基于随机森林算法的原始土壤图更新研究
Updating traditional soil maps based on random forest algorithm

以湖北省黄冈市红安县华家河镇滠水河流域为研究区，利用随机森林算法(random forest,RF)结合多源环境变量，对研究区原有的土壤图斑进行分解制图，将混合多种土壤类型的复合土壤图斑进行细化，在土壤多边形内部画出新的边界来代表单一土壤类型，并通过373个实地采样点验证更新后的土壤图。结果显示,更新后的土壤图其制图精度从原有的63%提高到了76%，展现了更为详细的空间细节和空间变化信息，表明利用随机森林算法进行数字土壤制图的可行性和可靠性。
The random forest algorithm (RF) combined with multi-source environmental variables was used to decompose the traditional soil maps of the Qinshui River Basin in Huajiahe Town,Hong’an County,Huanggang City,Hubei Province.The composite soil maps mixed with various soil types were refined,new boundaries were drawn inside the soil polygons to represent a single soil type,and the updated soil map was verified by 373 field sampling points.Results showed that the updated soil map improved its drawing accuracy from 63% to 76%,showing more detailed spatial details and spatial variation information,indicating that using the random forest algorithm to digitally map the traditional soil is feasible and reliable.