Squatter Settlements Consolidation:Spatial Analysis in an Agent-based Environment

Squatter settlements consolidation:

Spatial analysis in an agent-based environment

Fabiano Sobreira Faculdade de Ciências Humanas-ESUDA, Brazil

Keywords squatter settlements, agentbased models, complexity, economy movement, consolidation, fragmentation. fjasobreira@yahoo.co.uk

Abstract In this present paper the development of inner-city squatter settlements is analysed at two complementary scales, through distinct agent based models developed by the author: AxialAgents and Favela. The first one discuss the arising and development of spontaneous settlements in a virtual city centre where some aspects of the typical structure and dynamics of Third World cities are reproduced. In particular, its irregular and fragmented structure, sorted by open spaces (usually non-occupied public land), model processes which are usually the basis for inner-city squatter settlements’ development. This process is analysed according to the relative location of these vacant sites, in relation to the grid structure and to the existence of movement generators, through AxialAgents, an heuristic agent-based model in which decentralised agents run over an axial lines like environment searching for available space to settle. The second approach is developed at a smaller scale, and is related to the development of the squatter settlement itself. For this purpose we use Favelas, an agent based model which simulates, in a heuristic basis, the process of growth and consolidation of squatter settlements according to decentralised agent based movement from the edge (attractive and non-attractive boundaries) of vacant sites. Both models were built with heuristic purposes, based on random walk dynamics and built in StarLogo, a parallel programming tool developed by the Epistemology and Learning Group of the Massachusetts Institute of Technology.


1. Introducttion Recent studies in Space Syntax have pointed out that the level of consolidation and "the spatial configuration of settlements, and particularly the way in which the site is embedded in the larger spatial structure of the city, determines whether the spontaneous settlement is successfully incorporated to the city body or remains ‘slumlike’" (Hillier et al., 2000: 62). Other studies (Mujhika, 2001) argue that the location of settlements, the land use in them, the settlements’ layouts and the size of the lots within them, all can impact on the success of upgrading processes.

Proceedings . 4th International Space Syntax Symposium London 2003

Actually, when analysing the development of squatter settlement areas in third world cities, one can observe a feedback process: the urban growth determines the consolidation of spontaneous settlements and this consolidation changes the course of urban growth process, contributing to a fragmented urban form.

In this paper, special attention is dedicated to inner-city squatter settlements that are located close to the most integrated urban areas and generally present high densities, since most of these settlements are consolidated and have reached the limits of growth (UNCHS, 1982). These settlements are usually bounded by roads, which are highly integrated in relation to the overall structure of the city. Because they develop in spatially restricted non-occupied land, these settlements face spatial restriction, and therefore tend to present a highly dense spatial structure. Besides, the proximity of central sites to job opportunities makes them a more competitive


and overcrowded site (Figure 1).

Figure 1: Squatter settlement in Caracas, Venezuela
(source: author's collection)

In this paper, the inner city squatter settlements’ development is analysed at two complementary scales, through two agent-based models: AxialAgents and Favela. Both models are based on random walk dynamics and are built in StarLogo, a parallel programming tool developed by the Epistemology and Learning Group of the Massachusetts Institute of Technology for agent based heuristic experiments (Resnick, 2000).

Through AxialAgents we discuss the arising and development of spontaneous settlements in a virtual city centre where some aspects of the typical structure and dynamics of Third World cities are reproduced. This process is analysed according to the location of these void spaces (non-occupied land), in relation to the grid structure and to the existence of movement generators. AxialAgents is an heuristic agent based model in which decentralised agents run over an axial lines like environment, searching for available space to settle.
Squatter settlements consolidation

Favelas is also an agent based model which simulates (or illustrates), on a heuristic basis, some aspects of the process of settling dwellings inside non-occupied land surrounded by attractive and non-attractive boundaries. This approach is developed at a smaller scale, and is related to the development of the squatter settlement itself. We analyse the growth and consolidation of squatter settlements according to decentralised agent based decisions.

Before going to the models analysis, some considerations must be presented, about complexity as a theoretical basis for agent-based models construction.

2. Complex, but not complicated The morphological structure of the city is built from the interplay of different dynamics, offering an extra level of complexity to these systems. As Holland (1995: 1) suggests, “a city’s coherence is somehow imposed on a perpetual flux of people and structures?. From Holland’s words one can identify two different kinds of fluxes: the flux of people and the flux (or change) of structures. The ever-changing nature of cities, however, seems to require both interpretations for a better understanding. Not only it is necessary to understand the complexity of its structure, but it also seems to be necessary to understand the movement of people in it.


Spontaneous settlements are clear examples of complex subsystems within a complex urban system. Their morphological characteristics combined with their development process are misunderstood as chaotic and unorganised, when complex should be the appropriate designation. And so are Third World cities, traditionally known for their apparently chaotic feature, because of their discontinuous spatial patterns and rapid and fragmented development process.

The idea of a structure emerging from a bottom-up process where local actions and interactions produce the global pattern has been widely developed through Emergence based tools and technologies. Although a number of models have been developed using agent-based techniques to simulate urban scenarios, including land use, pedestrian modelling, and so on, the use of agent-based simulation to urban spatial change is not a consensus in the research community, specially when they are produced for deterministic instead of heuristic purposes. Indeed, agent-based models can give rise to too much complexity making it difficult to track elements and relationships within the model and thus, making difficult any kind of conclusion. But the choice of increasing the degree of complexity of the model or keeping it simple depends entirely on the researcher and the purposes of the model in hand. So, from this perspective, agent-based models are more suitable to simulations focusing on human behaviour in a given spatial environment. Viewed as such, they open up
Proceedings . 4th International Space Syntax Symposium London 2003

an avenue for analysis of dynamic processes that link spatial development with social issues. This kind of analysis is of fundamental importance when dealing with cases characterised by a highly decentralised development process, as is the case of urban growth in the Third World (Barros and Sobreira, 2002).

3. AxialAgents 3.1 The movement economy of third world cities The Movement Economy, proposed by Hillier (1996), relates spatial structure to social development. This is the basis of 'natural movement', which is the relationship between the structure of the urban grid and movement densities along lines in an urban structure: “natural movement is the proportion of movement on each line that is determined by the structure of the urban grid itself rather than by the presence of specific attractors or magnets? (Hillier, 1996: 161). So, according to Hillier, the


structure of the grid itself accounts for much of the variation in movement densities. In this sense, considering a city from the axial map perspective, the most integrated streets would be lined by the most developed areas, while the less integrating streets, as the most segregated areas, would line the poorest areas of the city.

But third world cities are characterised by a typical fragmented structure (Balbao, 1993). The urbanisation process in cities of developing countries is often insufficiently planned and poorly co-ordinated. The morphological result is a fragmented set of patches, with different morphological patterns often disconnected from each other. This fragmented pattern has its origins in the successive superposition of different urban typologies, including planned areas, spontaneous settlements, housing tracts, slums, vacant sites, institutional areas, shopping malls, informal town centres and so on. The Third World city is the result of the combined dynamics of fragments that are in constant mutation and evolution. It is fairly common to find, in highly integrated areas, lots of non-occupied land, some of them being occupied by squatter settlements, as well as less integrating areas occupied by high class private residential areas.

Spontaneous settlements fill some of the gaps in this erratic development, at the same time creating obstacles for any attempts to rationalise the development process and introduce effective land-use control measures (UNCHS, 1982). Spontaneous settlement is not only one of the elements that compose this fragmented pattern but also it is a fragmented object, composed by smaller fragments. It leads us to suppose that along with structure itself, other complementary elements could be considered when analysing squatter settlements in Third World cities.

Squatter settlements consolidation

According to Pacione (2001: 497), “the form of squatter dwelling and their location in the city vary. Most cities have squatter communities in their central areas adjacent to the major employment sources?. An example of that spreading dynamics is the city of Kuala Lumpur, in Malaysia. In the figure 2.C one can see that squatter settlements spread along the main roads of the city, from the city centre to the periphery. Squatters usually follow job opportunity areas, and then look for available and affordable places close to these areas. The following diagrams present some clues about these movement processes. In figure 2.B - Pacione (2001) - one can observe migration patterns among low-income neighbourhoods in a Third World city. There is a rural migrant flow (A) that leads people to low-income inner-city slums, searching for job opportunities and places close to them. From this point, when no affordable space is found, people move to intermediate and peripheral zones (L2). A similar approach is suggested by the diagram in figure 2.A, from which one can speculate that "land availability and proximity of high-intensity mixed land use, usually jobs opportunities, are the major controlling elements in spontaneous settlement location" (Dwyer, 1975).


Figure 2: (A) movement dynamics in Third World cities, source: Dwyer, 1975; (B) migration movement patterns, source: Pacione, 2001; (C) Kuala Lumpur squatter settlements map, source: Dwyer, 1975

So, despite the fact that there is no generally accepted or consensual theory about spontaneous settlement location, there is an agreement that because of the discontinuous, fragmented and decentralised growth process in Third World cities, the urban structure itself is not enough to analyse development patterns, as the poor and rich areas are located both in segregated and integrated areas. For example, it is common to find, in Third World cities, high class residential neighbourhoods located in highly segregated areas of the urban grid, while the poorest slums can be found along the main streets of a highly integrated and, at the same time, decaying city centre (UNCHS, 1982).

So, it seems that central business districts, commercial corridors, employment centres, transport connection points (train and bus stations) are complementary elements to be considered along with the spatial configuration of the grid, when
Proceedings . 4th International Space Syntax Symposium London 2003

analysing squatter settlements’ development in highly integrated areas. It seems important, therefore, to better understand the consolidation of inner-city squatter settlements (already located in highly integrated areas), to consider as complementary parameters two key elements: movement generators and people flux. Both of them can be analysed from the agent-based perspective.

3.2 An agent-based model in an axial like environment Through AxialAgents we analyse the interrelationship between the growth of spontaneous settlements and the movement flows in the urban grid. It is based on random walking agents which are set to search available spaces (in central nonoccupied land) to build their dwellings, in an axial like environment which represents a highly integrated area (city centre, for example) in a Third World city.


The agents’ movement depends essentially on randomised walking parameters and on the co-ordinates of movement generators. The model interface allows the user to control parameters, as well as these start-up co-ordinates, the number of agents running over the virtual environment, and the consolidation limit.

The consolidation limit parameter controls an important consolidation rule. This rule refers to a process in which spontaneous settlements are gradually upgraded, and, as time passes, turn into consolidated Favelas or, in other words, spontaneous settlements that are harder to be evicted. So, as a result of the introduction of the consolidation logic, it is possible to speculate about differences of development between settlements, according to the “time? (computational iteration) variable.

Technically speaking, the consolidation process is built into the model through a “cons? variable (Barros and Sobreira, 2002). This “cons? variable has its value increased at each iteration of the model and, at a certain threshold (defined by the user), the patch turns into the consolidated state (represented in the open spaces by the dark grey cells). In resume, once the "go" button is pressed, the model starts running and the agents go walking through the axial like structure, surrounded by open spaces ready to be occupied. The user has the option to hide or show the agents while running the program. Another element present in AxialAgent is the “flux gradient?, that is a spatial representation of the most visited streets in the system. Each time an agent passes through a cell, it adds value to the flux gradient variable of that cell. The resulting map is a network of streets varying from black (high flux) to light grey (low flux). This mapping process can be used as a complementary tool along with the integration maps of space syntax. Both of them illustrate street hierarchy according to movement pattern; the first one through a dynamic mapping and the latter by a static graphic representation.
Squatter settlements consolidation

Figure 3 shows a sequence of outputs originated from AxialAgents. Figure 3.A, the irregular grid that is the virtual environment that the agents search is presented. It is composed by lines and open spaces (both represented by light grey color) and represents a highly integrated section of an urban centre in a Third World city. The following parameters are common to the three experiments shown in Figures 3.B to 3.D: 1000 agents running in a random walk, defined by a random angle between +93º and -93º; consolidation threshold: 8; and final snapshots at 407 iterations (programming time). The only difference between the three experiments is the start up co-ordinate. In experiment 1 (Figure 3.B) the agents’ search starts up from coordinate (0,- 3), leading to a development pattern in which 628 cells were left as open spaces and 506 were occupied, resulting in a average density for the overall system of 44%. In experiment 2 (Figure 3.C) the origin co-ordinate is (12, -3), that is quite close to the previous one, but resulting in a distinct pattern of development. The average density in this case is 53%. Finally, in experiment 3 (Figure 3.D), the movement generator is situated in co-ordinate (-11, -11), and the snapshot shows a development where 725 cells refer to open spaces and 409 to occupied plots, resulting in an low average density of 36% (probably because of the relatively low integration of the co-ordinate).






Figure 3.A-3.D: AxialAgents outputs

From the experiments some observations can be made. Firstly, a quite obvious result that is confirmed by the model: by varying co-ordinates of the movement generators and keeping the same axial lines structure, we find distinct patterns of development and consolidation. This seems to reinforce the pre-supposition that movement based on people flux and movement generators can be an additional parameter to be considered, when analysing settlements that present a similar degree
Proceedings . 4th International Space Syntax Symposium London 2003

of integration in the structure. So, we conjecture that for each new co-ordinate of generators, there is a new consolidation map and development pattern. In other words, in a real urban centre of Third World cities, even keeping the same axial structure, if some generators, such as employment sources, central transport stations, business and services areas, are displaced or inserted into the structure, there will be a difference in the development pattern of informal growth.

So, if through integration maps it is possible to make conclusions about squatter settlements’ consolidation, through AxialAgents or other similar agent based models it is possible to speculate about development patterns, which will affect their configurational properties, growth tendencies and even details on consolidation degrees.


4. Favela 4.1 Growing from the edge Inner-city squatter settlements grow and consolidate from the edge (Sobreira, 2002). They are dynamic systems and result from the decentralised behaviour of local agents, that can be illustrated by generative models based on boundaries as attractors and on emergence as a process. It reinforces a conjecture (Sobreira and Gomes, 2001) according to which in spatially constrained squatter settlement development, boundary is the cause, packing is the process and diversity is the route.

The Favela project simulates the spatial development of spontaneous settlements at the local scale. The experiments are based on randomly walking agents over a cellular environment, constrained by attractive and non-attractive boundaries. It is based on the features of most of the inner-city settlements, which grow in vacant sites of usually highly integrated urban areas resulting from the fragmented and discontinuous development of Third World cities as described previously. These settlements grow from “attractive boundaries? towards the inner part of the urban site, the built structure is developed prior to any network and rough foot tracks arise in between built structures and often consolidate, connecting houses to local services situated on the site’s borders (Sobreira, 2002).

In such settlements, the streets of the existing city, which bound the site, tend to attract the development of building structures to the borders. Following this logic, the model’s rules are based on the idea that the spatial development of spontaneous settlements is both constrained and stimulated by the boundaries. In the Favela project the agent’s rules resemble some of the constraints and opportunities faced by actual people when looking for attractive urban sites to settle. The behaviour rules tell the agents to wander around the site and, when reaching an attractive boundary, find an available place to settle.
Squatter settlements consolidation

Among a number of interface options that allow the user to draw the site and change the parameters of development, the model presents a “feedback? procedure in which the agents are “fed? with information about the environment and, based on that information (such as global density, for example), change their behaviour, which in turn drives the spatial development to a different path. The most important parameters are the settling patterns (dwelling typology), the agents searching features, and the density threshold. Figure 4 illustrate a sequential output of Favela, in which the site is bounded by attractive boundaries at the bottom and right hand side only.

Figure 4: sequential output of Favela experiments

As one can see from the snapshots, the spatial configuration resembles the development process of a settlement in Acera, Ghana (Figure 5), starting with isolated building units combined with open areas (left) and as the density increases, the clustering and densification are inevitable (right). As the settlement gets dense and more agents come to the site searching for available space, agents take longer in the searching process, finally settling in more restricted spaces, occupying vacant spaces between existing dwellers and, thus, causing in this case the diversity of size of building clusters. Different initial conditions were also tested, in an attempt to explore to what extent path dependence influences the model’s behaviour.

Figure 5: squatter settlement in the city of Acera, Ghana. Early situation (left) confronted to the consolidated one (right).

4.2 Fragmentation: statistical properties of spatial complexity As visual comparisons are not enough to evaluate models outputs, some statistical analysis is necessary to better understand the results. In recent years a great deal of effort in pure and applied science has been devoted to the study of nontrivial spatial and temporal scaling laws which are robust, i. e. independent of the details of particular systems (Bak, 1997; Batty and Longley, 1994; Gomes et al., 1999). In
Proceedings . 4th International Space Syntax Symposium London 2003

recent work, Sobreira and Gomes (2001) show that spontaneous settlements tend to follow these scaling laws in both scales, local and global. This multi-scaling order is analysed here by a fragmentation measure which is related to the diversity of sizes of fragments (built units) in these systems. Diversity is understood here as a measure of complexity (Gomes et al., 1999) and an expression of universal dynamics. At the settlement scale, the fragmentation pattern refers to the diversity of size of islands (cluster of connected dwellings).

Figure 6 presents the maps of three squatter settlements showing in black the islands formed by groups of houses. The Figure 7 presents three samples run through the Favela project. The snapshots are related to the time when the development reached approximately the same number of houses as the real settlements of Figure 6 (around 250 dwellings), which allows a more precise statistical comparison.


Figure 6: Graphic representations of three squatter settlements situated, respectively, in Bangkok Thailand, Nairobi - Kenya and Recife - Brazil

Figure 7: Favela snapshots

The graphs in Figure 8 describe the correlation between the frequency of islands “f(s)?, according to its size “s?. The discrete variable “s? gives a measure of the size or area of an island (group of interconnected houses). The scaling distribution found for the settlements in Figure 6, represented by the graph of Figure 8.A, is statistically the same that related to the settlements generated by Favela (Figure 8.B), which reinforces our conjecture, which connects boundaries, packing and diversity as the interrelated key aspects to the internal development of squatter settlements. The distribution f(s) in both graphs of Figure 8 obeys a scaling relation given by f(s) ~ s-t, with t = 1.6 ± 0.2. The exponent t is robust and refers to the

Squatter settlements consolidation

degree of fragmentation of the settlement. The negative exponent t indicates a nonlinear scaling order, in which there is a great number of small units (islands, group of houses), a small number of big units, and a consistent distribution between them.

Figure 8: Fragmentation analysis: A) Scaling distribution - settlements in Bangkok, Nairobi and Recife; B) Scaling distribution - virtual settlements generated by the Favela agent based model

Fragmentation measures come as a way to compare the result of experiments on the dynamic processes, that is, on how these systems behave in time and space with the patterns found in empirical examples. In this aspect, we argue that such similarity of patterns is clearly empirical evidence of a decentrilised process of development of inner-city squatter settlements, which leads to a fragmented pattern, growing from the edge.

5. Conclusions From the AxialAgents experiments, preliminary results suggest that a complementary analysis considering flux and relative location of movement generators seems to be important to better understand the squatter settlements development dynamics. So, this approach, made possible through the AxialAgents model, can be a complementary one to other spatial analysis, such as the axial lines based mapping, adding a decentralised and dynamic view to the usual static perspective. Obviously, as suggested by Hillier (1996, p.161), “we cannont easily conceive of an urban structure as complex as the city in terms of specific generators an attractors, (...) because the city is a structure in which origins and destinations tend to be diffused everywhere?. But even such diffuse and multiple aspects of movement in the city can be considered by adding new elements and more complexity to the agent based model. It is important, however, to keep in mind the difference between complexity and complication, that is a tricky issue in any modelling experience.

Regarding Favela, experiments led us to come up with a suggestive indication that some spatial features of the decentralised dynamic of squatter settlements can be simulated by agent-based models, which describe the fragmented features of
Proceedings . 4th International Space Syntax Symposium London 2003

these self-organised systems. In these dynamics, boundaries (or edges) play an important role in the settlement development, which occurs through a decentralised packing process and results in a robust and universal pattern, as suggested by the scaling distribution of the squatter settlements fragmentation.

Finally, it is important to stress that both projects were built for strictly speculative purposes, that is, they focus on how simple local rules generate emergent and complex structures. In other words, how are people without collective leadership, searching for a place to settle their own houses, capable of building such intriguing and surprisingly complex spatial structures (Barros and Sobreira, 2002). Naturally, as it is suggested by Pacione (2001), the subsequent development of the squatter settlement into a settled community is a complex process that is related to the socioeconomic status and motivation of the inhabitants and dependent upon the attitude


of the government towards illegal settlement. Speculations presented in this paper are centred on rather specific spatial aspects of such complex and multifaced urban dynamics. Therefore, both AxialAgent and Favela models are set to be complementary (and not exclusive) tools for spatial analysis of decentralised and dynamic urban process such as those related to Third World cities.

References Bak, P., 1997, How Nature Works: The science of self-organized criticality, Oxford, Universirty Press Balbao, M., 1993, “Urban Planning and the Fragmented City of Developing Countries?, Third World Planning Review, 1(15), 23-35 Batty, M. and Longley, P., 1994, Fractal Cities: A Geometry of Form and Function, London, Academic Press Barros, J. and Sobreira, F., 2002, “City of Slums: self-organisation across scales?, Working Paper Series, Centre for Advanced Spatial Analysis, University College London, available on-line at www.casa.ucl.ac.uk. Dwyer, D. J., 1975, People and housing in Third World cities: perspectives on the problem of spontaneous settlements, London, Longman. Gomes, M., Garcia, J., Jyh, T., Rent, T. and Sales, T., 1999, “Diversity and Complexity: Two sides of the same coin??, The Evolution of Complexity, 8, 117-123, Dordrecht, Kluwer Academic Hillier, B., 1996, Space is the machine, Cambridge, Cambridge University Press Hillier, B., Greene, M. and Desyllas, J., 2000, “Self-generated neighbourhoods: the role of urban form in the consolidation of informal settlements?, Urban Design International, 5, 61-96 Holland, J., 1995, Hidden Order: how adaptation builds complexity, Massachusetts, Helix Books Mukhija, V., 2001, Upgrading Housing Settlements in Developing Countries: The impact of Existing Physical Conditions. Cities, Volume 4, no. 18, 213-222 Pacione, M., 2001, Urban Geography: a global perspective, London, Routledge Resnick, M., 2000, Turtles, termites, and traffic jams: explorations in massively parallel microworlds, Cambridge, The MIT Press Sobreira, F. and Gomes, M., 2001, The Geometry of Slums: boundaries, packing and diversity. Working Paper Series, Centre for Advanced Spatial Analysis, University College London, available online at www.casa.ucl.ac.uk. Sobreira, F., 2002, The Logic of Diversity: complexity and dynamics in spontaneous settlements, Doctor ate thesis (not published), Recife, Federal University of Pernambuco UNCHS–Habitat, 1982, Survey of slum and squatter settlements, Dublin, Tycooly International Publish ing Limited
Squatter settlements consolidation

Share & Embed

More from this user

Add a Comment

Characters: ...