Implementation and calibration of a new irregular cellular automata-based model for local urban growth simulation: The MUGICA model

Abstract

Cellular automata-based models have traditionally employed regular grids to represent the geographical environment when simulating urban growth or land use change. Over the last two decades, the scientific community has introduced the use of other spatial structures in an attempt to represent the processes simulated by these models more realistically. Cadastre parcels are a good choice when simulating urban growth at local scales, where pixels or regular cells do not represent the geographic space properly. Furthermore, the implementation and calibration of key factors such as accessibility and suitability have not been sufficiently explored in models employing irregular structures. This paper presents a fully calibrated model to simulate urban growth: Model for Urban Growth simulation using Irregular Cellular Automata. The model uses the irregular structure of the cadastre and its smallest unit: the cadastral parcel. The factors included are based on the traditional Neighbourhood, Accessibility, Suitability and Zoning Status modelling schema, frequently employed in other models. Each factor was implemented and calibrated for the irregular structure employed by the model, and a new approach was explored to introduce a random component that would reproduce illegal growth. Several versions of Model for Urban Growth simulation using Irregular Cellular Automata were produced to calibrate the model within the period 2000–2010. The results obtained from the simulations were compared against observed growth for 2010, adapting the traditional confusion matrix to irregular space. A new metric is proposed, called growth simulation accuracy, which measures how well the model locates urban growth.

Regresión Logística Geográficamente Ponderada para identificar los factores explicativos de la distribución de usos de suelo en escenarios futuros de crecimiento urbano

“…Por su parte, respecto a los factores identificados, los modelos locales RLGP constataron la existencia de variabilidad espacial en los factores que explican la localización de los crecimientos urbanos de acuerdo con los resultados del taller colaborativo. Observando los resultados es posible deducir que, de entre las variables físicas, la distancia a ríos y la altitud han sido relevantes en gran parte de los modelos estudiados, mientras que la variable pendiente no ha presentado tal relevancia a pesar de su aplicación en numerosos estudios (Barreira-González et al, 2017;Hu & Lo, 2007;. Por su parte, las variables relativas al transporte público y la movilidad (distancia a estaciones de tren y autobús interurbano) presentan influencia en un mayor número de usos, dejando en un segundo plano la accesibilidad por carretera, a excepción de modelo que ajusta la distribución espacial del uso "residencial unifamiliar" en el escenario Altos niveles de inseguridad ciudadana y el uso…”

Urban Land Allocation Model of Territorial Expansion by Urban Planners and Housing Developers

“…The input data used to implement the submodels were selected to reflect the drivers of urban growth, empirical knowledge of urban growth trends in the study area [47,48,50,51], and the most commonly used variables in other studies [8,33]. These latter included the municipal boundaries (surface), zoning status (surface), and housing distribution (pixel) used by agents to assess specific environments and conditions, as well as the other spatial information indicated in Table 1 below.…”

Pattern‐based calibration of cellular automata by genetic algorithm and Shannon relative entropy

“…Cellular automata (CA), a collection of fixed cells regularly arranged in a cellular space, is the simplest model and a special type of automata. A two‐dimensional regular grid is the most common CA (Liu, 2009), however, other arrangements such as one‐dimensional CA (Di Gregorio & Festa, 1981), or CA with irregular cells (Barreira‐González, Aguilera‐Benavente, & Gómez‐Delgado, 2019) have been developed to represent objects and elements with different shape and size. Each of the regular spatial cells that constitute a CA can only have one of a finite number of possible values or states (e.g.…”

“…CA modeling is widely used due to the ability to accurately and efficiently replicate observed complex dynamics using a simple rule system, and the incorporation of socioeconomic and biophysical forces driving land use change (Newland, Maier, Newman, van Delden, & Zecchin, 2015). Since cities can be conceived of as consisting of a two‐dimensional regular grid of n × n cells (Liu, 2009), with simple transition rules generating complex urban development patterns due to the CA’s self‐organization and self‐reproduction ability, CA have been adopted in modeling various natural, social, and economic phenomena (Sidiropoulos & Fotakis, 2011), including urban modeling to simulate urban growth (Barreira‐González et al, 2019; Liu et al, 2018; Rienow, 2016; Xia, Wang, Zhang, & Zhang, 2018), understand processes of land use change (Roodposhti, Hewitt, & Bryan, 2020), and in various optimization problems (Afshar & Shahidi, 2009; Guo, Walters, Khu, & Keedwell, 2007; Heinonen & Pukkala, 2007; Mathey, Krcmar, Dragicevic, & Vertinsky, 2008; Strange, Meilby, & Bogetoft, 2001). Since the early 2000s, urban growth and land use change were simulated by SLEUTH, a modified CA model whose name is an acronym for the input data layers of slope, land use, exclusion, urban extent, transportation, and hillshade (Clarke, Gazulis, Dietzel, & Goldstein, 2007; Clarke, 2008a, 2008b).…”