论文部分内容阅读
人类活动近年成为数字土壤制图亟需考虑的要素。本文以农业活动中轮作模式为例,将轮作信息应用于数字土壤制图,探讨其对土壤空间变异刻画的有效性。以安徽宣城两个县市的耕地平区为研究区,通过野外调查获得近年三种主要轮作模式,基于监督分类对多期遥感影像解译得到轮作类型空间分布图,使用方差分析探讨轮作对土壤表层有机质空间变异是否有显著性影响,采用随机森林重要性指标对自然环境因子、轮作模式、土地利用方式和归一化植被指数进行重要性排序,并构建不同的环境因子组合,利用基于相似度的土壤推测模型和随机森林模型进行制图和交叉验证。结果表明,轮作模式对土壤表层有机质有显著性影响,其重要性排序为第二,引入轮作使得基于相似度的土壤推测模型和随机森林模型制图精度分别提高4.8%~65.9%和1.9%~2.7%。
In recent years, human activities have become an important element to be considered in digital soil mapping. This paper takes the rotation pattern of agricultural activities as an example, applies rotation information to digital soil mapping, and discusses its effectiveness in characterizing soil spatial variation. Taking the arable land in two counties and cities in Anhui Province as the study area, three major rotation patterns in recent years were obtained through field investigation. Based on the supervised classification, the spatial distribution of rotation type was obtained by interpreting the multi-period remote sensing images. The analysis of variance Soil organic matter spatial variability was significantly affected by the use of random forest importance index of natural environmental factors, crop rotation patterns, land use patterns and normalized vegetation index order of importance and the construction of different combinations of environmental factors, the use of similar Degree of soil speculation model and random forest model mapping and cross-validation. The results showed that rotation patterns had a significant effect on the surface organic matter of soils, and the order of importance was the second. The introduction of rotation made the similarity between soil speculation model and stochastic forest model mapping accuracy increased by 4.8% -65.9% and 1.9% -2.7 %.