Representing territory

Central to the sfr-abes model is the territorial behaviour of agents. Agents have a spatial location, and can sow and harvest crops in their vicinity. They will generally occupy structures for habitation at their location (ie homes). They can provide and receive services (eg haircuts) from other agents when they are at the same location. They can use goods at their location. Goods can be moved between locations.

Sfr-abes must have a coherent model of the territory and its locations. The implementation must represent this in a manner suited to access while executing agent code.

In the economy, territory has significant features at a wide range of scales. A prairie might be hundreds of square kilometres and have a rather mundane agricultural function. An inner city, on the other hand might house many thousands of people per square kilometre, and be highly connected to transport systems. Areas of interest may be separated by large distances which are economic voids.

Representing territory as a uniform grid is rejected as likely to be inefficient and unscalable. Non-uniform or multi-level grids might be a viable approach. Sfr-abes intends to use a generalised network, in effect a bunch of nodes with whatever auxiliary data structures might be needed for particular types of access.

Associated with each node is a bunch of data comprising things like:

• Spatial coordinates (in 2-D, (latitude, longitude)?)
• Surface area, climate, vegetation/crops
• Freshwater coverage (proportion), average depth
• Saltwater coverage (proportion), average depth
• Subsurface mineral content and accessibility - oil, coal, gas, iron ore, bauxite, copper etc
• A list of physical capital (eg grain, fuels, fertiliser, buildings)

Perhaps these are some intrinsic and immutable properties of the location, and a dynamic representation of quantified natural and artificial resources.

An agent sowing a crop will need to use up grain at the location, and convert it into an (ungerminated) crop. As time progresses, the crop will develop until it can be harvested by an agent, and the crop removed becomes grain.

As the simulation becomes more complex, new types of natural resources and capital can be defined.

For display purposes, a user needs to be able to see a representation of the territory, perhaps with a variety of visualisation strategies. A simple approach would be to assume locations cover area implied by a Voronoi tessellation, shaded by a blend of blue, yellow, green, grey, white representing water, crops, forest, buildings and snow. Iconic representations of the resources and capital can be overlaid.

Travel time/cost between locations is a crucial aspect of the simulation, since this affects the ability of agents and capital to move. In the first instance, the travel time can be proportional to the distance between coordinates, assuming a walking pace. Ideally, real-world travel time estimates would be used. In this manner, questions about the economic impact of a particular infrastructure proposal could be examined.

One means to help initialise the territory representation would be to download a detailed, cloudless composite satellite photograph of the earth, such as these while ensuring only royalty-free images are used. Images are available for vegetation, rainfall, light emission (at night, primarily urbanisation).