摘要
土壤有机质(SOM)对改善干旱区土地盐碱化、沙漠化和草场退化等环境问题发挥着重要作用。为了探索分数阶微分方法在SOM高光谱反演的可行性,以渭干河-库车河绿洲73个土壤样本为研究对象,通过测定其SOM含量与光谱反射率,利用0.2阶微分为步长,进行0~2阶分数阶微分的数学变换,分析分数阶处理光谱与SOM含量间的相关性,运用支持向量机回归、偏最小二乘回归和随机森林(RF)等方法对SOM含量进行定量反演。结果表明:采用RF的1.2阶微分建立的SOM含量反演模型预测精度最高,决定系数为0.93,均方根误差为1.62,相对分析误差为3.65。研究结果为精准反演该地区的SOM提供了依据,也为其他地区的SOM反演提供一定的参考。
Soil organic matter(SOM)plays an important role in ameliorating environmental problems such as land salinization,desertification,and grassland degradation in arid areas.To explore the feasibility of the fractional differential method in hyperspectral SOM inversion,73 soil samples from Weigan River to Kuqa River oasis were considered as research objects.By measuring the SOM content and spectral reflectance,the mathematical transformation of a fractional differential of order 0‒2 was performed using a 0.2-order differential as the step size.Further,the correlation between the fractional processing spectrum and SOM content was analyzed.Support vector machine regression,partial least squares regression,and random forest(RF)methods were used to quantitatively invert the SOM content.The results reveal that the prediction accuracy of the SOM inversion model established by the 1.2 RF derivative is the highest,with Coefficient of determination of 0.93,Root mean squared error of 1.62,and Relative percent difference of 3.65.These results can provide a basis for accurate inversion of SOM in this study area,and they also have a certain reference significance for inversion of SOM in other areas.
作者
李武耀
买买提·沙吾提
买合木提·巴拉提
Li Wuyao;Mamat Sawut;Maihemuti Balati(College of Geography and Remote Sensing Science,Xinjiang University,Urumqi 830046,Xinjiang,China;Xinjiang Key Laboratory of Oasis Ecology,Urumqi 830046,Xinjiang,China;Key Laboratory of Smart City and Environment Modelling of Higher Education Institute,Urumqi 830046,Xinjiang,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第7期394-401,共8页
Laser & Optoelectronics Progress
基金
新疆自然科学计划(自然科学基金)联合基金(2021D01C055)。
关键词
分数阶微分
高光谱
土壤有机质
模型估测
fractional differentiation
hyperspectrum
soil organic matter
model to estimation