Hyperspectral inversion of soil heavy metals in Three-River Source Region based on random forest model

•Optimal combination of spectrum transformation and model were selected for different heavy metal.•Results showed that the accuracy of RF was higher than that of SVM and PLS.•This study can provide technical support for the nondestructive monitoring of soil heavy metal content. Hyperspectral remote...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Catena (Giessen) Ročník 202; s. 105222
Hlavní autoři: Zhou, Wei, Yang, Han, Xie, Lijuan, Li, Haoran, Huang, Lu, Zhao, Yapeng, Yue, Tianxiang
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier B.V 01.07.2021
Témata:
catenas
China
clay
heavy metals
Hyperspectral
pollution
Random forest model (RF)
reflectance spectroscopy
Soil heavy metal
soil organic matter
Support vector machine (SVM)
support vector machines
Three-River Source Region
China
Three-River Source Region
Support vector machine (SVM)
Soil heavy metal
Random forest model (RF)
Hyperspectral
ISSN:0341-8162, 1872-6887
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•Optimal combination of spectrum transformation and model were selected for different heavy metal.•Results showed that the accuracy of RF was higher than that of SVM and PLS.•This study can provide technical support for the nondestructive monitoring of soil heavy metal content. Hyperspectral remote sensing technology has considerable research value in monitoring and evaluating soil heavy metal pollution. In this study, the Three-River Source Region was taken as the study area. The occurrence relationship of six heavy metals in soil, such as Mn, Cu, Zn, Pb, Cr, Ni, with soil organic matter, clay minerals, and iron-manganese oxides, was studied through the determination and analysis of soil samples and the collection of soil reflectance spectrum. Spectral transformation was carried out by first derivative, second derivative, inverse-log, continuum removal and multiple scattering correction of the spectrum. The correlation between soil heavy metal content and soil spectrum was analyzed to select the characteristic band, and partial least squares (PLS) method, support vector machine (SVM) method and random forest (RF) model were used to build inversion model based on characteristic band. Then the best combination of spectral transformation and inversion model were explored. The results showed that Pb contents were the twice of the background in Qinghai province. The combination spectrum processing method can improve the correlation between spectrum and heavy metals. The location and quantity of characteristic bands of six heavy metals are different. The accuracy of RF was significantly better than that of SVM and PLS for all six heavy metal (i.e. pb: R2RF = 0.83, R2SVM = 0.62, R2PLS = 0.18), and the model effective of soil properties in non-polluted sites were reliable (i.e. clay: R2RF = 0.93, R2SVM = 0.87, R2PLS = 0.74). This study can provide technical support for the larger-scale monitoring of soil heavy metal content and heavy metal pollution assessment.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0341-8162
1872-6887
DOI:10.1016/j.catena.2021.105222