【作者】 张栋；

【导师】 柯长青；

【作者基本信息】 南京大学 ， 地图学与地理信息系统， 2013， 博士

【摘要】 迄今为止,由于科学认知水平与技术条件的限制,人类对高影响冰盖变化的动力机制研究比较薄弱,造成海平面上升的预估结果仍具有很大的不确定性。因此,为了更准确的预测海平面的上升结果,急需强化对极地冰盖由表及里的观测,获取高时空分辨率和高精度的数据,为冰盖动力机制的理论分析和数值模拟等研究提供观测支撑和数据基础。Lambert冰川流域是研究冰盖物质平衡对海平面影响的重点区域之一。但是,目前国际上对其观测仍有许多不足之处。在冰川流域表面,缺乏对其高时空分辨率的物质平衡分布研究,不易为不同区域冰流动力过程提供直观的认识；在其底部,缺少亚公里分辨率尺度的冰下地形特征,而这种尺度的地形对分析不同类型的冰流动力机制具有重要的意义。为此,本文采用测高数据提取Lambert冰川流域表面高时空分辨率的高程变化信息,识别冰流变化的区域性特征。对其亚公里分辨率尺度的冰下地形特征提取主要集中在两个区域。一个是Lambert冰川流域源头的Dome A区域,另一个是冰川流域东侧的中山站-Dome A断面。主要研究内容和结论包括：(1)为提取表面高程变化特征,在对ICESat相邻工作期中重叠点对因坡度引起的高程差进行消除的基础上,计算了重叠点对的高程变化。然后采用反距离加权、自然邻域、径向基函数和ANUDEM (Australian National University DEM)等方法对每个阶段内高程变化进行插值,并选用插值精度最高的结果作为整个表面的高程变化。研究结果表明,本文的高程差消除方法效果更好,ANUDEM插值结果的精度最高。在时间上,Lambert冰川流域高程平均每年增长约0.002cm,且增长的加速度在减缓。因此,其物质处于相对平衡的状态。除接地线附近存在高程减小超过2m/a的区域外,多数区域的高程变化量都小于1m/a。在空间上,高程变化具有显著的区域性特征。纬度方向上,下游地区高程呈减小的趋势且减小的加速度在升高；上游高程增长0.84cm/a,但增长的加速度在减缓。经度方向上,三大支流高程均处于增长的趋势,其中Fisher支流增长最快,平均增长1.09cm/a。但三大支流高程增长的加速度也在变缓,以Mellor支流变缓的程度最大。通过与流速数据结合分析可知,高程减小超过1m/a的区域主要分布于Lambert和Mellor支流,二者流速大于100m/a的区域均深入内陆约200km,而Fisher支流的相应区域仅深入内陆75km,表明其所受冰流作用小于其它两个支流。(2)为了提取Dome A区域的冰下地形特征,首先建立了一种半自动方法从冰雷达数据中获取了各断面上覆冰层的厚度。然后采用多种方法对Dome A区域的冰厚和表面高程进行插值实验,从中确定出精度最好的结果(ANUDEM获得)作为冰厚和表面高程模型。最后将冰厚与表面高程结合,生成Dome A底部冰下地形数据。研究结果表明,半自动法所得交叉点厚度绝对差值在50m之内的点占76%,100m之内的点占92%,而人工数字化方法的对应值则分别为61%和89%。且半自动法所得结果与国际上AGAP(Antarctica’s Gamburtsev Province)数据的一致性也明显优于人工数字化方法,表明半自动法所得结果的一致性和精度比之前方法均有所提高。通过采用AGAP数据将所得冰厚模型与人工数字化并插值后的结果进行对比,可知本文结果的厚度差均值仅为3.5m,远优于人工数字化方法的结果(17.7m)。因此,新的Dome A区域冰下三维地形精度最佳,对区域冰盖模式的模拟、冰流动力过程理解和冰盖演化过程分析均具有重要意义。基于此方法,创建了在Dome A区域运行Elmer/Ice模式所需的数据集,为Dome A区域底部温度场分布和冰层深度—年代关系研究提供数据支撑。(3)为提取中山站-Dome A断面的冰下地形特征,采用快速傅立叶变换计算了滑动窗口内冰岩界面的粗糙度双参数,用于定量描述冰岩界面高程的异常变化。在提取的过程中,除了获取冰岩界面的总粗糙度外,由于可以采用快速傅里叶变换将冰岩界面分解为一系列不同波长尺度的周期波,所以还将构成冰岩界面的波分为长波(大于1680m),中波(840-1680m)和短波(420-840m)三个类别,分析了不同波长尺度下的粗糙度分布。在本文采用的窗口尺度下,断面的冰岩界面主要有两种类型：ζ和η值均小对应高程起伏幅度小而水平变化频率快的冰岩界面类型；ζ和η值均大对应高程起伏幅度大而水平变化频率低的冰岩界面类型。结果表明,Lambert冰川支流南侧断面的冰岩界面起伏幅度普遍大于北侧,而且与北侧同为高大山峰的地形相比,其山峰的坡度也较大。通过不同波长尺度下的粗糙度分析,发现整个断面冰岩界面以长波尺度的地形为主要形式,但也存在一些特殊地形区域。例如,Lambert支流北侧断面两端高程起伏幅度不超过200m的区域,经分析发现以中短波的地形为主要特征。而南侧部分区域的冰岩界面则以中波尺度的地形特征为主要形式,短波尺度的地形则被侵蚀了。更多还原

【Abstract】 So far, because of the limitations of scientific cognition levels and technical conditions, the researches of the high-impact factors on dynamic mechanisms of ice sheet change are so weak that sea level rise projections still have great uncertainties. Therefore, in order to predict the result of sea level rise more accurately, the urgent requirement is to strengthen the observations of ice sheet from surface to interior, and to obtain high spatial and temporal resolutions and high precision data to provide observation supports and data base for the theoretical analysis and numerical simulation studies for ice sheet dynamic mechanisms.Lambert Glacier drainage basin is one of the key areas for the study of ice sheet mass balance effect on sea level. However, the current international observations of this basin still have many inadequacies. On the surface of the glacier basin, high spatial and temporal resolution distributions of mass changes are lacking, which cannot support intuitive understanding for the different regions of ice flow dynamics; at its bottom, the subglacial topography with sub-km resolution scale, which is great significance for the analysis of the different types of ice flow dynamic mechanisms, is lacking.Therefore, this paper extracts elevation changes with high spatial and temporal resolution on the surface of Lambert Glacier drainage basin using altimetry data to identify the regional characteristics of ice flow change. The feature extraction of subglacial topography with sub-km resolution scale mainly concentrates in two areas. One is the Dome A region which is the source of Lambert Glacier drainage basin, and the other is Zhongshan Station-Dome A transect in eastern basin. The main contents and conclusions include: (1) To extract the surface elevation change, based on the elimination of elevation differences of overlapping footprint pairs caused by the slope from adjacent operation periods of ICESat, we calculate the elevation change of overlapping footprint pairs. Then we use the inverse distance weighting, natural neighbor, radial basis functions and ANUDEM (Australian National University DEM) to interpolate the elevation change of each stage and chose the most accurate interpolation results as the elevation changes of the entire surface.The result shows that our method of eliminating the elevation difference is better, and the interpolation accuracy of ANUDEM is the highest. In the time scale, the elevation of Lambert Glacier drainage basin grows about0.002cm/a, and the acceleration of growth is slowing. Therefore, the mass of the basin is in balance. Except near the grounding lines with regions of elevation reduction exceeding2m/a, the elevation changes in most regions are less than1m/a. In the spatial scale, the elevation change has significant regional distribution characteristics. In the latitude direction, the elevation of downstream areas shows a decreasing trend and the reduced acceleration is increasing; the elevation of upstream areas increases0.84cm/a, but the acceleration of the growth is slowing. In the longitude direction, the elevations of the three major tributaries are in growth trends, and the Fisher tributary has the fastest growing, about1.09cm/a. However, the accelerations of growth in the three major tributaries elevation are also slowing, and Mellor tributary has the maximum extent of slowing. Combined with the ice flow data, the regions with elevation reduction exceeding1m/a are mainly in Lambert and Mellor tributaries, and the regions with velocity greater than100m/a of the two tributaries are affecting further200km of the inland region, while the corresponding region of Fisher tributary is only75km. This indicates the role of ice stream in Fisher tributary is less than the other two tributaries.(2) In order to extract the subglacial topography of Dome A region, we firstly create a semi-automatic method to extract the ice thickness of the transect from ice radar data. Then we interpolate the ice thickness and surface elevation of Dome A using a variety of methods, and choose the best accuracy results (ANUDEM obtained) as the ice thickness and surface elevation model. Finally, we combine the ice thickness with the surface elevation to produce the subglacial topography of Dome A region.The result shows the cross points with absolute thickness difference less than50m derived by semi-automatic method account for76%and less than100m account for92%, while the corresponding points derived by artificial digital method are61%and89%. And compared with the international AGAP (Antarctica’s Gamburtsev Province) data, the consistency of semi-automatic method is also better than artificial digital method. This indicates the consistency and accuracy of semi-automatic method have improved than previous methods. Using AGAP data to compare the ice thickness model derived by us with the one of previous method, we found the mean thickness difference of our result is only3.5m, which is far superior to artificial digital method (17.7m). Therefore, the accuracy of the new three-dimensional subglacial topography of Dome A is the best, which will play an important role in the regional ice sheet model simulation, the understanding of ice flow dynamics and the evolution of the ice sheet. Based on this method, we create datasets for Elmer/Ice model running in Dome A for the research on the basal temperature of ice and estimate the relationship between the depth and age of the ice layer.(3) For the extraction of the subglacial topography of the Zhongshan Station-Dome A transect, we calculate two-parameter roughness index in the sliding window using the Fast Fourier Transform to quantitatively describe the abnormal elevation change of the ice-bedrock interface. In the extraction process, because the Fast Fourier Transform can transform ice-rock interface into a sum of several various wavelength periodically corrugated surfaces, besides obtaining the total ice-rock interface roughness, we also divide the ice-rock interface into three scales:long wavelength (greater than1680m), medium wavelength (840-1680m) and short wavelength (420-840m), and the analyzed the distribution of roughness with different wavelengths scales.The ice-rock interface includes two main types in the window scale which this paper used:both of the ζ and η values are small relating to the ice-rock interface with small elevation fluctuation and fast frequency of horizontal change; both of the ζ and η values are large relating to the ice-rock interface with high elevation fluctuation and slow frequency of horizontal change. The result indicates that the vertical fluctuation of ice-rock interface in the south side transect of Lambert tributary is generally greater than the north side, and compared with the high peaks in the north side, the slopes of peaks in the south side are also high. By analyzing the roughness among the different wavelength scales, we find long wavelength scale terrain is the main characteristics for the ice-rock interface of the entire transect, but there are also some special terrain areas. For example, the analysis finds that short wavelength terrain is the main feature in the areas with elevation fluctuation less than200m in the two ends of the transect in the north side of Lambert tributary. The medium wavelength is the main feature in some parts of the transect in the south side of Lambert tributary, and the short wavelength topography has been eroded.更多还原