针对基于少量典型样点土壤属性空间分布推测模型中的土壤属性参数敏感性问题,以坡位渐变信息结合典型土壤样点的加权平均模型为例,利用东北地形平缓小流域的土壤表层有机质含量样点集,使用阶乘设计、箱线图分析、扰动分析法和本文新设计的MR指数评价该模型的参数敏感性。结果表明,该模型中土壤属性参数敏感性较大,其大小与典型样点空间分布有关。敏感性主要由应用该模型时采用的坡位分类体系的不确定性引起。该文的分析方法可用于对基于少量典型样点的土壤属性空间分布推测模型进行参数敏感性综合分析。
Although the typical-sample-based prediction models of the spatial distribution of soil property are promising in digital soil mapping, the sensitivity analysis of the soil property parameter in these models is still lack. This paper presents an attempt to establish weighed average model on a fuzzy slope position based on typical samples for digital soil mapping. The soil samples of A-horizon soil organic matter were collected in a low-relief small watershed in Northeast China. A "factorial design" method was applied to simulating scenarios of modeling point set. The performances of the model under different scenarios were analyzed with box plot. Further, we designed an MR index, which was calculated from the median and the range of the prediction error, for parameter sensitivity analysis based on the prediction performance of model. Both MR index and existing sensitivity index were used in this study. The result shows that the weighed average model is sensitive to the soil property parameter. The sensitivity degree is related to the locations of typical soil sample. The method proposed in this paper is also available for the parameter sensitivity analysis of other typical-sample-based prediction models of digital soil mapping.