Mapping urban dynamics (1992–2018) in Southeast Asia using consistent nighttime light data from DMSP and VIIRS

https://doi.org/10.1016/j.rse.2020.111980Get rights and content

Highlights

  • A new mapping framework is proposed to monitor annual urban dynamics.

  • This is the first work for mapping urban dynamics using fused nighttime lights.

  • A view of annual urban dynamics in Southeast Asia (1992–2018) is presented.

  • Our framework is transferable for monitoring urban dynamics over large areas.

Abstract

The long-term urban dynamics at regional and global scales is essential to understanding the urbanization processes and environmental consequences for providing better scientific insights and effective decision-making. The time series of consistent nighttime light (NTL) data generated by integrating the Defense Meteorological Satellite Program-Operational Linescane System (DMSP-OLS) and the Suomi National Polar-Orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) provide a longer consistent record of the nightscape beyond a single dataset for monitoring urban dynamics. In this study, we developed a new framework based on the spatial variation of NTL gradient (SVNG) to map urban dynamics in Southeast Asia using the consistent NTL data (1992–2018). First, we identified the potential urban clusters in the region using the cluster-based segmentation approach in 2018. Second, we applied the SVNG framework in each potential urban cluster to extract the initial annual urban extent from corresponding time-series NTL images (1992–2018). Finally, we performed a temporal consistency check on the initial urban extent to obtain the final urban sequence in Southeast Asia. The evaluation on the spatiotemporal patterns and consistency of urban dynamics using other urban products indicates that the SVNG framework can effectively capture the urban dynamics in areas with different development levels and patterns. Moreover, we investigated urban dynamics in Southeast Asia at the local, national, and regional scales. This study opens new research avenues for monitoring and understanding the long-term urban dynamics and the pathways of urban growth from local to global scales.

Introduction

Monitoring and modeling of urban dynamics are crucial for understanding the urbanization process and corresponding environmental consequences in a changing world (Li et al., 2019; Zhu et al., 2019). This need is especially urgent for parts of Asia, which have experienced rapid urbanization over the past few decades (Chen, 2007; Cohen, 2006; Zhao et al., 2018). For instance, the urban population in Southeast Asia has increased from 27 million in 1950 to about 262 million in 2010, accounting for 15% and 44% of the total population in this region, respectively, and this number is projected to increase to about 500 million by 2050, far more larger than the non-urban population (UN, 2018). Through the land use/cover change, urbanization intensifies urban heat island (Li et al., 2017; Shi et al., 2019), increases anthropogenic greenhouse gas emissions and energy consumption (Guneralp et al., 2017; Sugar et al., 2012; Wang et al., 2019), and profoundly influences ecosystem (Alberti et al., 2017; Pauchard et al., 2006; Zhao et al., 2019a). As such, long-term and consistent mapping of urban areas is essential to track, understand, predict the pathways of urban growth, and address these urbanization-induced issues for future sustainable development.

Satellite-based artificial nighttime light (NTL) observations provide a unique proxy measure for unveiling urbanization and regional development (Elvidge et al., 2017; Elvidge et al., 1999; Elvidge et al., 2001; Li and Zhou, 2017b; Yang et al., 2019). NTL images obtained from different satellite sensors, particularly the Defense Meteorological Satellite Program-Operational Linescan System (DMSP-OLS), and the Visible Infrared Imaging Radiometer Suite (VIIRS) instrument carried by the Suomi National Polar- orbiting Partnership satellite (Suomi-NPP), have been extensively used for urban mapping at the local, regional, and global scales (Chen et al., 2019; Dou et al., 2017; Liu et al., 2012; Shi et al., 2014; Zhou et al., 2018; Zhou et al., 2015). Studies on urban mapping and urbanization estimation using NTL observations, at various spatial scales, can be grouped into three major categories including the threshold-based method, supervised classification, and the statistic relationship between NTL indices and observations of urban areas (Zhao et al., 2019b). Among them, the continuously improved threshold-based methods, which provide reliable estimates of the optimal NTL thresholds for characterizing the urbanized areas in different regions and periods, always benefit from either the auxiliary data such as remote sensing products (Henderson et al., 2003; Liu et al., 2012; Xie et al., 2019) and statistical data (He et al., 2006; Shi et al., 2014; Zhao et al., 2016), or the spatial variations of NTL at different scales (Ma et al., 2015; Su et al., 2015; Zhou et al., 2018; Zhou et al., 2014; Zhou et al., 2015). The latter one, without additional information/datasets as references, always shows potentials in mapping urban extents in different regions/periods.

Limited understanding of urban-rural NTL spatial structures makes it challenging in well identifying the urban extent of different zones with diverse patterns. Because of the effects of blooming and shrink of NTL, the built-up settlements in urban cores are likely to be overestimated, while some small built-up patches in developing towns are easy to be ignored (Cao et al., 2009; Henderson et al., 2003; Imhoff et al., 1997; Small et al., 2005). Moreover, the diversity of urbanization patterns in different regions and periods also hinders a comprehensive capturing of the NTL spatial patterns in urban, suburban, and rural areas (Ma et al., 2015). These limitations challenge the applications of the threshold-based methods in acquiring the urban boundary matching the spatial extent of urbanized areas using NTL images. For instance, Ma et al. (2015) proposed a quantitative partitioning approach for DMSP NTL to identify different types of urban sub-regions based upon the quadratic relationship between the pixel-level NTL and corresponding NTL gradient. However, this approach cannot well overcome the blooming effect in the highly lighted areas with high intensity of human activity, likely leading to overestimating the urban extents in large cities. Zhou et al. (2018) proposed a quantile-based approach to extract the urban extents from NTL images by removing the rural and suburban areas sequentially for the potential urban domains in different regions and periods. Though this approach largely eliminates the overestimation of the urban areas in large cities caused by the blooming effect, the urban extent is likely to be misestimated when the variations of the urban-rural NTL gradient within the potential urban domain is gradual. In short, it tends to be difficult to portray the urban extent accurately over different space and time owing to the lack of effective measure and characterization of the locally spatial variations of NTL signals. Therefore, an improved approach to better identify the different patterns of NTL variations from rural areas to urban cores is pressing needed for capturing the urban dynamics from NTL observations.

Despite the valuable time series of NTL from DMSP (1992–2013) and VIIRS (2012–), a longer urban dynamics (1992–present) from regional to global scales is still lacking due to the inconsistency of NTL from different sources. Because of the significant differences of the sensor performance and overpass time between DMSP and VIIRS, the NTL brightness in the same period and the spatiotemporal trends and variations of NTL time series, obtained from the two sensors, are inconsistent and incomparable (Zhao et al., 2019b). The severe inconsistency of the two different sensors makes it difficult for the multi-source NTL observations to be used directly for long time series analyses from 1992 to present. Although recent studies have contributed to obtaining consistent NTL signals from DMSP and VIIRS (Zhao et al., 2020), the NTL-based monitoring of urban dynamics currently are limited to short-period studies, using either DMSP NTL data spanning 1992–2013 or VIIRS NTL data since 2012. Therefore, the consistent and continuous NTL data integrated from DMSP and VIIRS show great potential for the further applications of NTL in estimating long-term urban dynamics.

Considering the limitations in the current urban mapping approach and NTL data, a comprehensive characterization of urban-rural NTL spatial structures and further application of consistent NTL data are needed to monitor the long-term urban dynamics in areas with different development levels from regional to global scales. Therefore, we aimed to develop an improved framework for mapping annual urban extents and estimating urban dynamics in Southeast Asia from 1992 to 2018, using the consistent NTL data generated previously (Zhao et al., 2020). The framework in this study, based on the spatial variations of NTL gradient (SVNG), was developed by combing the advantages of the two existing methods using the DMSP stable NTL (Ma et al., 2015; Zhou et al., 2018). This improved framework aims at well capturing different patterns of NTL spatial variations from urban cores to rural areas for accurately extracting urban extents from the potential urban clusters containing urban areas, as well as suburban and rural areas. These urban extents delineated using the DMSP stable NTL, generally associated with dense human settlements, are contiguous areas with urban functions and indicative of urbanization-related socioeconomic activities. The remainder of this article details the study area and dataset (Section 2), the three major parts of the urban mapping approach (Section 3), the evaluation and discussion of the approach and results (Section 4), and conclusions (Section 5).

Section snippets

Study area

The study area in this paper is Southeast Asia (Fig. 1), with the uneven process of urbanization across the region. It can serve as an ideal experiment area to evaluate the performance of our approach in delineating urbanized areas with different development levels. Southeast Asia consists of two distinct geographic regions. Mainland Southeast Asia comprises West Malaysia, Laos, Thailand, Cambodia, Vietnam and Myanmar. Maritime Southeast Asia includes the countries of Philippines, Indonesia,

Methodology

A new method was developed for mapping the dynamics of urban extents from consistent NTL data (Fig. 2) based on a combination of two existing urban mapping approaches proposed using the DMSP stable NTL data (Ma et al., 2015; Zhou et al., 2018). First, we identified the potential urban clusters from the filtered NTL data using a cluster-based segmentation approach (Zhou et al., 2014). Second, we developed a framework based on spatial variations of NTL gradient (SVNG) to delineate the annual

Evaluations of mapped urban extents

Because of the difference between definitions, objectives or input data, urban extents could vary widely among different studies (Liu et al., 2014; Zhou et al., 2018). In this study, we performed visual and quantitative evaluations of the extracted urban extents. First, the NTL-derived urban boundaries were visually compared with the dense human settlements obtained from other remote sensing products and Google Maps. Second, total areas and their temporal trends calculated from the NTL-derived

Conclusions

In this study, we developed a new framework for mapping urban dynamics from the consistent NTL observations and estimated urban dynamics in Southeast Asia from 1992 to 2018. We first identified the potential urban clusters for the entire region using a segmentation approach. We then applied the SVNG framework to delineate the initial urban extents for each potential urban cluster, using two different strategies based on the changing patterns of urban-rural NTL gradient. Finally, we performed

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.

Acknowledgements

This study was funded by the National Key Research and Development Program of China (No. 2017YFB0504205), the Surveying and Mapping Geo-information Nonprofit Specific Project of China (No. 201512033), the Major State Basic Research Development Program of China (No. 2015CB954101), the National Natural Science Foundation of China (No. 41171332 and No. 41571388), and the program of China Scholarships Council (No.201806190136).

References (44)

  • A. Pauchard et al.

    Multiple effects of urbanization on the biodiversity of developing countries: the case of a fast-growing metropolitan area (Concepción, Chile)

    Biol. Conserv.

    (2006)
  • C. Small et al.

    Spatial analysis of global urban extent from DMSP-OLS night lights

    Remote Sens. Environ.

    (2005)
  • S. Wang et al.

    Urban− rural gradients reveal joint control of elevated CO2 and temperature on extended photosynthetic seasons

    Nat. Ecol. Evol.

    (2019)
  • Y. Xie et al.

    Temporal variations of artificial nighttime lights and their implications for urbanization in the conterminous United States, 2013–2017

    Remote Sens. Environ.

    (2019)
  • Y. Zhou et al.

    A cluster-based method to map urban area from DMSP/OLS nightlights

    Remote Sens. Environ.

    (2014)
  • Y. Zhou et al.

    A global record of annual urban dynamics (1992–2013) from nighttime lights

    Remote Sens. Environ.

    (2018)
  • Z. Zhu et al.

    Understanding an urbanizing planet: strategic directions for remote sensing

    Remote Sens. Environ.

    (2019)
  • M. Alberti et al.

    Global urban signatures of phenotypic change in animal and plant populations

    Proc. Natl. Acad. Sci.

    (2017)
  • Z. Chen et al.

    Mapping global urban areas from 2000 to 2012 using time-series Nighttime light data and MODIS products

    IEEE J. Select. Topics Appl. Earth Observ. Remote Sens.

    (2019)
  • Y. Dou et al.

    Urban land extraction using VIIRS nighttime light data: an evaluation of three popular methods

    Remote Sens.

    (2017)
  • C. Elvidge et al.

    A fifteen year record of global natural gas flaring derived from satellite data

    Energies

    (2009)
  • C.D. Elvidge et al.

    Why VIIRS data are superior to DMSP for mapping nighttime lights

    Proc. Asia-Pacific Adv. Network

    (2013)
  • Cited by (0)

    View full text