Chronological Analysis of Wind Erosion of Kitakami Highland using Satellite Images

本研究利用卫星影像对岩手县北上高地风蚀区进行了识别，并对其与环境因子的关系进行了分析。在定量分析方面，对卫星图像进行了几何校正和光照校正。针对基于模拟地貌图像的几何校正优化方法不适用于大范围的问题，提出了一种改进算法，将场景中心位置的移动建模为线性回归函数。通过使用新算法，整个北上高地被卫星图像无缝覆盖。在光照校正方面，应用了改进的余弦法，得到了阴影起伏较小的图像。由于研究区域的地表高度较高，我们没有进行任何特殊的大气层改正。改正后的卫星图像使用…更多地用于土地覆盖分类：首先用金字塔链接法对图像进行空间分割，然后用K-均值方法对分割的图像进行聚类，最后将其分为五类：(1)侵蚀土壤，(2)棕壤，(3)草地，(4)森林年龄*分析其与下列环境因子的关系：(A)地表高度，(B)坡度，(C)坡向角，(D)向西开阔，(E)西部森林存在，(D)春季的太阳辐射。分析了各因素对侵蚀速率的影响，并将各因素纳入广义线性模型。分析认为，森林的存在是阻碍冬季西风的最主要因素，确定侵蚀区存在一定的困难。例如，超过35度的陡坡和海雅坦山的岩石顶峰从卫星图像上看起来像是侵蚀区。这些地区被小心翼翼地从侵蚀类别中删除。然而，利用卫星图像识别侵蚀区域是非常有效的，我们发现了一些到目前为止还没有识别的侵蚀区域。较少

In this research, wind erosion areas in Kitakami highland of Iwate Prefecture were identified by using satellite images, and its relationship with environmental factors was analyzed. In addition, the effect of counter measures for wind erosion prevention were cholonologically analysed.For quantitative analysis, the satellite images were carefully processed with geometric correction and illumination correction. As the optimization method for geometric correction using simulated relief image cannot be applied to wide area, we proposed a modified algorithm, which model the shift of scene center position as a linear regression function. By using the new algorithm, the whole Kitakami highland was covered by the satellite images seamlessly. For illumination correction, we applied the modified cosine method and obtained the images with less shading relief. We did not apply any special atomospheric correction because of high surface altitude of the study region.The corrected satellite images w … More ere used for land cover classification : first the images were spatial segmented by using pyramid linking method, then the segments were clustered using K-Mean method, finally they were classified into five categories : (1)eroded soil, (2)bair soil, (3)grass, (4)forest *****ages were used for analyzing its relationship with following environmental factors : (a)surface elevation, (b)slope angle, (c)aspect angle, (d)openness to the west, (e)forest existence on the west, (d)solar radiation in spring. Each factor was analysied for its effect on the erosion rate, and all factors were incorporated into the generalized linear model. The analysis revealed that the forest existence was the most important factor, which prevent the west wind in winter.There was some difficulties to identify the erosion area. For example, acute slope of more than 35 degree and the rocked summit of Mt.Hayatine looked like erosion area from the satellite image. These areas were carefully eliminated from the erosion category. Utilization of satellite images, however, was proved to be very effective to identify the erosion area over a wide region, and we found some erosion area which had not been identified so far. Less