| The soil erosion, which happens most widely, is one of the most serious ecological environment problems in China currently. It not only exacerbates the ecological environment of the local region, but also its hazardousness often spread to a greater scope. The Danjiangkou reservoir is located at the juncture of Hubei, Henan, Shaanxi three provinces. It has aroused widespread interest from all walks of life, because at present it undertakes the middle route project of south-to-north water transfer which supplies water to the north China including capital Beijing.Along with the implementation of the project, the ecological environment question in the reservoir area especially soil erosion has become an important factor which affects the water quality of reservoir as well as the success or failure of the water transfer project. Therefore it is very necessary to do research on the present situation of soil erosion and its dynamic change in the water source area of Danjiangkou reservoir. Comparing with others kind of methods to investigate and appraise soil erosion, remote sensing has superiorities of multi-type, multi-platform, multi-period, multi-band characteristics, and rich information, short information cycle, strong presence and macroscopic dynamic and so on, undoubtedly, it becomes one of the fastest, effective advanced methods:Based on summarizing the research advance of soil erosion, remote sensing monitoring for soil erosion and its method, this article took Danjiangkou county which is the water source area of water transportation project from south to north, as the study object, utilized the technology of RS, GIS and GPS, as well as the theory of water and soil conservation, and did more detailed research on information extraction from remote sensing data with difference phases and different data sources, it chose wavelet transform fusion image and SVM(Support vector machine) classification method. According to the standard for classification and gradation of soil erosion, Soil erosion intensity map was obtained by using the method of expert classfier analysis based on the classfier function of Erdas. The article comparatively analyzed the three period of land use and vegetation coverage and soil erosion, it had also been done to find the rule of change in 1991, 2000, 2005. The main results were outlined as following:First, three period of satellite images, TM in 1991, ETM+ in 2000, IRS-P6 in 2005 has been selected. Then geometrical rectification, digital mosaic as well as statistical property analysis and combination choice between wave bands have been carried on to these images with different phases and different data sources. Based on the study of image fusion technology, using the combination method of biorthogonal wavelet transform and IHS transform, the result has been comparatively analyzed with db4 wavelet transform, IHS transform method. The study proved that using the fusion method referred in this article, the image after ETM+ fusion not only retained its rich spectrum information of multi-spectrum image, but also improved spatial resolution at varying degree, improved the effect of visual interpretation and automatic classification to a great extent.Second, based on supervised classification to three images, the support vector machines (SVM) classification method that surpasses the traditional supervised classifier has been selected. As a result, the classified precision has been improved. Take IRS-P6 image as an example, the classification kappa coefficient, and overall classification precision of SVM is 0.797, 82.37% respectively. But the kappa coefficient and overall classification precision of the most commonly used maximum likelihood method is 0.759, 78.43% respectively. The classification precision of ETM+ image has improved after fusion compared with before.Third, to extract the vegetation cover factor, TM and ETM+ images have used the revised NDâ…¥model, while IRS-P6 image has used the general NDâ…¥model. Then vegetation coverage was extracted through classification. All three images employed the correlation method after classification, and comparative analysis has been carried out to the land use change in GIS. From 1991 to 2000 and from 2000 to 2005, both forest land and habitation show positive growth tendency in two time intervals, while unused land and scrub forest show decreasing tendency at all. Dry land and pasture in 2005 both reduced at a certain degree compared to 1991. While water land is growth and the change of paddy field is not significant.Fourth, the result of research indicated that, from 1991 to 2005, the erosion area presented a decreasing trend overall, and the erosion condition has improved gradually. The erosion area was 1736.29 km~2 in 1991, which accounted for 55.63% of the total acreage; in 2000 was 1408.96 km~2, accounted for 45.14% of the total acreage; in 2005 was 1258.37 km~2, accounted for 40.32% of the total acreage. For erosion intensity graduation, the obvious soil erosion of all the three images is mainly moderate erosion and mild erosion. In 1991, moderate erosion and mild erosion accounted for 21.84% and 21.04% of the total acreage; In 2000 was 17.98%, 17.07% respectively; In 2005 was 15.79%, 15.56% respectively.
|
PDF Full Text Request