavior of a cell is a function of its neighbors' states, then the actions of the landscape can be characterized by CA rules.

In practice, implementation of this idea can be constructed from a combination of a few basic building blocks:

• a basic cell object which performs operations
• a list for keeping track of the objects
• a scheduling mechanism for calling the operations in the objects
• a data file manager for file input and output

In the case of the CA-surface, these generalized objects became:

1. the cell
2. the landscape
3. the scheduler
4. the data file manager

The purpose of each of these objects should be explained in greater detail.