轮作模式在农耕区土壤有机质推测制图中的应用

Mapping Soil Organic Matter in Farming Areas with Crop Rotation

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%。

Most of the current digital soil mapping uses natural environmental covariates. However, human activities have had a profound impact on soil properties and thus become an important factor influencing soil spatial variability. In this paper, we introduced crop rotation as one of the agricultural activities into digital soil mapping and explored its effectiveness in characterizing soil spatial variability. Flat cultivated areas of Xuanzhou City and Langxi County in Anhui Province was chosen as the study area. Three main crop rotations were obtained through field investigation in 2010. The spatial distribution of crop rotation in the studied area was obtained by multi-phase remote sensing image interpretation.One-way analysis of variance （ANOVA） was used to analyzewhether topsoil organic matter content had significant different contents statistically among three crop rotation groups. Factor importance of seven natural environmental eovariates, crop rotation, land use and NDVI were generated by variable importance measurements of Random Forest. Different combination of environmental covariates were constructed according to the rankings of their importance. The Soil Landscape Inference Model （SoLIM） and Random Forest were used to predict soil organic matter mapping and the cross validation was used to evaluate the mapping accuracies. The results showed that topsoil organic matter content was significantly different statistically between three crop rotation groups, and the crop rotation was more important than parent material, land use or NDVI. In addition, the introduction of crop rotation improved the accuracy of soil mapping. This study demonstrated the usefulness of human activities in digital soil mapping and indicated the necessity for using human activity factors in digital soil mapping studies.