南京市城市景观的特征尺度

The characteristic scales of the urban landscape in the Nanjing metropolitan region

在RS技术支持下,采用分辨率为10 m的SPOT-5遥感影像为数据源,分别应用Haar、Morlet、Mexican Hat 3种基小波,对南京市江北城郊、主城区和东南城郊3种不同景观类型进行了特征尺度研究。结果表明,Morlet基小波最适于进行城市景观的特征尺度检测,M exican Hat基小波对粒级嵌套结构检测效果稍差,而Haar基小波不适于对连续型数据源的景观进行特征尺度检测。Morlet基小波的一维小波分析结果表明,南京市江北城郊和东南城郊都存在1个对应于各自农田斑块平均粒径的特征尺度,分别为362 m和446 m,而主城区则检测出存在3个特征尺度,即:292、835m和2200m,分别对应于建筑小区、小型街区和大型街区的平均粒径,显示主城区存在具有等级结构特征的“建筑小区-小型街区-大型街区”的粒级嵌套结构,揭示了城区具有比城郊复杂得多的粒级结构特征。

The purpose of this scale-related study is to find the underlying mechanisms of ecological phenomena by using appropriate spatial-temporal scales. Scale holds the key to understand pattern-process interactions, but due to limitations of existing theories and methods, in-depth studies on scales in landscape ecology have been limited, especially in urban ecosystems with complex structures. Recently, many new methods were developed and applied in landscape ecology dealing with spatial-temporal processes, e. g. , semivariogram analysis, point pattern analysis, and wavelet analysis. Wavelet analysis can associate a spatial or temporal pattern with different scales and locations and is effective in characteristic scale detection. With one scene of SPOT-5 imagery, one dimensional wavelet analysis was conducted to study the structural features of the urban landscape in the Nanjing metropolitan region. One transect traversed the metropolitan in the main diagonal direction of the city, and three sections named A, B and C were distinguished which corresponded to the northern suburban, urban and southeast suburban area, respectively. To check the validity of main diagonal sampling, three short transects from each of the three areas were randomly placed near the main diagonal line. The results were then compared with the corresponding results along the main diagonal direction, and no remarkable differences were found, suggesting that we could use the results derived from the main diagonal line to represent features of the landscape. Three common mother wavelets, Haar, Morlet and Mexican Hat, were used. The results indicated that the Morlet wavelet was most suitable for detecting characteristic scales of the urban landscape. The Mexican Hat wavelet could be used for the same purpose but it lacks the ability to reveal some details of scales, while Haar wavelet was not capable of detecting characteristic scales with continuous imagery data. The one-dimensional Morlet wavelet transform showed that the landscape of both northern and southeastern suburban areas had one characteristic scale corresponding to the average grain size of cropland patches, 362 m and 446 m, respectively. While for the urban area, multiple characteristic scales （about 292, 835 m and 2200 m） were distinguished, which corresponded to architectural clumps, small blocks and large blocks, respectively. The results revealed that a grain-nesting structure existed in the urban area, in the order of architectural clumps, small blocks, and large blocks. This nesting structure is significantly more complicated than that in suburban areas.