Expansion Analysis of Yangtze River Delta Urban Agglomeration Using DMSP/OLS Nighttime Light Imagery for 1993 to 2012

Huimin Lu, Meiliang Zhang, Weiwei Sun, Weiyue Li
2018 ISPRS International Journal of Geo-Information  
Investigating the characteristics of urban expansion is helpful in managing the relationship between urbanization and the ecological and environmental issues related to sustainable development. procedures are applied to DMSP/OLS data, including intercalibration, intra-annual composition, and inter-annual series correction procedures. Spatial extents are then extracted from the corrected DMSP/OLS data, and a threshold is determined via the spatial comparison method. Finally, three models are
more » ... hree models are used to explore urban expansion characteristics of the YRDUA from expansion rates, expansion spatial patterns, and expansion evaluations. The results show that the urban expansion of the YRDUA occurred at an increasing rate from 1993-2007 and then declined after 2007 with the onset of the global financial crisis. The Suxichang and Ningbo metropolitan circles were seriously affected by the financial crisis, while the Hefei metropolitan circle was not. The urban expansion of the YRDUA moved from the northeast to the southwest over the 20-year period. Urban expansion involved internal infilling over the first 15 years and then evolved into external sprawl and suburbanization after 2007. urbanization processes and for estimating environmental consequences to address such urbanization problems in China. Currently, three main data resources are available for investigating urban expansion problems, i.e., socioeconomic data on population and economy, location-based social network data, and remote sensing imagery [11, 12] . Socioeconomic data directly reflect urban expansion patterns with respect to population and economic changes. Unfortunately, such data lack adequate spatial information and are usually aggregated within coarse administrative divisions. Location-based social network data reflect human activities and can be used to delineate urban boundaries. Unfortunately, such data only cover limited time periods due to their short history, and they thus cannot be used to monitor long-term urban expansion processes. Remote sensing, owing to its unique advantages regarding rapid and periodic data collection and spatial coverage, has served as a powerful technique to characterize and quantify urban extents and dynamic changes [13] [14] [15] . Widely used medium-to-high-remote sensing images, e.g., Landsat TM/ETM+, IKONOS and Quickbird, provide finer land cover information at city and smaller scales. Unfortunately, these remote sensing data sources are impractical to use at larger scales such as the YRDUA region. The above images are negatively affected by the presence of clouds and poor atmospheric conditions, especially in humid subtropical and tropical regions [16] . In addition, the limited spatial coverage of such data results in the generation of numerous images, necessitating considerable levels of image processing and thus involving the application of considerable time and financial resources to obtain an urban map of the YRDUA region [7, 9, 17] . Satellite remote sensing images generated from coarse spatial resolution sensors (>500 m) have accordingly served as major data sources for large-scale urban mapping. However, Moderate Resolution Imaging Spectroradiometer (MODIS) remote sensing imagery have coarse spatial resolutions, and the dataset only became available after 2000. The short-term nature of the dataset, along with its low level of data integrity due to the presence of clouds, impedes the generation of large-scale urban expansion maps of the YRDUA. Instead, the Defense Meteorological Satellite Program/Operational Line-scan System (DMSP/OLS) collects nighttime light (NTL) data at a daily temporal resolution, and such data provide spatially explicit observations of human settlements at night without moonlight [18] [19] [20] [21] . The archived annual NTL data cover a relatively long time span (i.e., 1992-2013) and are freely accessible. A number of efforts have been made to use overall measures of NTL data to estimate demographic and socioeconomic activities at large scales, e.g., for mapping imperious surface areas [22, 23], predicting GDP growth [24, 25] , estimating electricity consumption [26, 27] , evaluating light pollution [28] , and measuring population density and agglomeration [29, 30] . In particular, NTL data have been regarded as cost-effective tools for mapping urban areas [30-32] and urban expansion dynamics [12, 33, 34] . Various researchers have investigated urban expansion at the global, national, and regional scales. At the global scale, Zhang and Seto developed an iterative unsupervised classification based on NTL data for 1992 to 2008 and compared urban dynamic patterns for India, China, Japan, and the United States [35]. Using normalized difference vegetation index data drawn from MODIS and NTL imagery for 1992-2012, Liu investigated spatiotemporal relations between urbanization and vegetation degradation for 50 large metropolises [36] . Cai combined MODIS data and
doi:10.3390/ijgi7020052 fatcat:tenzbsgaejbcbotmjs7jjdl2im