以海上丝绸之路沿线的11个超大城市为例，基于长时间序列的Landsat MSS/TM/ETM+/OLI和HJ-1卫星CCD数据，利用基于面向对象的支持向量机SVM（Support Vector Machine）分类方法提取20世纪70年代到2015年的城市不透水层，并结合景观格局指数-最大斑块指数LPI（Largest Patch Index）、斑块密度PD（Patch Density）和欧几里得最邻近距离ENN（Euclidean Nearest Neighbor Distance）分析了超大城市的发展模式。研究结果表明：基于面向对象的SVM分类方法能够高效提取城市不透水层；平均总精度高于87.9%，平均Kappa系数高于0.87；过去40余年，各超大城市的面积扩张了4-13倍，中国和印度的超大城市扩张最快，广州、上海超过12倍；各城市以“中心-边缘”或“沿海-内陆”的方向扩张，表现为“扩散-聚集-再扩散”的扩张模式；总体来看，沿线的城市化进程仍处于上升期。本研究为建设“21世纪海上丝绸之路”提供了科学依据，对当地生态环境保护和新型城镇化建设具有重要意义。

A megacity is a city with more than 10 million inhabitants. Only two megacities existed in the 1950 s, namely, New York and Tokyo. The number of megacities eventually reached 29 in 2014. Experts predict that more than 40 megacities will exist by 2030, and at least 70% of them will come from the Maritime Silk Road. Uncontrollable urban sprawl in the Maritime Silk Road has resulted in serious environmental pollution and ecological damage, which have significantly impacted people's daily lives and health. Building the 21st Century Maritime Silk Road is currently a hot topic. Monitoring and analyzing the spatial dynamics and patterns of megacities along the Maritime Silk Road is critical to local resources and environment protection. This study attempts to monitor and analyze the dynamics of urban expansion in the 11 megacities along the Maritime Silk Road in the period of the 1970 s to the year 2015.
Five long-time series of Landsat MSS/TM/ETM+/OLI and HJ-1 CCD imagery (acquired in the 1970 s, 1990, 2000, 2010, and 2015) were adopted to extract the impervious surface of megacities along the Maritime Silk Road in this study. The images of the said megacities were geo-referenced with registration errors of less than 0.5 pixels in the data preprocessing. All images were resampled to 30 m spatial resolution under the Universal Transverse Mercator coordinates and WGS84 geodetic datum. The object-oriented Support Vector Machine (SVM) classification method was applied to all images after the data preprocessing. The classes involved in the classification maps were bare soil, impervious surface, vegetation, and water body. Approximately 200 points of each class were randomly selected as the validation points. The Overall Accuracy (OA) and Kappa values were calculated by cross-validation utilizing the Google Earth and Landsat MSS/TM/ETM+/OLI imagery. Three landscape metrics, including the largest patch index, patch density, and mean Euclidean nearest neighbor distance, were also applied to analyze and compare the spatial patterns and urbanization of the megacities along the Maritime Silk Road.
Consequently, the average OA and Kappa were above 87.9% and 0.87, respectively. The proposed method could accomplish the spatio-temporal change analysis of urbanization. Moreover, the megacities along the Maritime Silk Road experienced rapid expansion in the period of the 1970 s to the year 2015. These megacities have sprawled at least four times on average with reference to the impervious surfaces of the cities in the 1970 s. In particular, Guangzhou expanded 8 times. These megacities sprawled in a concentric circle or in a "dispersion, aggregation, and re-dispersion" pattern. The urbanization of the megacities along the routes, especially in developing countries such as China and India, exhibits an increasing trend.
This study offers a simple method to extract impervious surfaces on a large regional scale utilizing an object-oriented SVM classifier. The spatial expansion patterns of the megacities in developing countries along the Maritime Silk Road were analyzed utilizing applications such as spatial growth analysis and landscape metrics. The consistent monitoring of megacities in this study provides scientific data for policy makers to assess the potential impacts of urbanization in future urban planning, development activities, and population expansion.