【作者】 孔凡哲；

【导师】 芮孝芳；

【作者基本信息】 河海大学 ， 水文学及水资源， 2003， 博士

【摘要】 数字高程模型(DEM)已在分布式流域水文模型中得到了越来越广泛的应用，但DEM中的闭合洼地和平坦区域影响着流域河网的自动提取。目前已提出好多方法来处理这两种地形，但均针对已经形成的DEM单元网格进行处理，结果往往生成伪河道及平行河道。DEM中的闭合洼地和平坦区域是由于低质量的资料输入、生成DEM时的内插误差等引起的。通过增加输入地形高程信息，可避免DEM中平坦区域和闭合洼地的生成，使模拟的河网与实际河网能够精确拟合。实例分析表明，该方法效果明显。O’callaghan和Mark方法是广泛应用的提取流域河网的方法。为了得到连续的河网，方法中引入了汇水面积阈值的概念，但对于同一流域，利用不同的汇水面积阈值将得到不同的河网。提出了一种方法，利用河源密度(或河网密度)与汇水面积阈值的关系，确定汇水面积阈值，当河源密度(或河网密度)趋于稳定时对应合理的面积阈值。将该方法应用于沿渡河流域，得到了理想的结果。地貌单位线被看作是流域上各水质点在弱相互作用下，到达流域出口汇流时间的频率分布。对于一个典型的山坡型网格单元，汇流路径由两部分组成，即坡地部分和河道部分，为了得到汇流时间，必须首先确定汇流速度。坡地和河道的汇流速度随着区域位置而变，并且必然与坡度有关，因此，可首先计算流速的空间分布，进而得到汇流时间的空间分布。现行地貌单位线计算方法忽略了与坡地漫流有关的滞留时间。提出了一种汇流时间计算方法，汇流时间中包括坡地漫流时间和河道汇流时间。方法中坡地单元的汇流速度与河道单元的汇流速度采用不同的计算公式，同时考虑流速沿河道向下游的变化。流域中每一个网格单元的汇流时间得到后，将其看作随机变量，进行统计分析后，得到汇流时间的频率分布—GIUH。地形指数ln(α／tan β)是一些以物理概念为基础的水文模型的重要参数。TOPMODEL是以计算ln(α／tan β)指数及其分布为基础的。对于栅格DEM，α为上坡区域通过单位等高线长汇集到单元网格内的面积，反映径流在流域中任一点的累积趋势，tan β为单元网格的坡度，反映重力使径流顺坡移动的趋势。目前普遍使用的计算该地形指数的方法为多流向法。方法中计算α和tan β用的均是与流出单元网格流向垂直的等高线长。另外计算下坡单元网格累积汇流面积时没有考虑欲计算ln(α／tan β)值的单元网格的面积，这些是不合理的。计算α值应该用与流入单元网格流向垂直的等高线长，据此提出了改进后的ln(α／tan β)的计算公式。方法摘要中计算下坡单元累积汇流面积时包括了欲计算hi(a八anp)的单元网格的面积。分析了两种方法计算结果间的差值。 多流向地形指数计算方法在实际应用时比较复杂。提出一种简单的地形指数in(a/t anp)计算方法，该方法分别计算DEM网格单元的a和tanp，对坡度为。的网格单元进行专门处理，然后计算每一个网格单元的ln(a/tanp)。以长江上游沿渡河流域为例，将计算的地形指数用于TOPMODEL进行洪水径流模拟计算，得到了比较满意的结果。 流域地形控制着地表径流的路径和累积水量的空间分布，流域地形可以用地形指数的空间分布来表示。地形指数可用来作为水文相似性指数，即具有相同地形指数的各点对降雨具有相同的水文响应。TOPMODEL利用地形指数频率分布计算流域径流过程。因此可以认为:具有相同地形指数频率分布的流域具有水文相似性。分析认为长江流域上游的沿渡河流域与兴山流域具有水文相似性，利用兴山流域率定的模型参数和地形指数频率分布计算的沿渡河流域径流过程，与实测过程拟合较好。更多还原

【Abstract】 Digital elevation model(DEM) has become widely used in distributed hydrological models. It is found that the automatic detection of drainage networks from digital elevation models is affected by flat areas and closed depressions .A variety of methods have been proposed to treat these two features , but it can be seen that the flat areas and closed depressions are processed only after they have been created ,so that many spurious and parallel drainage courses are generated . It is known that flat areas and closed depressions often arise because of low quality input data, interpolation errors during DEM generation et al., therefore, they can be removed by adding elevation information to input topographic data. The proposed approach can lead to better match between observed and modeled flow structure and produce more realistic results in application.The algorithm for delineation of a drainage network proposed by O’callaghan and Mark is very widely used. A drainage area threshold is used to produce a continuous network, so that different network is obtained for varying drainage area threshold. A method is proposed, in which the relation between source density (or drainage density) and the threshold is used to produce an ideal drainage network ,the area threshold when source density (or drainage density) show stability is reasonable. The method is used in Yanduhe watershed and a good result is obtained.The geomorphological instantaneous unit hydrograph(GIUH) is viewed as the frequency distribution of the times of arrival of individual water deoplets at the catchment outlet.The travel path ,for a typical hillslope cell .consists of a hillslope fraction,corresponding to overland flow and a stream fraction,corresponding to concentrated channeled flow.To obtain the time of travel,velocities must be defined.Hillslope and stream velocities vary with location and must be strongly correlated with slope,and therefore a spatial distribution of velocities and hence of travel times could be obtained.The present methods of GIUH neglect any time delays associated with overland flow pathways.A new method is presented.The travel time,including the time delays associated with overland folw pathways, is obtained.It is expected that the hillslope velocity and the stream velocity are different,and different equation is used.In the method ,the fact that velocity increases going downstream in river systems is taken into account.Afterthe travel time of each cell being calculated,the frequency distribution of the times of arrival of individual water droplets at the catchment outlet -GIUH,is obtained.The topographic index ln(α/tan #) is an important parameter of many physically based hydrological models. TOPMODEL is based on the calculation of ln(α/tan #) index and its distribution .In terms of a DEM, a is the cumulative upslope area draining through per contour length to a pixel, which reflects the tendency of water to accumulate at any point in the catchment, tan # is the local slope angle of the cell, which reflects the tendency for gravitational forces to move that water downslide. The calculation method of ln(α/tan #) index widely used is the multiple flow direction algorithm developed by Quinn et al.It can been seen that, in the algorithm, the contour length normal to the direction of flow flowing out the current cell is used to determine both a and tan # , and that the calculated total cumulative contributing area of downslide grid cell does not include the area of the current cell. These are unreasonable. It is found that the contour length used to determine a should be that normal to the direction of flow entering into the current cell, rather than that flow out the current cell. The algorithm improved to calculate the ln(α/tan #) index is presented, in which the area of the current cell is included in the calculated total cumulative contributing area of downslide grid cell. The different between the results of the two algorithms is described.A simple algorithm for the calculation of topographic index ln(更多还原