论文部分内容阅读
坡位的空间渐变特征影响着小流域及坡面尺度上的土壤、水文、地貌等现象和过程,因此对精细尺度下的地理建模(如土壤空间信息推理)有重要作用。虽然目前已有多种模糊坡位信息定量提取方法,但所得到的模糊坡位信息还缺乏实际应用。本文以精细尺度下的土壤属性空间分布推测为例,对此展开探索。应用模型假设:(1)在小流域内,地形因素主导着土壤属性空间分布的变化;(2)典型坡位上对应分布着典型的土壤属性值,土壤属性与坡位之间存在协同变化关系。据此建立以模糊坡位信息对各类典型坡位上土壤样点属性值的加权平均模型,推测土壤属性的空间分布。模型应用于黑龙江省嫩江流域一个地形平缓的小区(面积约60 km2),通过一个以坡位典型位置作为原型的模糊坡位定量方法提取5类坡位(山脊、坡肩、背坡、坡脚、沟谷)的空间渐变信息,对土壤表层有机质含量的空间分布进行推测。推测结果通过研究区70个土壤采样点进行评价,以推测结果与评价样点集之间的相关系数、平均绝对误差、均方根误差作为定量评价指标,与使用常用地形属性的多元线性回归模型推测结果进行对比。评价结果表明,仅使用极少建模点的加权平均模型的推测结果优于多元线性回归模型的推测结果。
Spatial gradient characteristics of slope position affect the phenomena and process of soil, hydrology and landform on the scale of small watershed and slope. Therefore, it plays an important role in geographic modeling (such as soil spatial information reasoning) on the fine scale. Although there are many methods to quantitatively extract information of fuzzy slope position, the fuzzy position information obtained is still lack of practical application. This article takes the speculation of the spatial distribution of soil properties under the fine scale as an example to explore. The application of the model assumes that: (1) in the small watershed, the topographic factors dominate the spatial distribution of soil properties; (2) Typical slope values correspond to the distribution of typical values of soil properties, soil properties and slope position synergies . Based on this, a weighted average model of soil sample attribute values of various typical slope positions with fuzzy slope position information is established to estimate the spatial distribution of soil attributes. The model was applied to a flat terrain with an area of about 60 km2 in Nenjiang River Basin, Heilongjiang Province. Five types of slope positions (ridge, shoulder, back slope and foot of foot) were extracted by a fuzzy slope quantitative method taking the typical position of the slope as a prototype , Gully) spatial variability of soil organic matter content of the spatial distribution of speculation. The results were evaluated by 70 soil sampling sites in the study area. The correlation coefficient, average absolute error and root mean square error were used as quantitative evaluation indicators. The results were compared with the multiple linear regression model Conjecture results were compared. The evaluation results show that the weighted average model using only few modeling points is better than the multiple linear regression model.