Abstract:
Abstract: Based on the spatial distribution model of NDVI (normalized difference vegetation index) value which is extracted from remote sensing image, spatial distribution features of the watershed surface NDVI value can be understood to be composed of a large number of equal units, and the side length of the units is equal to the size of the remote sensing image pixels. NDVI value is stored in each unit of attribute sheet "VALUE". In this paper, we realize the measurement of NDVI increment at every point on the watershed by the development of GIS algorithm and the establishment of the moving window in the research process. The moving window statistical method divides the whole watershed into a number of equal DVM (digital vegetation model) cells (r×r, r is an odd multiple of the pixel size, r>1). Each cell is called a "window" and the cells don't have overlap for each other. We computerize the difference of NDVI values between each pixel point and the center pixel point in each cell. We calculate NDVI increment value of the watershed points at a certain spatial scale by the moving window statistical method, then computerize mathematics expectations of measure collection, which is composed of all NDVI incremental values. To study the characteristics of fractal dimension of the vegetation cover and its variation at different scales, it is divided into 4 scales according to the area of the basin, with 5 phases per level. The first level is the entire Dali River basin with an area of 3 906 km2, approximately 5 million times of pixel size (30 m × 30 m). At the second stage, Dali River basin will be divided into 3 parts i.e. upstream, midstream and downstream, with an average area of approximately 1 000 km2, approximately 1 million times of pixel size (30 m × 30 m). At the third stage, it will be divided into 14 small basins of Dali River, with an area of 179.4-392.5 km2, about 200 thousand times of pixel size (30 m × 30 m). At the fourth stage, it will be divided into 53 small basins of Dali River (No.1 to 53), with an area of 21.9-108.9 km2, about 40 thousand times of pixel size (30 m × 30 m). Firstly, on the basis of information source from 5 issues of TM/ETM images from 1990 to 2006, the image data are processed, the vegetation information at basin stage is extracted by the platform of geographic information system (GIS), and then watershed DVM is established. FBM (fractional brownian motion) fractal dimension for watershed vegetation cover is between 2.7311 and 2.8499. By calculation and analysis, vegetation cover FBM fractal dimension is increasing from upstream to downstream, so vegetation distribution is more uniform from upstream to downstream. Vegetation cover fractal dimension is increasing with the watershed area. Through analyzing spatial and temporal variation of each sub-watershed at all levels, it presents the curvilinear trend of decreasing firstly and then increasing as the change of time from 1990 to 2006; at the same time, the vegetation coverage fractal dimension of each sub-watershed at all levels is no more than its bigger sub-watershed.