摘要
本文以渭干河–库车河绿洲(简称渭–库绿洲)土壤颗粒为研究对象,采集了绿洲内50个典型表层(0~10 cm)土壤样本,通过相关软件,提取到遥感指数变量、地形和气候等环境变量,经过相关性分析确定环境变量和预测目标间的关系,使用R语言构建了预测土壤颗粒含量的随机森林(random forest,RF)模型和极端梯度提升(extreme gradient boosting,XGBoost)模型。研究结果表明:XGBoost模型的预测结果整体好于RF模型,其中相关系数介于0.39~0.78;土壤pH、高程及衍生变量、光谱变换变量均是两个模型预测土壤颗粒含量的重要因子;将模型预测结果、实测数据和世界土壤数据库(HWSD)中的3种土壤颗粒数据作对比分析,结果表现出模型预测数据的误差小于HWSD与实测数据的误差。综上所述,通过筛选环境变量建立的XGBoost模型,是预测渭–库绿洲土壤颗粒含量的有效方法。
In this paper,soil particles in the Weigan River-Kuche River Oasis(referred to as the Wei-Ku oasis)were used as the research object,fifty typical surface(0–10 cm)soil samples were collected from the oasis,and environmental variables such as remote sensing index variables,topography and climate were extracted through relevant software.After correlation analysis to determine the relationship between environmental variables and prediction targets,a random forest(RF)model and an extreme gradient boosting(XGBoost)model for predicting soil particle contents were constructed using R language.The results show that the prediction results of the XGBoost model are better than those of the RF model,with the correlation coefficients ranging from 0.39 to 0.78.Soil pH,elevation and derivative variables,and spectral transformation variables are all important factors in the prediction of soil particle contents in both models.The errors of model prediction data are smaller than those of HWSD and measured data.In conclusion,the XGBoost model established by screening environmental variables is an effective method for predicting soil particle content in the Wei-Ku oasis.
作者
顾永昇
丁建丽
韩礼敬
李科
周倩
GU Yongsheng;DING Jianli;HAN Lijing;LI Ke;ZHOU Qian(Key Laboratory of Smart City and Environment Modelling of Higher Education Institute,College of Geography and Remote Sensing Sciences,Xinjiang University,Urumqi 830046,China;Key Laboratory of Oasis Ecology,Xinjiang University,Urumqi 830046,China)
出处
《土壤》
CAS
CSCD
北大核心
2023年第2期426-432,共7页
Soils
基金
新疆维吾尔自治区自然科学基金重点项目(2021D01D06)
国家自然科学基金项目(41961059)资助。
关键词
土壤颗粒
高光谱
环境变量
机器学习
Soil particles
Hyperspectra
Environmental variables
Machine learning
