Embedded Files
- Source Code for Location Model Same Income (542 KB)
- Source Code for Location Model No Geography (547 KB)
- Shapefile (18 KB)
The Effect Different Size Areas page highlighted how altering the size of small areas has effects on different spatial distributions. This page will investigate the effect small areas have on the spatial distribution of the agents by comparing the location model presented in the previous section (which was the same model used in the Effects of Space and Income and Location Model Same Income pages) to a variation of the location model where no small areas are imposed (Location Model No Geography). Within this model, agents first chose a location based on the aggregate information of the area (i.e. the urban entity itself, see getPolygonNeigbourAttributes in the LocationModel class) rather than just a specific smaller area (i.e. the getLocalNeighbourhoodAttributes method in the Location model class in the Location Same Income Model) which may or may not overlap two areas. If the agent can afford the desired space in the area (i.e. in terms of cost per metre2), it can move to the area, otherwise it queries the next area. As with the previous model, if the agent can afford an area it moves to the area and then searches the area for a suitable location. To ensure that the majority of the area is searched, it can move up to 2000 times in each area when searching for a suitable location (see calculateAttractiveAreas method in Resident and Employer classes). Each time the agent moves up to 75m in any direction from its previous position (see movement field and getMovement methods in LocationModel). However, while the agent’s movement is restricted to within the area when searching for a new location, when calculating the average cost per metre2 based on its local neighbourhood buffer (see GIS Operations in the Basic Model), this local neighbourhood can cross the boundaries of the area.
Within both models (i.e. the No Geography and Same Income), there are the same number and type of agents with the agents having the same ranges for space requirements and income. Additionally, the same areas are used (see The Effects of Space and Income Table 1 . and Figure 1), all model parameters are kept the same, and small neighbourhoods were defined as 75m. In each case, several models were run for 100 iterations and the results that follow describe the average conditions from these runs. A selection of animations can be seen below.
The effect imposed by using smaller neighbourhoods when choosing a new location can clearly be seen in Figure 1 which highlights the typical distribution of agents through the course of 100 iterations (animations of these model runs can be seen below). When no small areas are used, the agents appear less clustered and occupy more areas than when small areas are used. In both models, the outer two zones remain vacant (i.e. the ones which where initially unpopulated) and this is supported by an examination of aggregate results. Table 1 charts the number of empty areas, the number of agents searching, and the number of agents that have moved for both the small areas and no small areas models. The effect of using small areas rather than the actual polygons result on average in a greater number of empty areas through the course of a simulation run. Additionally, the use of small areas results in a greater number of agents who are dissatisfied with their area (i.e. too expensive) and this is revealed in the number of agents searching for a new location compared to the case when no small areas are used.
Figure 1: Typical configurations of agents during the course of a simulation for models where small areas are used and not used.
In both models, the total number of moves made during the course of the simulation and during the model iterations is similar. This is to be expected as agents have to move; for employers this is every six iterations (i.e. when their tenure is up) while for residents it is based on a probability depending upon age of the agent (see movement in the Location model description). The number of moves made by the residents between every 10 iterations decreases over time due to the fact that the agents are becoming older and therefore less likely to move.
Table 1: Average results when no small areas and small areas models are used.
Further examination of the effect of small areas is presented in Table 2 which shows the average number of different types of agents in different zones during the first 100 iterations of the simulations and the differences between the two models (zones 7 and 8 are not included as they are never occupied by agents). In both models, the number of agents occupying the outer zones (5 and 6) decrease over time. However, it is only when small areas are used that both zones 5 and 6 become devoid of agents by 100 iterations. By 100 iterations, only the residents within the central area (zone 1) are industrial workers – who require the least amount of space, while the use of no small areas forces commerce residents to zone 4 as these residents require largest amounts of space. Table 2 also supports Figure 1 where using small areas, the agents are more clustered in the central areas compared to the case when no small areas are used. This can be seen where there are negative difference values in zones 4 and 5.
Table 2: Average number of agents in different zones over the course of multiple simulations for small area and no small area models.
Animations of the typical configurations of agents during the course of a simulation for models where small areas are used and not used.
Animations of the typical configurations of agents during the course of a simulation for models where small areas are used and not used.
Page updated
Google Sites
Report abuse