土壤有机碳是重要的土壤化学性质之一，了解农牧交错带地区土壤有机碳的垂直分布特征对于合理规划土地利用格局，开展精细农业，防止土地退化具有重要意义。本文结合多种环境协变量，运用随机森林模型，对河北省丰宁坝上农牧交错带地区0-5cm层和35-40cm层土壤有机碳的浓度差进行制图，探讨土壤有机碳垂直分布差异的空间分布和形成机理，将制图结果与先分层预测土壤有机碳浓度，再进行栅格减法的制图结果进行比较，讨论了制图精度与误差来源。结果表明，地形因素特别是海拔对研究区土壤有机碳的垂直分布影响较大，整体上看，海拔越高，土壤有机碳垂直差异越大，高海拔阴坡的林区土壤有机碳垂直差异最大；海拔较低地区分布的稀疏草地和耕地土壤有机碳垂直差异较小。两种制图结果在大趋势上相似，都出现了预测结果向中心集中的现象，相比于本文采用的制图方法，先分层预测再进行栅格减法的结果对研究区土壤有机碳垂直分布差异的细节预测更好，更接近真实情况。最后，在误差分析中指出，随机森林的回归模型的预测特点可能是误差主要来源。

Soil Organic Carbon (SOC) is one of the most important chemical features of soil. It is important to understand the vertical distributions of SOC in agro-pastoral transitional zone, which is meaningful for planning of land use, precision agriculture and protection of soil. In this study, combining field work data and environmental data such as terrain factor and vegetation factor, we use Random Forest regression model to predict and map the difference of concentration of SOC between two profiles (0-5 cm and 35-40 cm) (method Ⅰ). Based on the results, we discuss the horizontal distributions of vertical differences of SOC between two profiles and explore the possible formation mechanism of the distribution. To estimate the precision of the results, we compare the result with another mapping method (method Ⅱ) that predicts the concentration of SOC in two profiles separately, then takes raster subtraction as the result. The result shows that terrain factor especially elevation is the most important environmental variable. In general, vertical difference of SOC is significant in woodland with high elevation, while in sparse grassland and cultivated land the vertical difference is not quite significant. Both of the methods tend to overestimate the lower data, and underestimate the higher data. Compared with method Ⅰ, method Ⅱ performs better on showing the details of real distribution and spatial variability. At last, we conclude that characteristic of Random Forest regressor method is the probable cause of the error.