顾及地貌结构特征的黄土沟头提取及分析

沟头是黄土地貌中发育最活跃的地貌部位，沟头的个体与群体对黄土沟间地的蚕食，成为监测黄土高原地面侵蚀演化的重要标识。有效获取并分析沟头的空间分布特征，对于系统、深入地研究黄土地貌的空间形态结构具有重要的意义。以陕北黄土高原典型地貌类型区为实验样区，设计并实现了基于5 m分辨率DEM的沟头提取算法，获取了各实验样区的沟头空间信息；运用点格局分析方法，分析了沟头的空间分布特征。实验结果显示，所提出顾及沟沿线的沟头自动提取方法，具有较好的合理性与精确性；同时，从黄土塬区、黄土残塬区到黄土丘陵沟壑区，随着沟壑发育程度的增强，沟头逐步逼近分水线，其空间分布呈现“远分水线聚集—随机—近分水线聚集”的变化格局。

Extraction and analysis of loess gully heads considering geomorphological structures

The gully head is the most active development position in loess landform. The individual and group of gully heads to nibble the loess area between gullies has a marked influence on monitoring the surface erosion and evolution in Loess Plateau. The effective extraction and investigation of gully heads is of critical significance to have a deeper study on the spatial form structure of loess landform systematically. In this paper, three typical landform areas of Loess Plateau are chosen as study areas which are Chunhua, Yijun and Yanchuan. They are located in loess tableland area, loess broken plateau area and loess hilly area respectively, and distributed in the north-south direction of the northern part of Shaanxi province. A gully head extraction model is designed based on 5-m resolution DEMs considering geomorphological structures. In each test area, the spatial information of gully heads is obtained by the extraction model, and the spatial distribution of gully heads is analyzed via the point pattern method. The result of precision evaluation shows that the automatic extraction method considering the shoulder line has 78.7% accuracy on the precision of quantity and 83.4% accuracy on the precision of shape consistent degree. The experiment of the spatial distribution characteristic is completed by the improved point pattern method. The experimental result shows that gully heads gradually approach to the divide line along with the increasing development degree of ravines. And in loess tableland area, loess broken plateau area and loess hilly area, the spatial distributions of gully heads respectively present three statues: clustered far from divide line, random and clustered near divide line. The application example also proves that the spatial distribution characteristic of gully heads is more in line with the actual situation when using the improved method called Area Decomposition Method to convert polygons to points. In conclusion, the research could be valuable not only in deepening the understanding of the spatial pattern of loess landform from the view of gully head, but also in offering a methodological support for digital terrain analysis on Loess Plateau.