9. Limitacions i prospectiva de la recerca
9.2. Perspectives de futur
where , is the transition potential for land use g on cell i, , is the neighbourhood effect for land use g on cell i, , is the physical suitability for land use g on cell i, , is the zoning status for land use g on cell i, , is the accessibility for land use g on cell i, and ν is a stochastic perturbation term. The latter is added to represent the different preferences that individual actors have and to account for variation in factors that are not otherwise represented. Time dependent variables are indicated with superscripts, where t indicates that information is taken from the existing situation, and t+1 indicates that this information is used for activity allocation in the next time step. Factors that represent physical suitability, zoning and accessibility can have values between 0 and 1.
The neighbourhood effect is the dynamic component of the CA algorithm which accounts for the self‐organizing behaviour of the model. It is calculated from the land uses in the neighbourhood of a location:
, ∑ , , , Equation 6.2
where wd,g,l is the neighbourhood rule that describes the effect of land use l at distance d(i,j) on the potential of location i for land use g . This land use taken from the previous time step. Locations that do not have a feature of constrained land use will get the unconstrained land use for which they have the highest suitability.
6.3.2 Adding activities to the Metronamica framework
In the original Metronamica modelling framework, land uses are constrained by an area demand. However, in reality many land uses are not constrained by an area, but by an amount of activity instead. In the case of urban expansion it is not an amount of land surface that needs to be covered by urban land, but the population that needs a place to live. Therefore, in the activity‐based approached, the total demand is not constrained in terms of cells, but in terms of activity instead. Hence for each time step the amount of activity that needs to be allocated on the lattice is defined exogenously. After the activity allocation, land uses that are associated with activities are assigned based upon the new activity distribution. The number of cells per land use then depends on the density of the activity: a high density requires fewer cells than a lower density.
Additionally, the land uses in the surrounding of a location are input to the neighbourhood effect in the original constrained CA. For example, commercial areas in the vicinity make a location more attractive for residential development as they represent jobs and services. As quantitative information is available in the activity‐based model, it is now possible to define the attraction as a function of the number of jobs rather than just the presence of commercial land use.
Hence locations that have a high job density in their commercial areas are more attractive to residents than locations with only a low density. At the same time, the compatibility of existing land use will influence the amount of activity that can be allocated on a location.
Since not all land uses can be associated with an activity, we have three types of land uses in our model. These are activity‐constrained land uses, area‐
constrained land uses and unconstrained land uses, where the activity‐
constrained land uses are added to the existing Metronamica framework.
Activity‐constrained land uses are assigned based upon their activity distribution, area‐constrained land uses are allocated after that, and finally, all cells that are not assigned one of the constrained land uses get the unconstrained land uses. The order of allocation of land‐use types represents the economic influence or power that is associated with these land uses.
An obvious example of an activity‐constrained land use is residential land use, with population as the associated activity. It should be noted that each activity constrained land use has only one associated activity. An example of an area‐
constrained land use is agriculture, for which the demand is expressed as a number of cells. Unconstrained land uses are often natural land uses, which occupy all locations that are not occupied by land uses which are driven by an external demand. Similar to the original Metronamica framework, the model requires at least one unconstrained land‐use class to make sure that all cells will have a land use at any time.
To keep track of activities, an additional data layer is required per type of activity. Hence, a cell has no longer only one discrete cell state. Instead it has a (e). Activity‐constrained land uses are then assigned based on the updated activity distribution (f). Similarly, the total potential for area constrained land uses is computed based on the land use and activity distribution from the previous time step (g and h) and the suitability, zoning status and accessibility of
that location (i). These land uses are allocated accordingly (j) until the area demand (k) is fulfilled.
Figure 6.1: System diagram for land use and activity distribution. Arrows show the flow of information, where solid lines represent current values and dashed lines represent values from the previous time step. Other symbols are explained in the text.
An important characteristic of the proposed activity based model is that it is a generic model that aims to simulate urbanization from the bottom up. According to urban economic theory, urbanization is the result of the interplay between centripetal and centrifugal forces (Krugman 1996; Furtado et al. 2012). The activity‐based model incorporates both forces. Agglomeration effects are
simulated by the mutual attraction of activities creating economies of scale.
However, this agglomeration advantage will generate competition for space which leads to higher land prices, congestion, and pollution that are associated with higher activity densities. These diseconomies of scale cause centrifugal forces. Both forces are included in the proposed activity based model for which details are provided in the next section.