基于遥感数据和随机森林算法的土壤重金属空间分布模拟——以铬为例

Spatial Distribution Simulation of Soil Heavy Metals Based on Remote Sensing Data and Random Forest Algorithm-Taking Chromi⁃um as an Example

为评估土壤重金属的富集状态及空间分异态势,选取山东省章丘市为研究区,系统采集425处土壤样品,测定土壤中铬(Cr)元素含量,采用描述性统计特征评估重金属在土壤中的富集状态;获取与土壤采样同期的Landsat-8 OLI遥感数据,将土壤重金属的环境要素作为自变量,测定的土壤Cr元素含量为因变量,构建基于随机森林算法的土壤重金属空间模拟模型,完成土壤中的重金属含量预测和空间分布模拟。结果表明,土壤重金属Cr含量均值高出土壤元素背景值37.22%,但低于农用地土壤污染风险筛选值,表明土壤中Cr的富集在可管控范围内;随机森林算法支持的空间模拟模型具有较好的精度和稳定度,精度系数R2和RMSE值分别为0.87和7.19,优于普通克里格法(R^(2)=0.66,RMSE=13.15)对土壤重金属的空间分布模拟。

To evaluate the enrichment status and spatial differentiation of soil heavy metals,425 soil samples were systematically collected from Zhangqiu City of Shandong Province,to determine the elemental content of chromium(Cr)in soil,the enrichment status of heavy metals in soil was assessed using descriptive statistical characteristics;obtain Landsat⁃8 OLI remote sensing data from the same period as soil sam⁃pling,taking the environmental factors of heavy metals in soil as independent variables and the measured content of Cr element in soil as de⁃pendent variables,a soil heavy metal spatial simulation model based on random forest algorithm was constructed to complete the prediction and spatial distribution simulation of heavy metal content in soil.The results showed that the mean value of soil heavy metal Cr content was higher than the background value of soil elements by 37.22%,but lower than the screening value of soil pollution risk in agricultural land,indicating that the enrichment of Cr in soil was within the manageable range;the spatial simulation model supported by random forest algorithm had better accuracy and stability,with accuracy coefficient R2 and RMSE values of 0.87 and 7.19,respectively,which was better than the spatial distri⁃bution simulation of soil heavy metals by ordinary Kriging method(R^(2)=0.66,RMSE=13.15).