长白山北坡苔原/岳桦景观边界的定量检测

Quantitative Determination of Landscape Boundary between Mountain Birch and Tundra in the Northern Slope of Changbai Mountains

景观边界是不同景观类型之间的过渡带,是生态系统与环境(包括地形和地貌等)相互作用的产物,运用地理信息系统和多变量分析方法(主成分分析PCA和移动窗口分析法MSWA)分别对样线数据和TM遥感影像数据进行分析,定量判定长白山北坡苔原和岳桦景观边界的宽度和位置。结果表明:长白山北坡苔原和岳桦景观边界的宽度为60 m左右,这与野外样线调查的结果——该边界的宽度为50 m相一致;研究结果表明:与地理信息系统和其它统计方法相结合,TM遥感影像数据可以用于森林景观边界的定量检测;在景观边界的检测方面,移动窗口法比主成分分析法更可靠;只要样线布设合理,主成分分析法也可以用于景观边界的定量检测。

Landscape boundaries, or transitional zones among different landscapes, are also called ecotones. They are resulted from complex interactions among ecosystems, topography, and geomorphology. Landscape boundaries are inherent features of landscapes and play important roles in ecosystems dynamics. They control the flux of material between ecosystems and influence biodiversity. Characteristics of ecotones may be especially sensitive to environmental change. Studies have showed that high environmental heterogeneity and biodiversity are often found within these zones, and changes of their locations can be used as indicators of environmental change. However, landscape boundaries have often been ignored or reduced to lines on a map when ecologists studied homogeneous regions to characterize and understand ecosystem processes. Hence, to better study and understand the functional roles and dynamics of ecotones, quantitative methods to identify their lo- cation are needed. There are many methods to characterize and identify ecotones. Such as spatial clustering which considers the spatial relationships among sites, lattice-wombling, triangulation-wombling and categorical-wombling etc. In this study, we use GIS, RS and multivariate statistics techniques (PCA and Moving Split-windows Analysis (abbreviated as MSWA)) to analyze data from field transect inventory and Landsat TM satellite imageries, quantitatively determining the width and position of landscape boundary between tundra and mountain birch in northern slope of the Changbai Mountains. MSWA is a classical analysis method for one dimension value. First putting two windows on the even-interval samples (the number of samples among one of the windows equals to the other), and comparing the dissimilarity of the samples in the two half windows; then, moving the window backward by a sample until all the samples on the transect are used. There are a lot of methods to calculate the dissimilarity. Because the results got by SED (Square Euclidean Distance) was compatible with field observation, it is the most commonly used method. The formula is as follows: where n represents the midpoint of the two half-windows or the stop point of the window, a and b represent the two half-windows respectively, w denotes the width of the window, while m stands for the variable numbers of each sample plot. According to the graph plotted by SED as ordinate and the position of Sample Points along transect as abscissa in Cartesian coordinates, we can determine the condition of landscape boundary by the change of rate. Higher and narrower peaks denote abrupt landscape boundary, while lower and wider peaks denotes gradual landscape boundary. Our results show that the widths of the landscape boundary between mountain birch and tundra is 60 m or so. Such detected widths are consistent with field transect data that suggests a 50 m transitional zone width. Our results further suggest that TM data can be used in combination with GIS and statistical techniques in determining forest landscape boundaries; MSWA is more reliable than PCA, while PCA can also be used to determine the landscape boundary when transects are properly located.