Characterizing the Growth Patterns of 45 Major Metropolitans in Mainland China Using DMSP/OLS Data

Tao Jia, Kai Chen, Jiye WangRemote 2017 · DOI

TL;DR

This study uses DMSP/OLS nighttime light data from 1992-2013 to analyze urban growth patterns in 45 major Chinese metropolitans, distinguishing between "intensification" (increased brightness within existing areas) and "expansion" (spreading to new areas). The findings reveal distinct aggregate and individual growth trends, highlighting how metropolitan size influences these patterns and suggesting that intensification and expansion are not always directly linked, especially in larger cities.

What Problem Does This Paper Address?

Understanding how cities grow is crucial for sustainable urban planning and policy-making. While urban growth is a well-studied phenomenon, previous research often focused on either urban expansion (the physical spread of cities) or urbanization (the broader process of population shift to urban areas) in isolation. Furthermore, many studies examined individual cities or regions rather than providing a comparative analysis across multiple metropolitans. This paper aims to fill this gap by offering a systematic, comparative investigation of metropolitan growth in Mainland China, specifically looking at two key aspects: intensification (making existing urban areas brighter, reflecting increased human activity and development) and expansion (the outward spread of urban areas into previously undeveloped land). This dual approach provides a more nuanced understanding of urban development.

Methods

The core of this research lies in analyzing DMSP/OLS (Defense Meteorological Satellite Program/Operational Linescan System) nighttime light (NTL) data. To ensure data quality and comparability over time, the authors performed several preprocessing steps:

  • Inter-calibration: Correcting for variations between different satellites and sensor drift using a quadratic polynomial regression model.
  • Intra-annual composition and inter-annual correction: Further refining the data to ensure consistency and to reflect actual growth trends, assuming that NTL extent expands and brightness increases over time.

To quantify growth, two key metrics were developed:

  • Intensification Level (IL): Measured as the change in Digital Number (DN) values normalized by metropolitan area. This captures how much brighter existing urban areas have become.
  • Expansion Level (EL): Measured as the change in DN extent normalized by metropolitan area. This quantifies the outward spread of urban areas.

Expansion was further categorized into three types:

  • Adjacent Expansion: New lit pixels adjacent to existing lit areas.
  • Scatter Expansion: New lit pixels isolated from existing lit areas, often with irregular shapes.
  • Corridor Expansion: New lit pixels forming linear patterns, like roads.

The study also employed hierarchical clustering (using Spearman correlation coefficients) to group metropolitans with similar intensification patterns.

Datasets

The primary dataset used in this study is:

  • DMSP/OLS Nighttime Light (NTL) Data: Annual stable lights products from 1992 to 2013. This dataset provides a consistent, long-term view of human activity as indicated by artificial lighting.

The authors also mention using Landsat TM images for accuracy assessment of expansion patterns in selected metropolitans, and socio-economic data (GDP and population) for correlation analysis.

Key Insights and Ideas

  • Intensification and Expansion are Distinct but Intertwined: The study highlights that urban growth isn't just about spreading out; it's also about developing existing areas more intensely. These two processes can occur independently or in conjunction.
  • Metropolitan Size Matters: The size of a metropolitan area significantly influences its growth patterns. Larger cities tend to intensify more, while smaller and medium-sized cities tend to expand more.
  • Cyclical vs. Decreasing Trends: While intensification shows a cyclical pattern over time (with peaks around 1994 and 2010, possibly linked to economic policies), expansion shows an overall decreasing trend. This suggests that while cities are still growing outwards, the rate of outward spread might be slowing down relative to intensification.
  • Diverse Expansion Strategies: Different-sized cities employ different expansion strategies. Moderate cities favor adjacent expansion, while smaller cities are more prone to scattered or corridor-like growth.
  • Geographic Constraints Influence Intensification: The study found that geographical features, like mountainous terrain (Chongqing) or flat plains (Shihezi), can influence a city's ability to intensify, leading to counter-intuitive findings where smaller cities in favorable terrains show higher intensification.
  • Intensification and Expansion Aren't Always Linked: For small metropolitans, there's a strong positive correlation between intensification and expansion, suggesting that increased activity drives outward growth. However, for large and moderate metropolitans, this correlation is weak or non-existent, implying that these cities might be managing their growth more strategically, perhaps due to awareness of negative impacts.

Key Results

  • Aggregate Intensification: Showed a cyclical pattern with peaks in 1994 and 2010. Large metropolitans had, on average, 48.4% higher intensification than moderate and 133.9% higher than small ones.
  • Aggregate Expansion: Displayed a decreasing trend over time. Moderate metropolitans showed 71.2% higher expansion than large ones, and 21.0% higher than small ones.
  • Expansion Types: Moderate metropolitans were more likely to expand adjacently (73.1% higher than large, 24.1% higher than small). Small metropolitans were more likely to experience scatter or corridor expansion (28.6% higher than moderate, 80.0% higher than large for scatter).
  • Clustering of Intensification: Hierarchical clustering identified 7 groups of metropolitans with similar intensification trends, revealing new spatial divisions beyond administrative boundaries.
  • Correlation with GDP: Both intensification and expansion showed a high correlation with the change in GDP per unit area, validating the metrics' reliability. On average, IL and EL explained about 49.2% of the variability in GDP change.
  • Accuracy of Expansion Validation: Landsat TM image comparisons for three selected metropolitans showed overall accuracies of 86.9% (Hefei), 78.4% (Qiqihar), and 56.3% (Shihezi) for expansion analysis.

Limitations and Future Work

  • Spurious Variability in NTL Data: Despite preprocessing, the DMSP/OLS data can still contain spurious variability. Future work could explore advanced preprocessing techniques like empirical orthogonal function analysis and the use of pseudoinvariant features.
  • Saturation Problem: The DMSP/OLS sensor's saturation (DN value of 63) can limit the detection of further increases in brightness in already very bright areas. While the study found this to be a minor issue for most metropolitans (only 2% saturated pixels on average), it might affect results for cities like Shenzhen, Shanghai, and Beijing.
  • Coarse Spatial Resolution: The 1 km spatial resolution of DMSP/OLS data limits its application for small-scale metropolitan studies. Future research should consider higher-resolution sensors like VIIRS and NPOESS.
  • Underlying Mechanisms: While the study demonstrates the correlation between IL/EL and GDP, further investigation is needed to fully understand the underlying mechanisms driving these relationships.

Summary updated

Abstract

Understanding growth patterns at the metropolitan level is instructive for better planning and policy making on sustainable urban development. Using DMSP/OLS data from 1992 to 2013, this article aims to investigate growth patterns of major metropolitans in Mainland China from the aspects of intensification and expansion. We start by calibrating the DMSP/OLS data and selecting 45 major metropolitans. On intensification, results suggest that aggregately, metropolitans displayed cyclical pattern over time and large metropolitans tended to have higher levels of intensification than moderate or small ones. Individually, metropolitans with similar intensification over time could be clustered together using Dendrogram, and evolution pattern of the clusters exhibited similarity to the aggregated one. On expansion, results show that aggregately metropolitans displayed a decreasing trend over time, and moderate or small metropolitans tended to have higher levels of expansion than large ones. Particularly, moderate metropolitans were more likely to expand adjacently, and small ones were more likely to experience scatter or corridor expansion. Each metropolitan can be represented by a mixed expansion model over time, which might tell where and how much expansion occurred in the current year. Furthermore, intensification is highly correlated with expansion over time for small metropolitans, but they are poorly correlated for large or moderate ones. Lastly, the high correlation of intensification and expansion with the change of GDP in each year indicates the reliability of our work.
BibTeX
@inproceedings{jia2017characterizing,
  author = {Tao Jia and Kai Chen and Jiye Wang},
  title = {{Characterizing the Growth Patterns of 45 Major Metropolitans in Mainland China Using DMSP/OLS Data}},
  booktitle = {Remote},
  year = {2017},
  doi = {10.3390/rs9060571},
}
Figure