Monitoring urbanization in mega cities from space
Highlights
► Classification methodologies for optical and radar data of urban footprints. ► Multi-sensoral change detection for almost 40 years monitoring of megacities. ► Processing of 27 mega cities at 4 time steps - 1975, 1990, 2000 & 2010. ► Complete processing chain for monitoring of spatial effects in cities. ► Overcoming the problem of individual case studies.
Introduction
What has happened in the past 40 years in mega cities like São Paulo, Brazil, Mumbai, India or Lagos, Nigeria shows what we can expect for many other cities: Uncontrollable urban sprawl, stretched beyond capacity to provide residents with necessary systems, the mega cities have massive congestion, poor public transportation, and noticeable lack of proper sanitation. With it come extreme socioeconomic disparities, crime issues, high vulnerability to natural and manmade risks, etc. (Fuchs et al., 1994, Kraas, 2007, Kraas and Nitschke, 2008, Mitchell, 1999). And trends imply that almost all the expected future world population growth will be absorbed by urban areas (UN, 2009). The population development of the world is expected to increase continuously from currently 6.7 billion to 9.3 billion in 2050.
While the massive dynamics of urban sprawl often overcharge the capability to govern, organize and plan new settlements, it is often even a difficult task to document and measure what has already happened. In most cases, a large amount of spatial, quantitative and qualitative data and information on urban sprawl exist. However, these data sets are seldom easily accessible or unrestricted, they are seldom stored centrally, complete, consistent, standardized, substantially documented, available as time-series or up-to-date not to mention comparability with data sets of other cities.
Measuring, analyzing and hence understanding the complex processes of urbanization needs to proceed beyond isolated case studies. Difficult tasks as there are approximately 27 mega cities today and their number is constantly increasing. Not to talk about the rising number of ‘million cities’ around the globe — where the core of urbanization is predicted for. Thus, an independent and consistent data source is required to avoid the mentioned problems and to afford comparative studies focussing on global urbanization trends and patterns. Further needs are data availability for long-time monitoring, availability for any city and considering the extensive extents of mega cities' large area coverage. One data source fulfilling these requirements, at least from a spatiotemporal point of view, are earth observation data.
In the last decade, earth observation sensors developed to a stage where global maps have been made possible on low resolution (LR) from 250 m to 2 km (Potere & Schneider, 2009). Examples are global urban extent maps based on e.g. DMSP-OLS night-time lights imagery (Elvidge et al., 2001) or MODIS data (Bartholome and Belward, 2005, Schneider et al., 2005). A list, analysis and comparison of the various available global data sets is presented and discussed by Potere and Schneider (2009). However, most of them are provided for a single time step, and the coarse geometric resolution is a clear restriction tracing the small-scale urban outlines, extents and patterns.
Even though higher resolution sensor systems are available (e.g. Landsat, SPOT, RapidEye, IRS, IKONOS, QuickBird, WorldView I and II) the provision of a global coverage – or at least of a large amount of cities – is not an easy task. Limitations such as cloud coverage, on board storage capacity, sensor utilization and sharing of the same source with other EO projects cause a several years lasting acquisition period. Furthermore, data costs and the processing effort are significant. Thus, a global coverage at the scale covered by medium (MR: here defined as 10 m to 100 m) and high resolution (HR: 1 m–10 m) to very high resolution (VHR: < 1 m) EO sensors is inexistent. Consequently, the comparative mapping and monitoring of urban areas at these scales is mostly limited in terms of temporal and spatial coverage: Case studies show high geometric and thematic capabilities of VHR optical data (Baud et al., 2010, Kemper et al., 2011, March, Niebergall et al., 2008, Taubenböck et al., 2010) or multi-sensoral combination of VHR satellite data with HR digital surface models even allow the derivation of 3-D city models (Heiden et al., 2007, Rottensteiner et al., 2007, Wurm et al., 2011). However, most of these approaches remain at the stage of case studies for individual or few cities and/or provide only a single time-step.
Small (2003) compares 14 cities at very high geometric level using QuickBird data with a sub-meter geometric resolution deriving parameters such as vegetation fraction. Research studies on long-term monitoring of the spatial effects of urbanization are mostly based on MR data from sensors such as Landsat or SPOT, having lower geometric resolutions and thus allow for fewer thematic details. These studies range from individual case studies (Griffiths et al., 2010, Herold et al., 2002, Ji et al., 2006, Luck and Wu, 2002, Stefanov et al., 2001, Yuan et al., 2005, Zhang et al., 2004) to cross-city comparisons (Schneider and Woodcok, 2008, Seto and Fragkias, 2005, Taubenböck et al., 2009, Taubenböck et al., 2010). Angel, Sheppard, and Civco (2005) even classified 90 cities at two time steps (1990 and 2000) and applied spatial metrics for comparison of the dynamics and patterns of spatial urban growth.
Today remotely sensed data and techniques are increasingly used in projects aiming at operational monitoring of the effects of spatial urbanization. The aim of the MURBANDY/MOLAND project was to provide a spatial planning tool that can be used for assessing, monitoring and modelling the development of urban and regional environments. The main feature of the project was to allow quantitative and qualitative comparisons at pan-European level, among 35 cities, regions or urban corridors subject to transformation due to policy intervention at a scale of 1:25,000 (Lavalle, Demicheli, Turchini, Casals-Carrasco, & Niederhuber, 2001). Another example is the European Urban Atlas providing consistent pan-European land use data for a total of 305 cities (Seifert, 2009), but without a land cover change component so far (Steinborn, 2010). The project CORINE Land Cover (CLC) provides a pan-European classification of land-cover with few thematic classes for cities at the minimum mapping unit of 25 ha for the time steps 1990, 2000 and 2006 (EEA, 2010). The idea of the ‘Urban Environmental Project’ (UEM), also known as ‘100 cities project’, is a baseline scientific effort to collect and analyse remotely sensed data and classifications for 100 cities (Wentz, Stefanov, Netzband, Möller, & Brazel, 2009).
Different studies have also shown that radar imagery is an excellent basis for classifying, monitoring and analyzing urban conglomerations and their development over time especially in cases of large area mapping (e.g. Dell'Acqua, 2009, Dell'Acqua and Gamba, 2003). Due to the active use of an own source of illumination, SAR systems can acquire data at day and night independent of the weather or environmental conditions — a distinct advantage compared to optical sensors especially to provide data for consistent mapping of large areas at supra-regional level: e.g. Esch, Thiel, Schenk, Roth, Mehl, et al. (2010) show urban footprint delineation for 10 sample cities using TerraSAR-X stripmap data. Tatem, Noor, and Hay (2005) created a large area map of urban extent of Kenya based on RADARSAT-1 and Landsat ETM+ data by using different texture layers as input for a neural network classifier, while Grey and Luckman (2003) studied the utilization of interferometric coherence data of the ERS satellite for mapping urban extents over a large area from Cardiff to Bristol (South Wales). Esch et al. (2011) even use TerraSAR-X stripmap data to classify different land cover types such as water, forest, urban and open areas.
This study deals with the following main aims to add to existing studies and projects using remote sensing to map and monitor urbanization and with it provide the basis for analysis and better understanding of spatial effects. Thereby the presented research focuses on the mega cities of the world:
- •
maximizing the time period for monitoring the spatial effects of urbanization to a span of almost 40 years (1972–2010)
- •
implementing a pragmatic, consistent and efficient processing chain to derive urban footprint classifications from a capacious multi-sensoral data base
- •
multi-temporal change detection for mapping and quantifying spatial urban growth
Section snippets
Study sites ‘mega cities’
The term‚ ‘mega city’ refers to the largest category of urban agglomerations. The UN (2007) define mega cities quantitatively as conurbation having more than 10 million inhabitants. Today, based on official numbers (UN, 2009), there are 27 mega cities throughout the world and their number is expected to increase. Cities such as Bangkok, Hyderabad, Chicago, Caracas, just to name a few, are already close to becoming a mega city. India is a prominent example for dynamic mega city development —
Remotely sensed data
Mapping urban change for the large areas of currently 27 current mega cities would not have been a feasible task even a few years ago due to limitations in data availability and costs, storage capacities as well as costs and speed of processing.
The overall aim of maximizing the time period for monitoring the spatial effects of urbanization limits the range for commercially or freely available EO data sets. Today's large extents of the mega cities – e.g. the dimension of Istanbul in east–west
The “urban footprint” product
The terms ‘urban area’ or ‘urban footprint’ are widely used in literature and basically refer to the spatial extent of urbanized areas on a regional scale; however, it is a fuzzy and inconsistent definition. The most obvious approach – administrative boundaries of cities – does not meet the requirements as discussed above. The Office for National Statistics of the UK (2001) suggests three alternative approaches to spatially define an urbanized area. It may be defined in terms of the areas for
Discussion and applications
Global urban classification products, as presented in the Introduction, prove remote sensing to be capable to overcome the problem of isolated case studies at low resolution level. This study underlines the remote sensing capabilities to enable long-time multi-temporal approaches (almost 40 years) on a medium geometric resolution for large areas.
With the presented multi-temporal approach based on multi-sensoral remotely sensed data two-dimensional urban sprawl is mapped. However, urban expansion
Outlook
The constantly increasing availability and accessibility of modern remote sensing technologies provides new opportunities for a wide range of urban applications. A pragmatic, straight forward and operationalized process chain proves that to date remote sensing data and techniques allow for extensive monitoring of spatial effects in urban areas over a comparatively long time period providing an independent, area-wide, reproducible and consistent mapping result.
The techniques applied to derive
Acknowledgments
The authors would like to thank the TerraSAR-X Science Team for providing the SAR data for this study. Furthermore we would like to thank Lisa Kluin, Sarah Abelen, Tobias Ullman and Dennis Edler for their great support.
References (87)
- et al.
Understanding heterogeneity in metropolitan India: The added value of remote sensing data for analyzing sub-standard residential areas
- et al.
Multi-resolution, objectoriented fuzzy analysis of remote sensing data for GIS-ready information
ISPRS Journal of Photogrammetry and Remote Sensing
(2004) A review of assessing the accuracy of classification of remotely sensed data
Remote Sensing of Environment
(1991, Jul.)- et al.
Feature selection for classification
- et al.
Nighttime lights of the world: 1994–95
ISPRS Journal of Photogrammetry and Remote Sensing
(2001) - et al.
Mapping megacity growth with multi-sensor data
- et al.
Determination of robust spectral features for identification of urban surface materials in hyperspectral remote sensing data
Remote Sensing of Environment
(2007) - et al.
The spatiotemporal form of urban growth: Measurement, analysis and modeling
Remote Sensing of Environment
(2003) - et al.
Monitoring and modelling of urban sprawl using remote sensing and GIS techniques
- et al.
Characterizing urban sprawl using multi-stage remote sensing images and landscape metrics
Are European cities becoming dispersed? A comparative analysis of 15 European urban areas
Landscape and Urban Planning
Spatial–temporal gradient analysis of urban green spaces in Jinan, China
Building detection by fusion of airborne laser scanner data and multi-spectral images: Performance evaluation and sensitivity analysis
ISPRS Journal of Photogrammetry and Remote Sensing
High spatial resolution spectral mixture analysis of urban reflectance
Monitoring urban land cover change: An expert system approach to land cover classification of semiarid to arid urban centers
Remote Sensing of Environment
Urban sprawl: metrics, dynamics and modelling using GIS
International Journal of Applied Earth Observation and Geoinformation
Assessing the accuracy of satellite derived global and national urban maps in Kenya
Remote Sensing of Environment
Urbanization in India — Spatiotemporal analysis using remote sensing data
Computers, Environment and Urban Systems
Land use change analysis in the Zhujiang Delta of China using satellite remote sensing, GIS and stochastic modelling
Land cover classification and change analysis of the Twin Cities (Minnesota) Metropolitan Area by multitemporal Landsat remote sensing
Remote Sensing of Environment
A GIS-based gradient analysis of urban landscape pattern of Shanghai metropolitan area, China
Interactive classification of urban areas using decision trees
GLC2000: A new approach to global land cover mapping from Earth observation data
TanDEM-X mission overview and status
Mega cities without global functions
TerraSAR-X SAR processing and products
Urban crisis in India: new initiatives for sustainable cities
Development in Practice
Footprints on the planet: A search for an environmental ethic
ISBN-:
Hyperspectral imaging: Techniques for spectral detection and classification
An approach to linking remotely sensed data and areal census data
International Journal of Remote Sensing
KH-series satellite imagery and Landsat MSS data fusion in support of assessing urban land use growth
Quantifying and describing urbanizing landscapes in the northeast United States
Photogrammetric Engineering and Remote Sensing
The role of SAR sensors
Texture-based characterization of urban environments on satellite SAR images
IEEE Transactions on Geoscience and Remote Sensing
Radar backscatter analysis for urban environments
International Journal of Remote Sensing
Can Earth observation help to improve information on population? Indirect population estimations from EO derived geo-spatial data: Contribution from GMOSS
Characterization of land cover types in TerraSAR-X images by combined analysis of speckle statistics and intensity information
IEEE Transactions on Geoscience and Remote Sensing, IEEE Xplore
Delineation of urban footprints from TerraSAR-X data by analyzing speckle characteristics and intensity information
IEEE Transactions on Geoscience and Remote Sensing
Some urban measurements from Landsat data
Photogrammetric Engineering and Remote Sensing
An examination of some problems and solutions in monitoring urban areas from satellite platforms
International Journal of Remote Sensing
Cited by (377)
The analysis of the spatio-temporal changes and prediction of built-up lands and urban heat islands using multi-temporal satellite imagery
2024, Sustainable Cities and SocietyDoes urban growth mean the loss of greenness? A multi-temporal analysis for Chinese cities
2023, Science of the Total EnvironmentA multi-scale weakly supervised learning method with adaptive online noise correction for high-resolution change detection of built-up areas
2023, Remote Sensing of EnvironmentSDGSAT-1 nighttime light data improve village-scale built-up delineation
2023, Remote Sensing of Environment