Introduction

Gentrification is defined as the transformation of a working-class or unoccupied area of the inner city into higher-class residential and/or the other beneficial uses (Lees et al. 2008). It is one of the popular topics of urban inquiry. Originated from UK it has attracted widespread consideration in the other developed as well as developing countries (Lees et al. 2008; Diappi and Bolchi 2007; Torrens and Nara 2007; Guzey 2006; Bounds and Morris 2006; Ha 2004). According to Hamnett (1991) there are five reasons that gentrification attracted a great deal of interest: First, It has introduced a new and interesting urban phenomenon for geographers and sociologists to investigate. Second, Traditional theories of residential location and social structure are challenged seriously by gentrification. Third, as gentrification is concerned with regeneration at the cost of displacement, it is a political and policy-relevant issue (Sabri et al. 2006; Hatami Nejad et al. 2008). Forth, the ‘leading edge’ of contemporary metropolitan restructuring has been believed that is greatly constituted by gentrification. And fifth, According to the consumption and production explanations of gentrification it represented one of the key theoretical and ideological battlegrounds in urban geography.

Moreover, recently after about one decade, Loretta Lees and her colleagues in their book “Gentrification” (Lees et al. 2008) added three reasons to Hamnett’s showing that not only gentrification is still very alive, but it has also evolved, base on the global transformations in terms of politics and economy. They considered Gentrification as the leading edge of neoliberal urbanism. Then, according to them Gentrification is following the globalization process. Ultimately, it is no longer limited to the inner city or to First World metropolises.

They touched on some new definitions that are named as mutation of gentrification. New-build gentrification is as such that questions the historic built environment of gentrification. All authors are not agreeing that inner-city new-build gentrification developments are a form of gentrification. Some choose the term ‘reurbanization’ instead. Modeling and simulation techniques are believed to formulate a clearer image of future impacts of urban phenomena (Batty 2005; Benenson and Torrens 2004), such as new-build gentrification, to see whether it can be considered as type of gentrification or not.

There are notable achievements in modeling and simulation of gentrification (O’Sullivan 2002; Diappi and Bolchi 2007; Torrens and Nara 2007; Jackson et al. 2008). Nonetheless, the developing of new concepts remained untouched by model- and simulation- approach. Probably, the rapid development of derivatives and mutation of the phenomenon together with its politico-social nature is the reason for less effort in this regard. This paper tries to put one step forward and outline a new conceptual framework after the abstract model of O’Sullivan which started in early 2000 and followed by Diappi & Bolchi in Milan after about 5 years and then Torrens & Narra, who introduced a hybrid approach at the same time and finally Jackson et al. at 2008 that touched on new agents (students) as new comers of the city.

According to nature of New-Built Gentrification, the paper focuses on Multi-agent approach with considering at two scales of neighborhood and housing type schemes. The social and politico-economic aspects of gentrification will be formulated using Geosimulation as an integrated approach in urban modeling and simulation process. Multi-criteria evaluation (MCE), especially Analytic Network Process (ANP) (Saaty and Vargas 2006) is proposed to drive evaluation and calibration weights in the modeling and simulation process. This combination can translate the expert knowledge into social and physical transformation rules. Therefore, the conceptual framework can demonstrate the spatial dynamics of urban areas.

The paper proceeds as follows:

Related literature will be reviewed in section two, which talks about residential transformations in New-build gentrification theory, and land use change (LUC). Then the role of Multi-Criteria evaluation (MCE) in the context of urban modeling will be discussed in section three and the reliability of Analytic Network Process (ANP) will be highlighted in same section. In section four, the conceptual framework will be presented through a series of enquiries and approaches separately based on each part of the modeling process and at different scales. Ultimately, this paper will be concluded in section five with a discussion of potentialities and the next step of transferring the model to a real world experiment.

Related Theories in the Conceptual Framework

Urban modeling approaches have seen a considerable transformation from an urban ecological approach to systems approach (Briassoulis 2008; Liu 2009). The General Systems Theory (GST) constructs the fundamentals of systems approach (Batty 2005; Benenson and Torrens 2004; Liu 2009). Von Bertalanffy (1972), the founder of GST believes that everything can be considered in a system in which it becomes an element. The link and interrelation of all elements are obvious and these are linked to the system’s environment as well. In an urban context for example, an urban system is comprised of elements or subsystems such as residents, land, transport, authority. These elements interact with each other through social, economic, and spatial mechanisms. This study considers the systems approach to examine the transformations of the city.

The new-build gentrification is a concept that explains the most recent social transformations in urban context. A brief introduction of its implication and the components are described in this section. The two main techniques which are going to be used in the conceptual framework are scrutinized as well. Cellular Automata and agent-based modeling as more recent and sophisticated tools are reviewed to highlight their role in modeling and simulation of new-build gentrification appraisal.

New-Build Gentrification

When British sociologist Ruth Glass coined the term “gentrification”, her description was focused on the transformation of the working class residential neighborhoods based on the rehabilitation process that was conducted by middle classes, landlords and developers (Davidson and Lees 2005). She explicitly separated the redevelopment which involves in construction of new buildings on previously developed land, and rehabilitation of old structures.

In contrast, Neil Smith (1996) raised a new enquiry that how we can distinguish between the rehabilitation of nineteenth-century housing and all the new establishments in city centers in terms of social transformations. He did mention about the new condominium towers, festival markets that are attractive for local and international tourists and increase of wine bars and boutiques as well as the establishment of modern and post-modern office buildings that thousands of professionals are employed and need to have their desired style settlements. This opened a new chapter for gentrification researchers (Davidson 2006; Davidson and Lees 2005; Lees et al. 2008).

There are still controversies about considering the construction of luxury condos on reclaimed industrial land as a gentrification process. These are not old houses, and there is no displacement of a low-income community. However, according to Davidson and Lees (2005), indirect displacement, urbane new-middle class comers, new landscape or aesthetic aspect, and redevelopment of Brownfield sites are several facts that prove them as part of a gentrification process.

The new-build gentrification as a multi-actor process that authority and investors are the key actors is considered a complex urban phenomenon. Definitely, the household status and the resistance degree in terms of affordability and desirability are the indicators of new-build gentrification. The indirect effect of this process can be investigated in neighborhoods (Lees et al. 2008).

To conclude, according to literature (Davidson 2006; Davidson and Lees 2005; Lees et al. 2008) two types of transformations will occur during the process. The first one is landuse change (LUC) mostly from brownfield to residential and the second one is the socio-cultural transformation. In next section we will develop the discussion about the most recent tools which are suitable for study of such a complex urban phenomenon.

Modeling Urban Land-Use with Cellular Automata

There are currently two schools in CA land use modeling (Dietzel and Clarke 2006). The first approach treats the urban system as a fundamental entity, urban and non-urban units (Ward et al. 2000; Wu 2002; Li and Yeh 2001). The second school of CA disaggregates the urban land use and takes into account the multiple land uses of a broad landscape level or within the city itself. The feedback and dynamics among different land uses are necessary in second approach (Dietzel and Clarke 2006).

According to the nature of “new-build gentrification”, the second approach definitely is more compatible. This is because of four main uses of land use dynamics. These are: pre-industrial, residential, commercial and institutional. Although the mixed-use that is known as a new concept of planning, especially in inner cities, can be regarded as the fifth type. Nevertheless, there are some limitations in application of CA models. Therefore, in order to avoid the uncertainties that are mentioned in recent literature (Benenson and Torrens 2004; Crooks et al. 2008; Hammam 2008; O’Sullivan 2001, 2009; Yeh and Li 2006), the mixed-use can be assumed as its horizontal description in each neighborhood (O’Sullivan 2002).

Besides, urban cellular automata are regarded as spatially discrete systems that consisted of different simple elements (White et al. 1997). The integration of human urban objects is necessary in understanding the realistic urban dynamics. The GAS idea, therefore, provided a general framework for outlining and modeling more components of urban system by integration of multi-agent systems (Benenson and Torrens 2004).

Modeling Urban Dynamics with Multi-Agent Systems

Basically, agents have certain characteristics from practical point of view (Batty 2005; Benenson and Torrens 2004; Jackson et al. 2008; Ligtenberg et al. 2001; Macal and North 2005; Parker and Filatova 2008):

  • Identifiable; a self-contained discrete individual with exclusive characteristics and rules. The distinguishable boundary reveals the capability to determine whether something is part of an agent or not and maybe is a shared characteristic. This conveys the nature of cellular automata, which is regarded as an illustration of the basic ideas of agent-based modeling and simulation.

  • An agent is living in an environment in which it interacts with other agents “Reactive agents”.

  • An agent has its own goal, which drives the agent’s behavior.

  • An agent can function independently in the environment or in reaction with other agents according to its own protocols and plans “Autonomous”.

  • An agent can learn and adapt its behavior in each experience, which needs some sort of memory “Flexible”.

The diverse nature of agents makes agent modeling interesting. Agents are divers, heterogeneous and dynamic in their characteristics and rules. Multi-agent systems are useful in modeling and simulation of dynamic systems that are the outcome of interaction between varieties of components (Parker and Filatova 2008). Land use change can be regarded as such dynamic systems, the importance of integration the human system modeling into spatial models is emphasized by Ligtenberg et al. (2001), Benenson and Torrens (2004), Batty (2005), Torrens (2007), Parker and filatova (2008).

Modeling the complex systems with the variety of components and their interdependencies thus increasingly interested in using an agent-based idea. The roots of the agent-based modeling system (ABMS) are strongly in fields of multi-agent systems (MAS) and robotics in the field of artificial intelligence (AI). However, its main roots are in modeling human social behavior and individual decision making (Macal and North 2005; Trajkovski 2007).

There is a wealth of literature on using multi-agent models in which a wide variety of application domains can be investigated including traffic (Benenson et al. 2008; Khalesian et al. 2008), demographic change (Chen et al. 2007; Wu et al. 2008), decision support system (Sengupta and Bennett 2003; Sengupta et al. 2005), residential dynamics (Benenson 1998, 2004; Crooks 2008; Crooks et al. 2008), land market (Parker and Filatova 2008) and so many other applications. In this context, we touch on two aspects of human decision making. The first one would be the decision of multi-agents that reveals the new pattern of land-use, based on multi-actor spatial planning concept (Ligtenberg et al. 2004). The second approach is residential decision making for housing. These are collected based on objective of the paper which is going to examine the consequences of new-build gentrification.

Land Use Models

As mentioned in section Modeling Urban Land-Use with Cellular Automata, CA are used in several experiences of land-use modeling. In multi-agent system models also we can find several attempts that are categorized in three types of representation of land use change in the city or the urban growth patterns (Benenson and Torrens 2004).

The general land use models discuss about simple agents and their interaction with the environment. They are not going further than reaction to changes in close areas, and sometimes they do not even move, Diffusion-Limited Aggregation (DLA) can be mentioned as such. These models make tree-like structure, resembling the pattern of a city based on distance from a city center and provide the opportunity to estimate the fractal dimension of the obtained pattern (Batty and Longely 1994). Percolation of the developers’ efforts is another general model that was employed by Makse et al. (1995, 1998) to portray the growth of Berlin and London. Correlated percolation method is an alternative for DLA to show that developments are correlated rather than attached to previous developments randomly and the monocentricity of DLA is not a limitation in percolation (Makse et al. 1995, 1998). Intermittency of local development as another general model is based on attraction and repelling statues of the location (Benenson and Torrens 2004). Spatio-demographic processes and diffusion of innovation are studied by Durrett and Levin (1994) which again examine the evidence of urbanization and depend on parameters of reproduction and mortality of agent population.

Abstract models in urban phenomena can more explicitly explain the socioeconomic dynamics (Benenson and Torrens 2004; Benenson et al. 2008). Multi-agent systems also like the CA models at their first implementations consider the fixed agents. This is based on two closed urban phenomena: the first one is “Models of election voting”, in which the agents’ political opinion can be changed based on the influence of political environment. The second one is “Diffusion of innovation in the city”, which is quite same as the first model as the acceptance of innovation can be regarded as the voter “for”. Therefore, in an urban context the agents can influence on each other and can force to change their neighbors’ states.

Portugali and colleagues (1997) considered the human agents’ behaviors which are influenced by power at the three urban hierarchy levels, individual, local and global. Individual agents react to external forces by their internal ability and can decide to accept or deny. Local agents react to their neighbors influence and the difference between their own property and that of their neighbors makes them to show dissonance. The global agents act as their perception of the spatial structure of the whole city. This ability resembles to human feature according to Benenson and Torrens (2004) since an agent reacts to the new appearances of the systems in a holistic approach.

If we are not going to ignore the human basis for land-use change, the model that was conducted by Ligtenberg et al. (2001) is more suitable to discuss about. This model assumes a multi-actor planning process to allocate a land use based on the agent vote. The voting power and land use type depends on the agent situation and goal. The model focuses upon the interactions between agents and their environment.

The desired future structure of the space is defined by each agent separately. This is based on the distance-dependent weighted function (White and Engelen 1997) using CA. Then the synchronized decision making based on the characteristics of each agent is conducted to assign the final land-use features in next time (t + 1).

Benenson and Torrens (2004) believe that the aforementioned experience seems to be applicable to real-world modeling with the assumption that the number of land-uses will decrease and the weight functions will be justified by experiments. There are further attempts to provide planning support tools based on multi-actor decision making in spatial planning and particularly land use change decision making using the multi-agent concept (Ligtenberg et al. 2004; Carsjens and Ligtenberg 2007; Wachowicz et al. 2001)

There are three aspects that according to authors, should be developed to provide the idea of “the artificial planning experience”:

  1. 1.

    The spatial preferences and knowledge should be dynamic. Meaning that through time, the decision of agents regarding the land use type can be changed and the information they collect is changing in each time elapse. This is based on the definition of agents who have a learning ability and are flexible in reaction with the other agents and their environment (Macal and North 2005).

  2. 2.

    They mentioned the spatial organization that was regarded as discrete entities. The spatial pattern of neighborhoods was not defined as disjoint neighborhoods, which cannot show the topological relation of them. The one that in new-build gentrification has a significant concern as the price shadowing and indirect sociocultural transformations are the key challenge (Davidson 2006; Davidson and Lees 2005; Lees et al. 2008).

  3. 3.

    The last aspect relates the methodology in which the agent decides the land use based on one characteristic of land-use. According to Ligtenberg et al. (2001), “methodologies like game theory, belief networks, fuzzy set theory, machine learning, etc. perhaps offer a wealth of techniques that should be explored to enhance the decision-making procedure”

Parker and Filatova (2008) designed a conceptual framework for land use change procedure using agent-based modeling. The model is proposed from an urban economic perspective but moves to local spatial interactions and the different aspects of land use change (LUC) are taken into account. The socioeconomic criteria as well as ecological factors are tight coupled to land market modeling (Parker and Filatova 2008). The use of agent-based modeling in their work is based on the ability of incorporating heterogeneity, interactions and non-equilibrium dynamics of real-world land markets. According to authors the heterogeneous drivers for LUC are such as incomes, interest rates, social preferences, and credit availability. The interaction of agents is defined in a process of price negotiation. Land characteristic and personal preferences can influence on buyers and sellers to form bid and ask for properties, which are based on their “willingness to pay (WTP) and willingness to accept (WTA)”. The residential price in their framework is composed of Alonso’s bid-rent theory (Alonso 1964) and household’s budget.

This framework actually can bring the urban economic and environmental economic concepts into spatial modeling and is useful to provide a decision support for simulating the future social context based on proposed housing type and land uses.

Modeling Residential Dynamics in the City

There is a wealth of literature about agent-based modeling of residential dynamics in urban environment. Since the ability to change location is a key aspect of agent activity it gives the opportunity to explore the vast variety of theories regarding urban residential dynamics (Benenson and Torrens 2004; Chen et al. 2007; Parker and Filatova 2008).

The residential decision making for housing type and neighborhood is simply simulated based on several economic or social theories. Speare et al. (1974) classified the factors that determine the selection of a specific residence. These factors are divided among four categories: 1) Individual, 2) household, 3) housing and 4) neighborhood.

The individual residential behavior follows two main approaches: Revealed preferences approach and stated preferences approach (Benenson and Torrens 2004). The first approach takes the real-world data as consequences of residential choice. However, it is likely to be biased in case of adapting external constrains, such as lack of information about vacant dwellings. The second approach emphasizes on controlled experiments. The stated combination of characteristics can be considered in evaluation of potential residences by householders.

The revealed preferences are helpful in verifying models, while outcomes of stated preferences approach can be considered in establishing behavioral rules and their parameterization (Van de Vyvere 1994; Timmermans and Noortwijk 1995).

Portugali et al. (1997), intensively studied the aspects of the process of sociocultural emergence from theoretical point of view. They show the appearance of different forms of cultural and economic segregation. This was conducted in a multi-agent simulation model of urban residential dynamics. The active autonomous agents were utilized since they have the ability to interact, change their locations and their own properties together with the ability of imitation the residential behavior and development of human beings. The economic status and cultural identity of an individual agent by uni-dimensional quantitative variables were considered in this study.

Benenson (1998) rejected this oversimplified cultural identity and considered it as a multidimensional and quantitative variable. He considered the economic version of the multi-agent model as rather simple and cultural version as more complicated. The satisfactory parameter is what makes the individual free agents to either change themselves or leave the city. However, the interaction between urban infrastructure and residential distribution which can be comprised of non-residential and residential urban components was not considered in his model. Ultimately the application of such a model in real-world was questioned by author himself.

Daffertshoffer et al. (2001), tried to find how agents as individual residents or families occupy specific locations according to attractiveness of the settlement. They translated the preferences of individuals in a mathematical formulated model to find the optimal distribution of persons over flats. They outlined the self-organization concept which suggested there exist no planning presses in the city but consider individuals, households, firms or government agencies as urban planners in various scales. The issue of decision making at these scales highlighted the dynamics of cities.

Benenson (2004), introduced the satisfying hypothesis as an alternative for optimization approach for modeling residential dynamics. The stress-resistance approach allows households to avoid or relocate in unsatisfied and negative conditions of their locations. Thus the local factors can determine the global structure in this approach. The state of the house, the socio-economic characteristics of neighborhood, distances to work, recreation and effective transportation all are considered as the main parameters in this approach. Torrens and Nara (2007) utilized this view in their simulation model to examine the gentrification process in Salt Lake city. They tried to conduct a revealed preference approach to delineate the future social transformation of the city.

There are plenty of experiments based on individual decision making after Schelling (1971) model of segregation. The notion of his model is based on the simple logic of individual preferences regarding to the other social groups (Crooks 2008). Thus it seems that incorporation of these ideas into the process of New-build gentrification can reveal a peer example of social dynamics in reaction to emergent structure of space. In most experiences that were described in this paper, the choosing of transformation drivers can be different based on each location. Accordingly, it is necessary to determine the factors and the agents that have more influential in residential dynamics. Definitely, the weight of their influence is crucial as well, and it makes us to apply a multi-criteria evaluation (MCE) in order to accomplish a real world model. In following section we explore the effectiveness of incorporating the MCE in agent-based modeling.

Multi-Criteria Decision Making

Understanding the notion of new-build gentrification or any other urban phenomena needs to define the most reliable criteria and indicators based on time and place that are happening (Guzey 2006). These criteria are tangible or intangible that should be compared by an effective mechanism. Moreover, the dynamic nature of urban phenomena makes the decisions and the influence of each factor dynamic (Wu 1998). Therefore, Multi-criteria decision making (MCDM) can help in this regard.

Malczewski (1999) categorized the MCDM techniques into multiple objective mathematical programming (MOMP) and multi-attribute decision making (MADM). When there are discrete set of explicit alternatives, which are usually small numbers, MADM is utilized. In contrast, by MOMP, the set of alternatives can be simply defined using a set of constrains to be satisfied; consequently, the outcome would be an infinite set of decision alternatives. According to recent attempts in modeling and simulation of gentrification, and based on the literature about new-build gentrification, the criteria are discrete and can be distinguished explicitly (Davidson 2006; Davidson and Lees 2005; Diappi and Bolchi 2007; O’Sullivan 2002; Jackson et al. 2008; Torrens and Nara 2007). Therefore, in this study the MADM approach is more reliable for use.

There is a wide range of MADM techniques in literature, utility theory, outranking methods, regret-based, goal programming, ranking matrix, and analytic hierarchy/network process. Describing of all these approaches is out of scope of this paper but for more reading you may refer to (Chakhar and Lamsade 2008; Levy 2005).

The Analytic Hierarchy Process (AHP) and the Analytic Network Process (ANP) are proposed by Saaty (1985, 2006). The ANP addresses the determination of the relative importance of a set of activities in a multi-criteria decision problem (Saaty and Takizawa 1986; Saaty and Vargas 2006; Pourebrahim Abadi 2008). This is more reliable than Analytic Hierarchy Process (AHP) method that shows the more realistic outcomes when there are interdependencies of parameters (Sabri & Yakuup 2008a, b, c). The process employs pair-wise comparisons of the alternatives as well as those of the multiple criteria.

Fulong Wu (1998) adopted an integration of AHP and CA in SimLand workstation to retrieve the more realistic behavior into the urban development simulation. AHP as a technique in Multi-criteria Evaluation (MCE) is used in simulation to translate the expert opinion through different scenarios into the definition of transition rules.

The good decisions depend on the conditions of the future. Since conditions are varied over time, to make a good decision require judgments of what is more likely or more preferred over different time periods (Saaty 2007). There are essentially two analytic ways to study dynamic decisions. The first way is structural; by which, scenarios and time periods will be included as the elements in the structure that represents a decision. And the second way is functional that in which time will be explicitly involved in the process. A possible third way would be a hybrid of these two. What we need in study of new-build gentrification is to make the decisions dynamic which can be done in either of the above two mentioned ways.

Conceptual Framework

The aforementioned literature supports the conceptual framework that is provided in this section in a stepwise fashion. The simulation process comprised of a number of events that are integrated together to show the residential dynamics. Land use change based on transformation of spatial attributes is the most significant part in the new-build gentrification process. Housing type is completely relied on current spatial attributes of the area and the agents’ preferences in the space in each time step. The interrelation of all components will represent the total process of new-build gentrification.

Temporal Flow of Events

At each time step five main processes will happen: current distribution of fixed and mobile agents, land use characteristics that are based on local authority. Then the developer and the planning regulations will interact, and the housing type characteristics based on future demand and planning schemes will be defined. Next, households will choose to whether move from neighborhood or remain. Finally, housing type will be chosen by remainder of residents.

The temporal flow of events in Fig. 1 shows a number of models are proposed to implement and interact within an urban environment. Each model employs the different techniques, so a range of techniques and approaches will be introduced in this framework.

Fig. 1
figure 1

The temporal flow of events in simulation

Full size image

Land Use Change

this event is a complex procedure that normally can be pretended as a seed event for future dynamics. Land use change process follows the multi-actor-based land use modeling approach (Ligtenberg et al. 2001, 2004). Actors as players in the process of spatial planning have individual and synchronized behaviors. They have their own preferences in the space and communicate, negotiate and decide for the future pattern of their environments. The land use types according to new-build gentrification literature are chosen as developable land (vacant or disinvested), residential, commercial and institutional.

Housing Type Change

this event also follows the previous technique but comprising of different actors. The housing type planning according to literature depends on future demand, property characteristics, socio-economic characteristics, neighborhood quality, location factors and some planning concepts such as sustainability and Smart growth (Howley et al. 2009; Keskin 2008). This event considers the future residents’ characteristics as well as neighborhood characteristics together with planning schemes as effective actors in determining the housing type.

Inflow of Residents

the new residents as consumers of real estate come to the city area to engage in housing search inside the neighborhoods. The procedure follows the stress-resistance hypothesis (Benenson 2004), if the search is successful they will be added to the existing neighborhood population and may displace an existing resident in case the vacant rate is equal to zero. Otherwise, the current resident decides to remain or relocate based on generated social and economic characteristics of neighborhood.

Housing Type Choice

Existing household wishes to relocate or new entering household will decide to choose a housing type in this event. The housing types are comprised of single house, double story house, condominium, three or four unit apartment, and five or more units’ apartment (Torrens and Nara 2007). The concept of hierarchical nested choice will be used in formulating these events.

In every time elapsing the distribution of residents and land use as well as housing type will be updated and outcome of previous stage will be considered as the initial stage for next time.

Land Use Change

New-build gentrification literature suggests that the process is a consequence of land use change from pre-industrial or brownfield lands to residential, commercial or institutional uses (Davidson 2006; Davidson and Lees 2005; Smith 1996). Besides, the key drivers are the state intention to take part in city competence (Guzey 2006, 2009) and developers as well as property value (Hamnett and Whitelegg 2007).

  1. Question 1:

    How can we model the land use change process by taking into account the drivers of new-build gentrification?

Following Ligtenberg et al. (2001, 2004) ABM model for land use change, the voting power in multi-actor fashion is more akin to nature of new-build gentrification. Ligtenberg et al. (2001) suggests a two step decision making in the process of land use change. Firstly, agents construct their individual preference image of future land use. Secondly, the agents communicate and share the ideas. Thus they finally synchronize the ideas to make the final decision.

As the individual agent decision making process, agent carries out two tasks, first evaluates the spatial environment and second assigns ranks to every parcel. A simple ranking indicator can be used as a transition rule in cellular automata (CA) to carry out the agents’ spatial evaluation. The distance based weighted sum approach is desired for this transition rule (White and Engelen 1997). In order to apply the rules to objects rather than cellular spaces, distance-based weights can be calculated by spatial preference functions (Ligtenberg et al. 2001; Wu 1996).

$$ f = f\left( {Parcel_{{Sxy}}^t,{\Omega_{{Parce{l_{{xy}}}}}}} \right) $$
(1)

Where, \( Parcel_{{Sxy}}^t \) is the land lot at time t in activity state S with centroid coordinate x, y. \( {\Omega_{{Parce{l_{{xy}}}}}} \) represents the neighborhood for Parcel xy with a specific distance buffer around it and centroid coordinate x, y . Then a CA rule generates the ranking indicator:

$$ {R_z} = \sum\limits_{{i = 1}}^n {\sum\limits_{{j = 1}}^s {{D_{{jzp}}}{J_{{ij}}}{F_{{if}}}} } $$
(2)

Where, n indicates the number of parcels in the neighborhood \( {\Omega_{{Parce{l_{{xy}}}}}} \), and s is the number of land use classes as a part of S. R z is the aggregate land use indicator for suitable land use class z. D jzp represents the agent specific ranking of land use class j for land use class z based on aggregate value from proximity of p including the central neighborhood district, major (city scale) commercial land use, and entrance highway.

J ij = 1 if cell in distance d is in state j otherwise, J ij = 0, and F if = 1 if the property f of parcel i sets to false, otherwise F if = 0. The aggregate ranking indicator certainly is different for each agent. Therefore, each agent has a personal set of ranks that indicates the agents’ future spatial preference.

After defining each agent’s desire of future spatial organization, the process of decision making starts. The local authority handles the task by asking for other agents’ opinion. Following four steps define the voting process:

  1. 1.

    There are two conditions in each voting time. Firstly, agents are agreeing in land use type of a parcel, so the parcel can be updated based on agreed land use. As the second condition ideas are not identical, so the location will be stored in a conflict list.

  2. 2.

    For the conflict list agents are asked to vote. It should be mentioned that the voting power is different based on agents’ hierarchy. Weighted agent vote P vote is based on agent’s voting power V w .

$$ {P_{{vote}}} = {V_w}i $$
(3)

In this case if the agent’s desirable land use (A s ) is similar to that of proposed by scenario (Prop s ), the vote is in favor of new land use:

i = {1, if A s= Prop s , else, 0}

The vote against is giving when the desired land use is not similar to that of proposed in scenario.

i = {1, if A s≠ Prop s , else, 0}

  1. 3.

    A new list can be extracted from conflict list containing the locations with majority of votes in favor of the proposed change. This list is a subset of conflict list and the number of favor votes is more than against votes.

  2. 4.

    The local authority agent ranks the remainder of the conflict list based on the weighted number of votes. This will continue until the summation of all votes in favor is greater than votes against. It should be mentioned that the updating of neighborhood cells will change the agents’ attitude toward the parcels that are remained in the conflict list in several iterations.

Housing Type Transformation Process

Neil Smith (1996) pointed out the construction of new condominium towers as one of the peer signs of new-build gentrification. Chris Hamnett and Whitelegg (2007) reviewed the conversion of loft to luxury apartments and consequently, the new lifestyle in the Clerkenwell, next to the city of London. Torrens and Nara (2007) considered the housing type in their modeling and simulation of gentrification in South Lake City. All these literatures are agreed that the housing type and its transformation are key factors in the new-build gentrification process of the urban area.

  1. Question 2:

    How to put the housing type transformation in the model and who are the drivers in this process?

Most of the literatures consider the government, developers and planning legislations as the key drivers of housing type change (Bounds and Morris 2006; Guzey 2006, 2009; Hamnett and Whitelegg 2007; Ha 2004). At this point, no study is found about modeling the housing type change. However, if we assume the housing type change will follow a same process of land use transformation, we can apply the proposed model in last section for housing type change.

Modeling Residential Mobility

Following the explanations of new-build gentrification the indirect socio-cultural displacement is the consequence of in-movers characteristics. This is mainly a social aspect (Davidson and Lees 2005; Lees et al. 2008).

  1. Question 3:

    How to highlight the socio-cultural displacement and consider the indirect effect of the new-build gentrification?

There is a wealth of literature in residential transformation from different points of view. Nevertheless, we adapt the model of Torrens and Nara (2007). In which, both socio-economic and physical characteristics of the urban set are considered. The integration of the property characteristics and in terms of price, value, size and so forth with social characteristics such as ethnicity, economic status is the advantage of their model. The “propensity for mobility” in their work is the result of household preference of existing property.

$$ L{h_{{ij}}} = 1 - L{p_{{ij}}} $$
(4)

Where,

Lh ij :

Shows the probability that household i will move from property j.

Lp ij :

Indicates the probability that household i prefers to remain in property j.

The probability of household preference is the outcome of a nested spatial choice that included all the related characteristics and attributes.

$$ L{p_{{ij}}} = \sum {\left( {{W_p} \cdot {P_C}} \right) + \sum {\left( {{W_N} \cdot {N_C}} \right)} } $$
(5)

Where,W p and W N are weights of housing type and neighborhoods respectively, they are retrieved from ANP method based on expert opinion. Definitely the value of weights should be different in each time elapse, which is possible by considering a Dynamic Network Process, DNP (Saaty 2007). In this research the DNP is considered as a future work which idea needs to be developed. Therefore the initial values of weights are considered to be static.

P C :

Is a bundle of housing characteristics such as value, type, size and distance to work, service and entertainment.

W N :

Is consisted of the characteristics of the neighborhood which are mostly the socio-cultural aspects such as economic status and ethnicity.

Sensitivity Analysis

To ensure that a complete set of evaluation criteria is provided, also the true preferences of agents are extracted, the research needs to carry out a sensitivity analysis (SA). It helps to investigate the potential changes and errors in the data and assumptions and their impacts on the outcomes (Nyerges and Jankowski 2010).

  1. Question 4:

    How to ensure that the model solution is stable in plausible changes?

This stage is the necessary part of the verification process in general literature on simulation (Benenson and Torrens 2004). The stochastic perturbation of model components will be investigated by adding stochastic terms to each part of it. By running the simulation several times the normal distribution of the results will be obtained.

Geographic Information Systems (GIS) can provide a sound platform for such a complex model that needs a variety of data type. The programming ability in GIS environment as well as visualization power can bring an opportunity in different scenario evaluations for planners and decision makers (Sabri and Yakuup 2008a). The compatibility of current simulation software’s like Netlogo 4.0.3 and Repast Symphony with GIS is it’s another advantage (Crooks 2008). Figure 2 illustrates the schematic procedure of residential decision making as well as land use transformation and housing type change.

Fig. 2
figure 2

Schematic representation of the household preference of Neighborhood and Housing types

Full size image

Conclusion

In this paper, we have outlined a conceptual design for gentrification appraisal using Geosimulation (Benenson and Torrens 2004) and ANP (Saaty and Vargas 2006) approaches. The model is comprised of several events with heterogeneous agents who make it dynamic. The new-build gentrification approach is considered to determine the drivers of physical as well as social transformations. This makes us to use a variety of concepts and theories during the simulation process; which touch on the social, political and economic aspects of urban space. Thus, we hope that the model will be of interest to all who are involved in urban studies.

In order to achieve a sustainable development and smart growth, today planners take the advantage of advanced computation in supporting the decision making. GIS has provided the sound opportunity for decision support systems. Especially, when it is integrated by the other reliable software and analytical techniques.

The review of previous attempts in modeling of Gentrification highlighted the lack of a systematic approach in parameter validation. Therefore, this study suggested an integration of Geosimulation and ANP to achieve a more realistic model. Besides, the study adapted new conceptualization of gentrification, which is likely to be more akin to current urban transformations (Lees et al. 2008).

There is a long way to go. Since the presented model is a conceptual framework, it needs to be implemented to evaluate the suggested methods. Definitely, after detailing of the model the implementation phase is an important stage; in which we will face with some limitations and need to revisit the proposed models with time.