Green growth? On the relation between population density, land use and vegetation cover fractions in a city using a 30-years Landsat time series

https://doi.org/10.1016/j.landurbplan.2020.103857Get rights and content

Highlights

  • Framework assessing paths and threats for green in times of urban densification.

  • We integrate population- with vegetation density from Landsat remote sensing data.

  • Compact and dispersed districts feature positive developments, i.e. green growth.

  • Berlin developed into a greening growing city almost citywide.

  • We find sealing in residential land use classes and greening in public greenspaces.

Abstract

Both compact and dispersed green cities are considered sustainable urban forms, yet some developments accompanied with these planning paradigms seem problematic in times of urban growth. A compact city might lose urban green spaces due to infill and a dispersed-green city might lose green in its outskirts through suburbanisation. To study these storylines, we introduce an operationalised concept of contrasting changes in population density (shrinkage or growth) with vegetation density (sealing or greening) over time. These trends are ascribed to different land use classes and single urban development projects, to quantify threads and pathways for urban green in a densifying city. We mapped the development in vegetation density over 30 years as subpixel fractions based on a Landsat remote sensing time series (for 2015: MAE 0.12). The case study city Berlin, Germany, developed into a city that is both gaining in vegetation–greening–and population–growing–in recent years but featured highly diverse trends for both compact and green city districts before that. Pathways to achieve a greening-growing scenario in a compact city include green roofs, brownfield and industrial revitalisation, and bioswales in predominantly green city districts. A threat for compact cities pose infill developments without greening measures. A threat for dispersed-green cities is microsealing in private residential gardens–gravel gardens–or car parking infrastructure. We conclude that neither a compact nor a dispersed-green city form concept logically leads to a development towards more environmental quality–here vegetation density–in times of densification but rather context specific urban planning.

Introduction

An increase in urban green cover and increases in population density are often considered to be antagonists (Haaland & van den Bosch, 2015). A number of studies find that growth in population density – whether compact inner-city growth or dispersed suburbanisation on the urban fringes - is inevitably connected to negative outcomes such as a reduction in vegetation cover (Elmqvist et al., 2018, Nuissl et al., 2009, Wolff and Haase, 2019). This notion of over-crowding in regard to population density increases can be found across the literature. However, none of these studies actually measures how vegetation cover has developed in its full spatial-temporal configuration for an entire city with population density developments. We therefore ask ourselves, whether both entities can grow at the same time and under different conditions in a city or parts of it and what the paths and threats for this development are?

From an urban ecological view, growth in population density should go hand in hand with an increase in vegetation density to prevent a degradation of viable ecosystem services that are critically important for human well-being such as regulating air temperature or local recreation (Pauleit et al., 2019). Therefore it is worrying that a wide range of authors have observed both a trend in the degradation of urban vegetation, and a worldwide increase in population density in cities, leading to a reduction in per capita ecosystem service provisioning (Elmqvist et al., 2018, Fuller and Gaston, 2009).

One prominent way of dealing with densification in urban planning and the green city is the implementation of greening strategies following the concept of green infrastructure according to Pauleit et al. (2019), as integrating vegetation into the city provides numerous beneficial flows of ecosystem services to its inhabitants (Andersson et al., 2019, Haase et al., 2014) as well as offering considerable potential for biodiversity conservation (Schwarz et al., 2017). Important tools for growing cities include small-scale elements of green infrastructure such as green roofs, roadside vegetation such as trees and hedgerows or bioswales and the formation of new green spaces such as parks. However, the processes on private grounds around buildings i.e. front and backyards or gardens are overlooked. Such areas can make up about 40% of total vegetation cover in a city and are very important for comprehensive spatial scale assessments (Haase, Jänicke, & Wellmann, 2019b).

The current European planning paradigm largely draws on the compact cities approach to sustainability (De Roo & Miller, 2019). The compact city paradigm emphasizes the need to integrate different land uses in close proximity to enable short daily commutes to save energy and land (Echenique, Hargreaves, Mitchell, & Namdeo, 2012). The opposite of the compact city is the concept of the dispersed city (in Europe also referred to as the ‘garden city’), which emerged in industrial England in the 19th century. Here, urban structure is shaped by open built-up space, low rise and low-density housing, where buildings, streets, green spaces and other urban amenities form a mosaic-like and more dispersed pattern (Holden & Norland, 2005, p.2148). In a dispersed city, we have less congestion but higher segregation. In terms of environmental quality, the dispersed city is also supposed to be a green city due to its higher share of open space and less air pollution (Westerink et al., 2012).

Most literature suggests that cities can generally be characterised as a compact or a dispersed city (Holden and Norland, 2005, Jenks and Burgess, 2000, Westerink et al., 2013). However, experiences from different cities across Europe over time tell us that we often observe both: growth in the form of densification and infill as well as dispersed growth at the periphery (Wolff & Haase, 2019). The processes however, of both population and built space densification in terms of the change in vegetation cover are often unequally distributed in a city, raising questions of equity and justice about access to green spaces (Kabisch & Haase, 2014). From this arises the need for a spatially and temporally explicit analysis of population- and vegetation density developments in an integrated manner to evaluate and design better policies. Important is thereby to enable active citizenship for instance with open data and result policies, working towards a kind of mosaic governance (Buijs et al., 2016).

To portray spatial-temporal trends in urban vegetation cover, Earth observation based time series analysis is very appropriate (Li, Song, Liujun, Yanan, & Di, 2017). Landsat satellite imagery is particularly suitable to monitor land cover and land use change because of its long recording history and open data policy (Wulder & Coops, 2014). As the resolution of Landsat (30 m) is still higher than the size of many urban objects, the problem of mixed pixels consisting of multiple land cover classes is a remaining challenge (van der Linden et al., 2018). In contrast to discrete land cover classification that tends to underrepresent smaller elements that occur comparatively frequently in an urban vegetation context (Zhu et al., 2015), spectral unmixing methods describe the quantitative composition of land cover within pixels and thus the continuous character of land cover throughout urban spaces (Small & Sousa, 2016). Spectral unmixing is a widely used concept in Earth observation and non-parametric machine learning regression approaches are becoming more frequently used in different urban and non-urban land cover composition mapping applications (Suess et al., 2018, van der Linden et al., 2018). However, contributions to urban ecological studies (Haase et al., 2019b) or for long time series analyses in cities (Michishita et al., 2012, Schug et al., 2018) are still rare.

Having the new time series based on Earth observation technologies at hand, we are able to assess more comprehensively whether green growth—or green shrinkage, respectively—is occurring or if this is merely envisioned and identify properties and characteristics that are typical for both types of city paradigms: the compact and the dispersed-green city, as outlined by various authors. Such information helps us to uncover what Neuman (2005) calls ‘the compact city fallacy’ that cities are a continuous co-evolutionary process rather than a form or a state. In doing so, in this study, we quantify vegetation and population density changes and co-relate both on different spatial scales for the city of Berlin, Germany to answer the following questions:

  • How can Landsat time series data and spectral unmixing contribute to a better understanding and hypothesis testing for the spatial realisation of different urban planning concepts?

  • What are long-term urban vegetation trends in different land use classes?

  • Can densification of built space be detected in specific land use classes and recommendations for action be provided?

  • What are the paths and threats for green-growth in compact or dispersed areas?

Section snippets

Study area

Berlin is a Central European city (52° N, 13° E) with an area of about 900 km2 (Fig. 1) featuring 3.6 million inhabitants (Statistical Office for Berlin-Brandenburg, 2018). The main constituents of green and blue infrastructure are forests with 18%, parks and allotment gardens with 12%, a river network with 7% and agriculture with 4% of Berlins surface coverage (Senate administration Berlin, 2018). Settlement and transportation infrastructure, in turn, cover about 60% of the city’s surface.

Data & methods

We acquired Landsat satellite imagery for seven years between 1988 and 2018 (Fig. 3). Imagery was pre-processed and a regression-based unmixing approach was performed in order to generate fraction maps of vegetated and non-vegetated surfaces. Results were validated for the year 2015. Two indicators were derived from the resulting historic and recent fraction maps: Firstly, a time-series contrasting vegetation and population densities portraying the dynamics of both entities in an integrated way

Spectral unmixing: Accuracy assessment

Fig. 4 illustrates the accuracy of the binary unmixing product for the year 2015 (see Fig. A1 in the Appendix), featuring an overall MAE of 0.12. Based on a linear model fitted to the estimate and reference data, the model overestimates vegetation in areas with lower to medium vegetation shares but shows a generally good fit in areas with higher vegetation shares with an intercept of 0.15 and a slope of 0.81.

Regarding the long-term vegetation trends, we found that 73% of the profiles match with

Discussion

Contrary to European (Pauleit et al., 2019) and international studies (McDonald et al., 2019), in the study undertaken here for Berlin, green growth can be found in both compact and dispersed areas of the city along a comparably long temporal trajectory from 1989 to 2018. This highlights that the urban form is not necessarily the only decisive factor or essential in the analysis of a city’s environmental characteristics and green land cover, as other authors have also suggested (Echenique et

Conclusions

Compact and dispersed city dynamics have long been understood as archetypical antagonists in urban planning. With this Earth observation data supported research, we want to contribute to a non-binary vision of urban (form) as we conclude that neither a compact urban form, nor a dispersed city leads to transparently deducible increases in environmental quality, measured here in terms of vegetation density. Rather, both areas have their unique configurations and processes that shape these

Open data policy

The fractional vegetation cover time-series for Berlin is available for download here: https://doi.org/10.5281/zenodo.3870592.

CRediT authorship contribution statement

Thilo Wellmann: Conceptualization, Methodology, Software, Writing - original draft, Data curation, Formal analysis, Visualization. Franz Schug: Conceptualization, Methodology, Writing - review & editing. Dagmar Haase: Supervision, Writing - review & editing. Dirk Pflugmacher: Supervision. Sebastian van der Linden: Conceptualization, Supervision, Methodology, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

Thilo received a Humboldt research track scholarship and receives a scholarship by the Deutsche Bundesstiftung Umwelt DBU (German Federal Environmental Foundation). Thilo and Dagmar wish to thank the CLEARING HOUSE (Collaborative Learning in Research, Information-sharing and Governance on How Urban forest-based solutions support Sino-European urban futures) Horizon 2020 project (No 1290/2013). Dagmar also benefited from the GreenCityLabHue Project (FKZ 01LE1910A), the Horizon 2020 innovation

References (52)

  • A. Okujeni et al.

    Extending the vegetation-impervious-soil model using simulated EnMAP data and machine learning

    Remote Sensing of Environment

    (2015)
  • A. Okujeni et al.

    Support vector regression and synthetically mixed training data for quantifying urban land cover

    Remote Sensing of Environment

    (2013)
  • S. Pauleit et al.

    Advancing urban green infrastructure in Europe: Outcomes and reflections from the GREEN SURGE project

    Urban Forestry and Urban Greening

    (2019)
  • F. Schug et al.

    Mapping patterns of urban development in Ouagadougou, Burkina Faso, using machine learning regression modeling with bi-seasonal Landsat time series

    Remote Sensing of Environment

    (2018)
  • N. Schwarz et al.

    Understanding biodiversity-ecosystem service relationships in urban areas: A comprehensive literature review

    Ecosystem Services

    (2017)
  • C. Small et al.

    Humans on Earth: Global extents of anthropogenic land cover from remote sensing

    Anthropocene

    (2016)
  • S. Suess et al.

    Characterizing 32 years of shrub cover dynamics in southern Portugal using annual Landsat composites and machine learning regression modeling

    Remote Sensing of Environment

    (2018)
  • J. Tratalos et al.

    Urban form, biodiversity potential and ecosystem services

    Landscape and Urban Planning

    (2007)
  • T. Wellmann et al.

    Urban land use intensity assessment: The potential of spatio-temporal spectral traits with remote sensing

    Ecological Indicators

    (2018)
  • T. Wellmann et al.

    Earth observation based indication for avian species distribution models using the spectral trait concept and machine learning in an urban setting

    Ecological Indicators

    (2020)
  • J.R. Wolch et al.

    Urban green space, public health, and environmental justice: The challenge of making cities “just green enough”

    Landscape and Urban Planning

    (2014)
  • S. Yu et al.

    View-based greenery: A three-dimensional assessment of city buildings’ green visibility using Floor Green View Index

    Landscape and Urban Planning

    (2016)
  • E. Andersson et al.

    Enabling green and blue infrastructure to improve contributions to human well-being and equity in urban systems

    BioScience

    (2019)
  • Berlin Brandenburg Statistical Office. (n.d.)....
  • L. Breiman

    Random forests

    Machine Learning

    (2001)
  • G. De Roo et al.

    Compact cities and sustainable urban development: A critical assessment of policies and plans from an international perspective

    (2019)
  • Cited by (61)

    View all citing articles on Scopus
    View full text