集成土壤-环境关系与机器学习的干旱区土壤属性数字制图

张振华,丁建丽,王敬哲,葛翔宇,王瑾杰,田美玲,赵启东

Digital Soil Properties Mapping by Ensembling Soil-Environment Relationship and Machine Learning in Arid Regions

ZhenHua ZHANG,JianLi DING,JingZhe WANG,XiangYu GE,JinJie WANG,MeiLing TIAN,QiDong ZHAO

表3 土壤属性建模集与验证集性能比较

Table 3 Performance comparison between soil attribute Calibration and Validation

土壤属性 Soil Attribute 模型 Model 建模集 Calibration 验证集 Validation R2 ρc RMSE R2 ρc RMSE pH RF 0.7929 0.7786 0.1681 0.6779 0.6084 0.2182 Cubist 0.5182 0.6356 0.2308 0.4383 0.5805 0.2252 Bagging 0.6389 0.6409 0.2092 0.5829 0.6159 0.2214 MLR 0.4154 0.4559 0.2971 0.3004 0.4683 0.2445 SSC RF 0.9067 0.9219 2.6680 0.7945 0.8377 3.1803 Cubist 0.8820 0.9237 2.9190 0.7135 0.6194 7.5771 Bagging 0.7417 0.8331 4.2156 0.6974 0.7791 3.9743 MLR 0.6478 0.7687 4.0600 0.5050 0.6670 5.7829 SOM RF 0.8565 0.7872 3.3215 0.7472 0.7009 3.5456 Cubist 0.5204 0.5774 4.4070 0.4795 0.4305 5.4758 Bagging 0.7653 0.7551 3.6085 0.6402 0.5494 4.2667 MLR 0.3835 0.5559 5.0483 0.3250 0.5151 4.8424