A framework for modelling spatio-temporal informal settlement growth prediction

https://doi.org/10.1016/j.compenvurbsys.2021.101707Get rights and content

Highlights

  • A framework is proposed for modelling informal settlement growth and emergence.

  • The framework is generic and may be adopted to model informal settlement development in any developing country.

  • The framework is applied for demonstration purposes within the context of a South African real-world case study.

  • A novel machine learning-inspired population diffusion model is proposed for inclusion in the framework.

  • An objective feature selection approach is adopted instead of a subjective feature selection for the transition rule.

Abstract

Many developing countries grapple with the problem of rapid informal settlement emergence and expansion. This exacts considerable costs from neighbouring urban areas, largely as a result of environmental, sustainability and health-related problems associated with such settlements, which can threaten the local economy. Hence, there is a need to understand the nature of, and to be able to predict, future informal settlement emergence locations as well as the rate and extent of such settlement expansion in developing countries.

A novel generic framework is proposed in this paper for machine learning-inspired prediction of future spatio-temporal informal settlement population growth. This data-driven framework comprises three functional components which facilitate informal settlement emergence and growth modelling within an area under investigation. The framework outputs are based on a computed set of influential spatial feature predictors pertaining to the area in question. The objective of the framework is ultimately to identify those spatial and other factors that influence the location, formation and growth rate of an informal settlement most significantly, by applying a machine learning modelling approach to multiple data sets related to the households and spatial attributes associated with informal settlements. Based on the aforementioned influential spatial features, a cellular automaton transition rule is developed, enabling the spatio-temporal modelling of the rate and extent of future formations and expansions of informal settlements.

Introduction

The world's population growth rate is expected to decrease from 1.24% in 2000 to 0.5% in 2050, while the population of many developing regions is expected to nearly double over the next 30 years (United Nations, 2015). As of 2007, more than half of the world's population lives in cities (Friesen, Rausch, Pelz, & Fürnkranz, 2018). Due to the projected global population growth, these cities continue to attract thousands of new residents from rural areas every year in search of employment and a better standard of life (Abbott, 2002; Kohli, Sliuzas, Kerle, & Stein, 2012). Urban housing backlogs coupled with shortages in housing subsidies in developing regions, such as in Africa (United Nations, 2002), South America (Friesen, Taubenböck, Wurm, & Pelz, 2019; Kamalipour & Dovey, 2019) and South-East Asia (Friesen et al., 2019), for example, have led to large numbers of urban dwellers having no alternative but to live on the peripheries of cities in informal settlements. Many urban areas in developing regions are, in fact, anticipated to contain 60% of the global population by 2030, with 90% of the projected urbanisation expected to occur in less developed countries (He, Okada, Zhang, Shi, & Li, 2008).

The emergence or expansion of urban areas is related to their topographical features, the local transport infrastructure, the current surrounding land-uses, the surrounding social structure, and economic activities associated with the underlying area. As the populations of cities increasingly require additional space to live in, it is generally accepted that the demographies and economies of these populations will be the primary driving factors of spatial urban and informal settlement emergence and expansion (He et al., 2008).

Although developed areas currently only cover a small fraction of the earth's land surface, there is evidence that spatial urban and informal settlement expansion has significantly altered the earth's natural landscape, resulting in a very significant impact on the surrounding environment and ecosystems at all geographical scales (Li, Zhu, Sun, & Wang, 2010; Seto, Güneralp, & Hutyra, 2012; Wei & Ye, 2014). Moreover, temporal urban and informal settlement growth is expected to contribute to global environmental changes, including land-use and land cover changes, changes in biogeochemical and hydrological cycles, climate, and biodiversity (McDonald, Kareiva, & Forman, 2008; Seto et al., 2012).

In order to predict land-use changes and settlement growth quantitatively and in a spatial fashion, cellular automata (CA) have typically been adopted as the primary modelling paradigm in the literature. In CA models, the landscape is represented by cellular land parcels that can develop into different land-use types according to local transition rules, in much the same way as fractal growth occurs. Global patterns of urban growth then emerge as a result of rural to urban conversions. Whereas previous urban development modelling and analyses of the associated impacts of such development have taken place mainly in the context of developed countries (Linard, Tatem, & Gilbert, 2013), it may be argued that developing countries (which exhibit the highest urbanisation rates) are more in need of such modelling and analysis projects.

A data-driven framework for modelling spatio-temporal informal settlement emergence and growth is proposed in this paper which leverages the considerable power of machine learning (ML) instead of relying on researchers to postulate the factors influencing settlement development subjectively. This framework is deemed unbiased as the factors driving informal settlement emergence and growth are identified and weighted from a large pool of data presented by a user, resulting in a black-box emergence of prominent driving factors to be employed within a predictive CA model.

The remainder of this paper is structured as follows. A brief overview of existing frameworks and models for settlement emergence and growth is given in §2, emphasising the strengths and shortcomings of these frameworks and models. The framework proposed in this paper is then presented in some detail in §3. The practical workability of the framework is showcased thereafter, in §4, in the form of a real-world case study. The paper finally closes with a short appraisal of its contributions and some proposals for future work in §5 and §6.

Section snippets

Background and literature review

Urbanisation and the emergence and growth of informal settlements have been the subject of research interest for decades in several disciplines (Batty & Longley, 1986; Batty & Xie, 1994; Couclelis, 1985; Sietchiping, 2005; Torrens & O'Sullivan, 2001). A relatively recent realisation that urbanisation and informal settlement development are largely dynamic processes have led to an adoption in the literature of a variety of modelling paradigms aimed at explaining the driving forces behind

Framework

The review of previous settlement expansion models presented in §2 indicates that, to the best of the authors' knowledge, no model or framework exists for investigating the intricate nature of informal settlement emergence and expansion based on dynamic spatial feature selection by an ML approach. A data-driven framework, called the informal settlement emergence and growth (ISEG) framework, in proposed in this paper for predicting the future temporal population growth and spatial population

Illustrative case study

In order to demonstrate the practical workability of the proposed ISEG framework, a concept demonstration is carried out in this section according to the framework with a view to model informal settlement emergence and growth in a developing country context. Although a South African case study is conducted in this paper for demonstration purposes, the generic nature of the ISEG framework allows for informal settlement growth and emergence analysis in any area of interest.

Discussion

In this paper, a generic framework for modelling spatio-temporal informal settlement emergence and growth was proposed. This framework addresses two major shortcomings of similar, existing frameworks. The first pertains to comprehensiveness and generalisation requirements in a variety of contexts. The second is that it renders irrelevant the typical subjective selection of factors driving informal settlement emergence and growth in existing models and frameworks, and then incorporating these

Conclusion

The generic data-driven ISEG framework proposed in this paper provides an objective feature selection procedure which proves to be more desirable than the subjective feature selection approaches (i.e. feature selection approaches in which a user selects certain spatial features based on their preference) typically employed in existing, similar modelling frameworks. During the objective feature selection procedure the ML component of the ISEG framework identifies and selects an appropriate

Author statement

(CRediT author statement as per Elsevier website https://www.elsevier.com/authors/policies-and-guidelines/credit-author-statement)

CRediT authorship contribution statement

P. Cilliers: Conceptualization, Methodology, Software, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administratio. J.H. van Vuuren: Conceptualization, Formal analysis, Resources, Writing – original draft, Writing – review & editing, Supervision, Project administration, Funding acquisitio. Q. van Heerden: Resources, Data curation.

References (57)

  • M. Sokolova et al.

    A systematic analysis of performance measures for classification tasks

    Information Processing and Management

    (2009)
  • A. Tsoularis et al.

    Analysis of logistic growth models

    Mathematical Biosciences

    (2002)
  • M. Batty

    Urban evolution on the desktop: Simulation with the use of extended cellular automata

    Environment and Planning A: Economy and Space

    (1998)
  • M. Batty et al.

    The fractal simulation of urban structure

    Environment and Planning A: Economy and Space

    (1986)
  • M. Batty et al.

    From cells to cities

    Environment and Planning. B, Planning & Design

    (1994)
  • J.R. Cannon

    The one-dimensional heat equation. 23

    (1984)
  • J. Canosa

    On a nonlinear diffusion equation describing population growth

    IBM Journal of Research and Development

    (1973)
  • A. Cecchini

    Urban modelling by means of cellular automata: Generalised urban automata with the help on-line (augh) model

    Environment and Planning. B, Planning & Design

    (1996)
  • A. Cecchini et al.

    Ficties (fictitious cities): A simulation for the creation of cities

  • J. Cheng et al.

    Modelling urban growth patterns: A multiscale perspective

    Environment and Planning A: Economy and Space

    (2003)
  • P. Cilliers

    ML and CA components of the ISEG framework code

    (2021)
  • K.C. Clarke et al.

    Methods and techniques for rigorous calibration of a cellular automaton model of urban growth

  • K.C. Clarke et al.

    A self-modifying cellular automaton model of historical urbanization in the San Francisco bay area

    Environment and Planning. B, Planning & Design

    (1997)
  • H. Couclelis

    Cellular worlds: A framework for modeling micro-macro dynamics

    Environment and Planning A: Economy and Space

    (1985)
  • A. Fick

    Ueber diffusion

    Annalen der Physik

    (1855)
  • J. Friesen et al.

    Determining factors for slum growth with predictive data mining methods

    Urban Science

    (2018)
  • J. Friesen et al.

    Size distributions of slums across the globe using different data and classification methods

    European Journal of Remote Sensing

    (2019)
  • S. Gao

    Spatio-temporal analytics for exploring human mobility patterns and urban dynamics in the mobile age

    Spatial Cognition and Computation

    (2015)
  • Cited by (3)

    • Simulating urban expansion using cellular automata model with spatiotemporally explicit representation of urban demand

      2023, Landscape and Urban Planning
      Citation Excerpt :

      Over the past few decades, researchers have developed various urban growth models to investigate the mechanisms that drive urban dynamics and their future pathways, thus guiding urban systems toward directions that improve human welfare and promote global sustainability (Balmaceda & Fuentes, 2016; Batty, Bettencourt, & Kirley, 2019; Chen, Li, Li, & Fang, 2021). Among these urban growth models (either process- or pattern-based), cellular automata (CA) has become one of the most prevalent approaches for spatially explicit urban growth modeling because of its strength in simulating dynamic phenomena in geographic space via the incorporation of neighboring interactions (Cilliers, van Vuuren, & van Heerden, 2021; Stanilov & Batty, 2011). Most CA models use discrete cell-based representations of geographic space in which spatial entities (cells) update their states following a bottom-up engine driven by specified transition rules (Batty, 2005; Tobler, 1970).

    View full text