رگرسیون لجستیک و مدل سازی مبتنی بر اتوماتای ​​سلولی توسعه خرده فروشی، تجاری و مسکونی در شهرستان احمدآباد هند

This study presents a hybrid simulation model that combines logistic regression and cellular automata-based modelling to simulate future urban growth and development for the city of Ahmedabad in India. The model enables to visualize the consequence of development projections in combination with present zoning and development control regulations. The growth in activities’ floor space is projected at a zonal level using time series data. Then, a logistic regression model is used to calculate a probability surface of development transition, while a cellular automata-based spatial interaction model is used to simulate change in activity floor space per activity, and thus urban growth. The developed model has the capacity to simulate urban growth space and hence vertical growth. The structure of the model allows for a detailed urban growth simulation and is flexible enough to incorporate changes in development control regulations and settings for spatial interaction. Therefore, it carries scope of being used to visualize growth for other, similar, cities and help urban planners and decision makers to understand the consequences of their decisions on urban growth and development.

Rapid urbanization and urban growth are recognized facts in India (MOUD., 2011). The planning of urban areas in India and, specifically, the methods used to prepare statutory urban development plans have been criticized as being ineffective (Adhvaryu, 2011). A likely reason for this poor practice of urban planning methods is mainly attributed to a limited understanding of the spatio-temporal processes and dynamics of urban growth (Adhvaryu, 2011) and the lack of tools that can help in making informed urban planning decisions (Geertman and Stillwell, 2004 and Masser and Campbell, 1991). The dynamics of urban growth are driven by the availability of space in urban areas for urban areas to change or grow and the ageing process of development (referring to the ageing and deterioration of buildings, as a result of which buildings might have to be restored or reconstructed [Batty, Xie, & Sun, 1999a]). Nivola (1999) states that urban growth can progress in four directions: “in, up, down and out. Indicating that the objective of any urban growth model that is used to make informed urban planning decisions should be to identify what land use will develop, when and where, and what its repercussions are on other developments” (Cheng, 2003). Over the years, urban growth models have proven to be effective in describing and estimating urban development (mostly outgrowth) and have consequently proven to be valuable for informed urban planning decisions (Clarke and Gaydos, 1998, Herold et al., 2003 and Vaz et al., 2012). Previous studies (e.g., Clarke et al., 1997 and Dubovyk et al., 2011; Arsanjani, Helbich, Kainz, and Darvishi Boloorani (2013); Lo and Yang, 2002, Thapa and Murayama, 2012 and Aljoufie et al., 2013a; (Aljoufie, Zuidgeest, Brussel, van Vliet, & van Maarseveen, 2013b) also emphasized the need and importance of performing a spatio-temporal analysis of urban growth, particularly in assessing the impact of future (land use) scenarios in terms of locations, characteristics and consequences. The use of urban growth modelling and prediction dates back to the 1950s (Torrens, 2006). The early urban growth models were developed by Chapin and Weiss, 1962 and Tobler, 1970 and Nakajima (1977). However, the interest in urban growth modelling faded in the following years and regained momentum only in the 1990s (Cheng, 2003) as a result of improvements in the availability of spatial data and computing ability (Wegener, 1994). Allen and Lu (2003) observed that in empirical studies on urban growth modelling, urban systems have been viewed in several contrasting ways, which have led to several subsequent theories and models (Southworth, 1995, Thapa and Murayama, 2012 and Wegener, 1994) of land use change and urban growth, including rule-based models such as Cellular Automata (CA) models (Clarke, Hoppen, & Gaydos, 1996). As observed by Zhao, 2011 and Tobler, 1979 work initiated several modifications to the then-existing CA models to make them suitable for simulating and predicting urban growth (Batty et al., 1999b, Clarke and Gaydos, 1998, O’Sullivan, 2001, White and Engelen, 2000 and Wu, 2002; Yeh & Li, 2002; Yen & Li, 2001). These modifications include the coupling of cellular automata with fuzzy logic (Liu, 2012), Markov Chain algorithms (Cheng & Cao, 2011), relative probability and/or regression models (Hu and Lo, 2007 and Pinjanowski et al., 1997), statistical models (Landis, 2001), and artificial neural network models (Almeida, Gleriani, Castejon, & Soares-Filho, 2008). The logistic regression-based Cellular Automata (CA) model was first proposed by Wu (2002). This hybrid approach, which was also implemented by Paulmans and Van Rompaey (2009), helps to overcome the main limitations of logistic regression, which is the inability to quantify spatial and temporal changes (Arsanjani et al., 2013), and of the CA approach that oversimplifies urban reality and does not provide enough evidence for informed urban planning (Allen and Lu, 2003 and Sui, 1998) (Batty, 2000 and Sipper, 1997). CA-based models are simple and allow for dynamic spatial simulation (Torrens & O’Sullivan, 2001). CA, when used in combination with logistic regression models, as done by Wu, 2002 and Arsanjani et al., 2013, has the ability to predict a complex urban reality and thus has a potential application in Indian cities. This hybrid framework can form the basic structure of the model; however, when such models are applied in the context of Indian cities, several due considerations will have to be made. The first important consideration deals with adapting the CA framework to suit the realities in the city in the sense that the outputs of the model should provide information at an appropriate scale for urban policy making. The second consideration deals with the input to the logistic regression models. Given the availability of data, it is important to identify the methods that can be used to quantify the inputs that go into the logistic regression model. The third consideration is to build a link with the policy framework, that is, the CA-based framework, and to operationalize the model by using inputs from the logistic regression model and urban policies in such a way that the model is able to establish forward and backward linkages with urban development and control policies. This consideration should ultimately be reflected in the simulation of urban development and growth. Therefore, this study furthers the work conducted on modelling the urban development and growth of Indian cities (Sudhira et al., 2004, Lata et al., 2001, Jat et al., 2008), which are mostly examples of modelling urban sprawl and are not able to model the development of different land uses. Thus, this study showcases how some recent applications, as found in the empirical literature on urban growth modelling, can be adopted for the Indian situation. The paper is divided into six sections. The next section briefly introduces the study area, followed by a description of the data used in the study. Section 4 details the modelling approach and discusses how the models used in the study are built on the existing approaches and how they are adapted in the context of Indian cities. The results of the application of the modelling approach for the city of Ahmedabad and the validation of the results are presented in Section 5. The paper ends with a discussion of the conclusions in Section 6.

The purpose of the present study was to develop an urban growth model that is able to simulate land use development and urban growth for the city of Ahmedabad in India and that is replicable in another, but similar, context. The presented tool uses a hybrid approach: a combination of a logistic regression and a Cellular Automata (CA)-based model to simulate the future urban development and growth of the city. The development projections are calculated and validated for each property tax zone, and these estimates are used as an input into the urban growth simulation model. Existing zoning and Development Control Regulations, outputs of the logistic regression as well as other activity location rules were used to calculate the transitional probabilities for activities in each grid cell. These probabilities are input into the location-allocation model and used to simulate urban growth. The model allows the possibility of designing alternate scenarios with different zoning and Development Control Regulations. The projections made on a disaggregated scale are distributed to the grid cells depending upon their propensity to develop or redevelop in combination with the attractiveness of the location for the activities to develop. The method used to simulate urban growth and development enables the visualization of the consequences of development projections in combination with (present) zoning and development control regulations. It is observed that there is a tendency for activity locations to cluster around locations that have a good mix of activities, as demonstrated in the logistic regression, which leads to a densification of activities at these nodes. On the other hand, the city of Ahmedabad is rapidly expanding due to rapid growth in combination with the limitations posed on development through FSI and other Development Control Regulations. Thereby, the results from this study produce a correct extrapolation of past urban growth trends and should give enough indication to the development authorities about the type of development that the present policy regime will lead to. This should allow them to make more informed decisions. The developed model has proven to be capable of accurately modelling the heterogeneous urban development pattern of Indian cities. The model is flexible and has the scope to incorporate changes in the Development Control Regulations and in the spatial interaction of activity locations with a number of structural and spatial indicators, which was considered as a drawback of many earlier models predicting urban development (Maithani, 2010). Thus, the model carries the scope and the potential to be used to project urban development and growth in other similar cities in India and elsewhere. Lastly, if such models were to be used by the development authorities, they will have to be inexpensive; the approach presented in this paper uses conventionally available software in most urban development offices in India to operationalize the model.