GF-2在城市绿地分类及景观格局度量中的应用

城市植被是维护城市生态系统平衡的重要组成元素,不同类型的植被具有不同的生态服务效益,然而基于植被精细分类的城市绿地景观格局度量研究较少。因此本文以北京城市副中心为研究区,使用夏、冬两季GF-2卫星遥感影像,基于随机森林特征优选和面向对象分类,划分了植被类型,并在此基础上使用景观指数法和移动窗口法分别度量了功能区尺度和栅格尺度绿地景观特征。研究结果表明：针对GF-2数据,使用多尺度分割后影像对象的光谱、纹理特征可以有效地提取植被信息;不同时相的影像能反映各类型植被的物候特征,相比于单时相数据,其分类精度更高,达到了87.7%;各功能区绿地景观格局特征差异较大,城市绿心拥有最丰富的植被且分布集中,商业区植被覆盖度和多样性均较低,绿地景观破碎;北京城市副中心景观多样性及各类型植被分布特征的空间差异显著,当前副中心绿地景观格局已形成规划的基本轮廓,但城市绿心和老城区公共绿地的建设仍有不足。研究探明了北京城市副中心绿地建设现状,证明了国产GF-2卫星在城市生态环境监测中具有较高的应用价值,有助于推动GF-2在城市生态环境领域的应用,并为副中心建设中的绿地系统监测和优化提供参考。

Urban Green Space Classification and Landscape Pattern Measurement based onGF-2 Image

Urban vegetation is an important component of an urban ecosystem, which provides important ecological service including purifying urban air, regulating urban climate, and beautifying urban landscape. Different types of vegetation provide different ecological service. To date, there have been many studies on characterizing urban green space landscape pattern and analyzing its spatiotemporal dynamics. However, the green space landscape patterns of different vegetation types have not been well quantified. In this study, we took Beijing Vice-City Center as the study area and focused on different types of urban vegetation (broad-leaved trees, conifers, farmland, grasslands, and shrublands). Due to the small size of urban vegetation patches and their high spatial heterogeneity, the object-oriented classification method was adopted in this study. Specifically, the image objects using multi-resolution segmentation were generated first. Then, combing the field sampling data, we applied random forest in feature selection and classification using GF-2 images in summer and winter. Based on the classified vegetation map with detailed information, landscape metrics and moving window approach were used to quantify the landscape pattern of urban green space at functional scale and grid scale, respectively. Our results show that, for GF-2 images, the spectral and textural features of image objects after multi-resolution segmentation can effectively improve the extraction of different vegetation types. Multi-temporal images can also provide phenological information of different vegetation types. Compared with the classification results using either summer images or winter images, our classification accuracy was improved to 87.7%. In our study area, the fragmentation of shrublands was the highest. The area proportion of conifers was the smallest, but its patch shape was the most complex. While the distribution of broad-leaved trees, grasslands, and farmland was spatially centralized and contiguous with a regular shape. We found that the green space landscape pattern in different functional areas were quite different. The urban green center had the most abundant vegetation. While the commercial area had less green space that was also fragmentized. The landscape diversity and the distribution of different vegetation types showed a spatial heterogeneity in the Beijing Vice-City Center. At present, the landscape pattern of green space in Beijing Vice-City Center has formed the basic outline of planning, but the construction of urban green center and the public green space in city center are still insufficient. Our study evaluates the current situation of green space patten in Beijing Vice-City Center and proves that GF-2 can be applied in urban ecological environment monitoring, which provides a useful reference for the future monitoring and optimization of urban green space.