模糊坡位信息在精细土壤属性空间推测中的应用

Application of fuzzy slope positions in predicting spatial distribution of soil property at finer scale

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。

The spatial gradation of slope position has a great effect on the soil,hydrological,geomorphic phenomena and processes in small watershed or on slope.The fuzzy slope positions extracted with various methods can quantify the spatial gradation of slope positions and are considered as a kind of promising information to geographic modeling,such as digital soil mapping at finer scale.However,few studies have actually applied the fuzzy slope positions in geographic modeling.This paper attempts to examine the possibility of application of fuzzy slope positions in predicting spatial distribution of soil property at finer scale.In this case,two fundamental assumptions are made as follows： 1） terrain condition which can be comprehensively reflected by slope positions shows most important effect on spatial distribution of soil property in small catchment,and 2） soil property on typical slope position generally represents a typical value when soil property co-varies spatially with slope position.Based on these two assumptions,a weighted average model in which typical values of soil property on typical slope positions are weighted with fuzzy slope positions is developed to predict the spatial distribution of soil property（A-horizon soil organic matter in this study）.The fuzzy information of a system of five slope positions（i.e.,summit,shoulder slope,back slope,foot slope,and valley） was derived by a method based on typical locations of slope positions.The weighted average model was evaluated in a low-relief catchment（about 60 km2） of Nenjiang watershed in Heilongjiang Province,Northeast China.The multiple linear regression model based on topographic attributes was also applied to comparison of model performance.Three indices,i.e.correlation coefficient between predicted and observed values,mean absolute error（MAE） and root mean square error（RMSE） based on a validation set of 70 soil samples were calculated for quantitative assessment of the model performance.Results show that the weighted average method with very few modeling points can better predict the spatial distribution of A-horizon soil organic matter than the multiple linear regression model does.