近20年京津唐地区不透水面变化的遥感监测

Monitoring of the Impervious Surface with Multi-resource Remote Sensing Images in Beijing-Tianjin-Tangshan Urban Agglomeration in the Past Two Decades

不透水面是衡量城市化程度的重要指标之一,对京津唐城市群的不透水面进行深入研究,可以量化城市群扩张过程及其影响,对该区域多城市协调发展及规划布局具有重要意义。本文结合高分辨遥感影像、生长季及落叶季的Landsat TM遥感影像和夜间灯光数据等,采用分类和回归树(Classification and r Regression Tree,CART)算法,构建了适于京津唐地区不透水面盖度提取的技术方案,获取了京津唐地区1995-2016年共5期地表不透水面盖度专题信息,并分析了地表不透水面的时空演变规律,结论为:(1)适于京津唐地区不透水面盖度提取的CART算法的最佳输入变量组合为:生长季和落叶季的Landsat TM图像以及对应的夜间灯光数据;其次为生长季Landsat TM遥感图像和夜间灯光数据组合方案。利用该组合方案,ISP估算输出结果的交叉验证精度R值可以达到约0.85,可以满足地表不透水面纵向对比分析的需要。(2)从地表不透水面总面积数量值来看,1995-2016年京津唐主体城市区域整体上呈增长趋势,其中2011-2016年地表不透水面积增加愈加明显;(3)从地表不透水面盖度值的高低来看,1995-2016年京津唐中、高盖度不透水面的占比都是在不断增长的,低盖度不透水面占比存在少量下降现象,且京、津、唐3城市的主体城区各阶段变化差异较大,反映出了各城市扩张具有各自不同的时空演变特征。

Impervious surface refers to the surface unable to allow water to percolate through, such as pavements that are covered by impenetrable materials and rooftops. Increased impervious surface area is a consequence of urbanization. Impervious surface percent（ISP） is an indicator to quantify the urbanization level. Therefore,accurate mapping and estimation of ISP in Beijing-Tianjin-Tangshan urban agglomeration are significant for multi-city coordinated development and urban layout. Based on classification and regression tree（CART）algorithm, a technical scheme of extracting ISP which is suitable for Beijing-Tianjin-Tangshan urban agglomeration was constructed in this paper. High-resolution remote sensing data（i.e. Quick Bird images）, medium-resolution remote sensing data（i.e. Landsat TM images in leaf-on and leaf-off seasons）, and nighttime light data were used as basic data in this scheme. Five-year ISP results from 1995 to 2016 were estimated to analyze the spatialtemporal evolution patterns of ISP using this scheme. The main conclusions are as follows：（1） The optimal input variables are the Landsat TM images in leaf-on and leaf-off seasons and the corresponding nighttime light data.Since the number of Landsat TM images in leaf-off season is less in line with the quality requirements, the alternative choice is to use the Landsat TM images in leaf-on season and the corresponding nighttime light data as the input variables. After the accuracy verification, the correlation coefficient（R） is about 0.85, which can meet the need of the comparison of ISP results between different years.（2） During 1995 to 2016, the total impervious surface area increased gradually in Beijing-Tianjin-Tangshan urban agglomeration. Within the period, the most dramatic growth was between the year 2011 and 2016.（3） ISP results were divided into areas with high-, medium-and low-density impervious cover. During 1995 to 2016, the high-density and medium-density impervious cover increased gradually in Beijing-Tianjin-Tangshan urban agglomeration, while the low-density impervious cover decreased slightly. The changes of ISP results in each stage were significantly different among cities of Beijing, Tianjin and Tangshan. It shows that the spatial-temporal evolution patterns are different in the process of urban expansion of each city.