坡位渐变信息的模糊推理

Fuzzy inference of spatial gradation of slope positions

坡位的空间变化通常是渐变的,定量的坡位空间渐变信息对于精细尺度上的坡面土壤侵蚀、预测性土壤制图等应用具有重要意义。现有基于栅格DEM的坡位模糊识别方法,或是仅在属性域内模糊聚类,忽略了空间信息;或是需要繁琐的规则进行模糊分类,实用性受限。本文建立了一种基于相似度的模糊推理方法,根据各类坡位在空间上的典型位置,计算其他位置与典型位置间的相似度,从而对坡位空间渐变信息进行系统、定量的描述。应用表明本方法能够合理地描述山脊、坡肩、背坡、坡脚、沟谷等重要坡位类型的渐变信息,所获得的坡位渐变信息也能够合理地解释土壤样点的A层土壤含砂量随坡位渐变的变化趋势。

An important characteristic of slope positions （e. g. backslope, footslope） is that the transition between them is gradual. And the quantification of gradual transition of slope positions is very useful in many terrain-related geographical or ecological modelling （such as soil erosion on slope, soil surveying and mapping）, especially on fine scale. Among current approaches of fuzzy slope position which are based on the gridded DEMs, fuzzy k-means approaches often focus on the parameter space and ignore the spatial information. Moreover, this type of approach can not extract some slope positions when they just exist with a very small proportion of application area. And the rule-based approach requires intensive operations and has a high demand for user knowledge of local landform. So the practicability is limited. This paper proposed a similarity-based approach to quantitative fuzzy representation of spatial gradation of slope positions. This approach includes two steps： the first is to extract the typical locations of each slope position. Then based on both local topographic attributes and regional terrain index, the similarity between other locations and typical locations is computed. This approach is carried out in both attribute domain （i. e. parameter space） and spatial domain. Both local topographic information and terrain context are taken into account. Application shows that results of proposed approach can quantitatively describe the spatial gradation of slope positions （such as ridge, slope shoulder, back slope, footslope, and valley）. The analysis combining with spatial gradual transition of sand percentages in the A horizon of soil samples also indicates the reasonability of the results of the proposed approach.