Spatial-Temporal Evolution and Regional Differentiation Features of Urbanization in China from 2003 to 2013

Peiyu Zhang, Jianjun Pan, Longtao Xie, Tao Zhou, Haoran Bai, Yanxiang ZhuIsprs Int. J. geo inf 2019 · DOI

TL;DR

This paper quantifies the spatial and temporal evolution of urban impervious surfaces (IS) in China from 2003 to 2013. By comparing three IS extraction methods and selecting the most accurate one (EANTLI), the study reveals significant regional disparities in urbanization and provides crucial data for understanding China's rapid urban development and its environmental implications.

What Problem Does This Paper Address?

Urbanization is a complex phenomenon with significant environmental and ecological impacts. Quantifying impervious surfaces (IS) is crucial for understanding urban development, managing resources, and planning policies. However, accurately mapping IS across large scales and over time, especially in a rapidly developing country like China, presents challenges. Existing methods often struggle with data saturation, mixed pixels, and regional variations in land cover. This paper aims to address these issues by developing and applying a robust method for extracting urban IS across China and analyzing its spatial-temporal patterns and regional differentiation.

Methods

The study employed a multi-step approach:

  1. IS Extraction Index Development and Comparison: Three established indices for extracting urban IS were utilized:
  2. Human Settlements Index (HSI): Relies on Normalized Difference Vegetation Index (NDVI) and nighttime light (NTL) data.
  3. Vegetation-Adjusted NTL Urban Index (VANUI): Adjusts NTL data with vegetation information.
  4. EVI-adjusted NTL index (EANTLI): Combines NTL data with the Enhanced Vegetation Index (EVI), specifically using the maximum annual EVI to enhance differentiation between urban and non-urban areas, and to mitigate NTL saturation issues.
  5. Benchmark Data Acquisition: High-resolution Landsat-8 imagery was used to create a benchmark dataset for urban IS in eight representative Chinese cities. This provided a ground truth for evaluating the accuracy of the three indices.
  6. Accuracy Verification: The extracted IS from the three indices were visually compared and quantitatively verified against the Landsat-8 benchmark data using overall accuracy and Kappa coefficient metrics.
  7. Selection of Optimal Index: Based on the accuracy assessment, the EANTLI index was identified as the most suitable for extracting urban IS in China due to its superior performance in distinguishing between urban and non-urban areas, especially in regions with varying vegetation cover and mitigating NTL saturation.
  8. Spatio-Temporal Analysis: The selected EANTLI index was then used to extract urban IS for China from 2003 to 2013. Quantitative indicators such as Urban IS Percentage (UISP), Year-On-Year Growth rate (YOYG), and Urban IS Expansion index (UISEr) were calculated to analyze the temporal and spatial evolution and regional differentiation of urban IS.

Datasets

  • DMSP-OLS NTL (Nighttime Light) data (Version 4): Used for its broad spatial coverage and ability to capture human settlement intensity. Data from 2003, 2008, and 2013 were utilized.
  • MODIS NDVI and EVI data (MOD13A2): Provided vegetation information to adjust NTL data and differentiate between vegetated and impervious surfaces. Annual data from 2003, 2008, and 2013 were used.
  • Landsat-8 OLI multispectral data: High-resolution imagery used to create benchmark urban IS datasets for accuracy verification. Data from 2013 for selected cities were used.
  • Google Earth images: Used for visual verification of Landsat-8 imagery in specific cities.
  • MOD44W (Global land water mask): Used to eliminate the influence of water bodies on classification results.
  • Provincial administrative division map of China: Used for regional analysis.

Key Insights and Ideas

  • EANTLI's Superiority: The EANTLI index, by incorporating the maximum annual EVI, proved to be significantly more accurate than HSI and VANUI for mapping urban IS across China. This is attributed to its enhanced ability to differentiate between urban and non-urban areas, especially in regions with low vegetation cover, and its effectiveness in reducing NTL saturation.
  • Regional Disparities in Urbanization: Urbanization in China is not uniform. The study highlights a clear spatial differentiation, with Eastern Coastal China exhibiting the highest urban IS percentage, while Northwest China shows the lowest. This pattern strongly correlates with regional economic development levels.
  • Accelerated Urban Growth: China's urban impervious area has more than doubled between 2003 and 2013, with a notably faster growth rate observed in the period 2008-2013 compared to 2003-2008. This rapid expansion is linked to China's economic development (GDP growth).
  • Specific Growth Hotspots: Cities and provinces like Beijing, Tianjin, Jiangsu, and Shanghai experienced the most significant growth in impervious areas, suggesting a concentrated impact on their respective ecological environments.
  • Temporal Trends in Growth Rate: While overall urbanization is increasing, the year-on-year growth rate of urban impervious areas showed a slight decrease from 2008-2013 compared to 2003-2008, potentially indicating a maturing stage of development in some areas.

Key Results

  • Total Urban IS Area: In 2013, China's urban IS area was 70,179.06 km², representing 0.73% of the country's land area. This is a substantial increase from 20,565.24 km² (0.21%) in 2003.
  • Accuracy of EANTLI: The EANTLI index achieved an overall accuracy of 95.41% and a Kappa coefficient of 0.91, outperforming VANUI (92.13% accuracy, 0.87 Kappa) and HIS (88.75% accuracy, 0.78 Kappa).
  • Regional UISP: Eastern Coastal China (ECC) had a UISP of 7.26% in 2013, while Northwest China (NWC) had only 0.11%.
  • YOYG: The expansion of impervious areas in the Middle Reaches of the Yangtze River (MRYTR) and Middle Reaches of the Yellow River (MRYLR) regions showed a significant four-fold increase over the decade (2003-2013).
  • UISEr: Beijing, Tianjin, Jiangsu, and Shanghai showed the greatest growth in impervious areas from 2003 to 2013.

Limitations and Future Work

  • Limitations:
    • The study acknowledges that while EANTLI is superior, it still has limitations, particularly in distinguishing between urban and bare land in areas with very low vegetation cover.
  • The use of annual average EVI in some previous studies (not this one, but mentioned for context) can weaken vegetation abundance and stabilize change, which the authors addressed by using maximum annual EVI.
  • The DMSP-OLS NTL data itself has inherent limitations, such as saturation effects, although EANTLI mitigates this.
  • Future Work:
    • The paper suggests that understanding the rapid increase in impervious surface areas in certain regions is crucial for addressing potential impacts on the ecological environment and informing policy development.
  • The UISEr index is presented as a valuable tool for future comparative studies of urbanization differences between cities.
  • Further research could explore more advanced methods to refine IS extraction, especially in challenging environments.

Summary updated

BibTeX
@article{zhang2019spatial,
  author = {Peiyu Zhang and Jianjun Pan and Longtao Xie and Tao Zhou and Haoran Bai and Yanxiang Zhu},
  title = {{Spatial-Temporal Evolution and Regional Differentiation Features of Urbanization in China from 2003 to 2013}},
  journal = {Isprs Int. J. geo inf},
  year = {2019},
  doi = {10.3390/ijgi8010031},
}
Figure