影响坡面径流法河网提取精度的关键问题研究

The DEM-based slope runoff approach is the most commonly used drainage line extraction method, in which the determination of the flow directions of depressions and flats and the calculation of the contributing area matrix are the two key techniques that effect the extraction accuracy. These issues are hot spots in this field and extensively studied in the past 20 years...As for the first issue, most of the present work focus on the development of various algorithms that incorpate the topographical informaiton derived from either DRLN (Digital River and Lake Network) or remotely sensed feature extraction data with DEM to designate flow directions for depressions and flats. However, few studies exploit the full potential of both data sources by combining them together with DEM. The latest research also found that this method is prone to commit ommison and commission errors in river network extraction, which arises from the asssumption that the surface of the watershed is considered homogeneous and each grid cell can generate the same amount of flow in computing the flow contributing area. Thus, the aim of this proposal is at improving the mentioned two key techniques...In this proposal, a flow direction designation algorithm for depressions and flats are developed which is capable of exploring the topographical information from DRLN and remote sesensing images. Thus, reasonable flow directions could be determined based on the combined multi-source information. Subsequently, rainfall-runoff plot experiments with different underlaying surfaces of the watershed are designed to study their contributions to flow generation amounts. A BP neural network is then employed to extend the results from the plot scale to the watershed scale to estimate flow generation capacity for each grid cell of the watershed. These measurements are normalized into 0 and 1 to form a weight matrix that has the same size with DEM. The flow accumulation matrix is recalculated with a weight matrix that take into account the variations of properties of watershed characteristics. The drainage lines are extracted based on the weighted flow accumulation matrix and the accuracy is assessed...The proposed approach incorporates the physical mechanism of flow generation into the data-oriented drainage line extraction, thus, increases the accuracy of drainage line extraction, and extends the spatial analysis in geoscience fields.

坡面径流法是最为常用的河网提取方法，其中DEM中洼地及平地流向的确定与流量累积矩阵的计算是影响提取精度的关键问题，也是国内外研究的热点与难点问题。本研究针对1)现有洼地和平地流向确定算法中对辅助数据利用不够充分，2)在计算流量累积矩阵时因假设流域地表均质而造成的水系错提与漏提等问题，拟在整合DRLN（数字河流湖泊网络）与遥感特征提取信息等数据优点的基础上，充分挖掘辅助数据中的有效地形信息，设计算法来确定DEM中洼地和平地的合理流向。在此基础上，结合野外降雨-径流小区试验及流域实测水文数据，研究土壤、植被、土地利用及地质条件等流域特征对产流量的贡献，进而构建BP神经网络，来反演非均质流域下垫面栅格的相对产流系数，用于计算相应的实际流量累积矩阵。本研究将传统基于数据的河网提取方法与坡面径流形成的物理机制相结合，从根本上提高了河网提取的精度，具有较高的理论研究意义与实用价值。

坡面径流法是最为常用的河网提取方法，其中DEM 中洼地及平地流向的确定与流量累积矩阵的计算是影响提取精度的关键问题，也是国内外研究的热点与难点问题。本研究拟在充分挖掘辅助数据中的有效地形信息，设计算法来确定DEM 中洼地和平地的合理流向。在此基础上，结合研究土壤、植被、土地利用及地质条件等流域特征对产流量的贡献，来反演非均质流域下垫面栅格的相对产流系数，用于计算相应的实际流量累积矩阵。本研究将传统基于数据的河网提取方法与坡面径流形成的物理机制相结合，从根本上提高了河网提取的精度，具有较高的理论研究意义与实用价值。.本项目研究了DEM填洼方法。根据DEM数据及水流的特点，提出了一种基于洼地特征点的定向侵蚀算法来解决洼地处理方法。算法通过划分集水洼地，寻找洼地出口、入口、最低点，对出水路径、入水路径沿特定方向进行合理侵蚀，使断续、封闭的洼地依次开放，合并，扩大，从而获得无洼地的DEM。该方法在较大限度的保留DEM高程信息的前提下，有效解决了洼地地区河网提取时的伪河道、平行河道、假坝体等问题。研究结果表明，无论是提取水系的精确还是原始DEM的保真度，本文方法都明显的优于传统的算法。.本项目研究了DEM配准方法。本文提出了一种基于流域中心点的DEM配准方法。该方法利用数字水文分析方法，从不同的DEM中提取相同尺寸的子流域，分别计算子流域的Hu矩并将其作为相似性描述子(descriptor),在此基础上利用最小距离与第二小距离比值法将对应的子流域进行匹配，然后提取匹配后子流域的中心点作为候选控制点。利用RMSE阈值法对候选控制点进一步筛选，确定最终高精度的控制点，最后利用三维相似变换模型实现DEM配准。研究结果表明，该方法稳定性强，可以通过改变子流域尺度灵活提取不同数量的控制点，配准精度可达到亚像元水平。.本项目研究了考虑流域下垫面异质性的栅格相对产流系数计算方法。以南北环境差异较大的雅砻江流域为研究区域，本研究考虑了降雨量、地形起伏度、岩性渗透性和植被覆盖度等因子，通过环境因子分区和环境因子权重矩阵等两种方法，计算下垫面异质性的栅格相对产流系数，从而消除流域环境差异对河网提取准确度的影响。结果表明，考虑环境因子空间差异的方法提取了更多的河道，且并没有多出支流河道的现象，结果与实际更相符。