Title: Cadmium concentration estimation in pen-urban agricultural soils: Using reflectance spectroscopy, soil auxiliary information, or a combination of both?

Author: Hong, YS (Hong, Yongsheng); Shen, RL (Shen, Ruili); Cheng, H (Cheng, Hang); Chen, SC (Chen, Songchao); Chen, YY (Chen, Yiyun); Guo, L (Guo, Long); He, JH (He, Jianhua); Liu, YL (Liu, Yaolin); Yu, L (Yu, Lei); Liu, Y (Liu, Yi)

Source: GEODERMA Volume: 354 DOI: 10.1016/j.geoderma.2019.07.033 Published: NOV 15 2019

Abstract: The regular monitoring of soil cadmium (Cd) in peri-urban agricultural soils is critical in maintaining soil security and balancing the local environment and ecosystem. Visible and near-infrared (Vis-NIR) spectroscopy provides an alternative to soil Cd estimation. Previous studies found a mild to moderate correlation of soil Cd against soil auxiliary information, including soil organic matter (SOM), Fe, and pH. This study aimed to determine whether the combination of spectral data and soil covariates would improve the model accuracy for soil Cd estimation. We collected 93 samples from a peri-urban agricultural area of Wuhan City, Central China. The reflectance spectra, Cd concentration, and soil properties, including SOM, Fe, and pH, were measured. In addition to the original spectra, the first (FDR) and second derivative reflectance were compared to process the spectra. Estimation models were developed using random forest. In particular, we tested two spectral reduction methods, namely, principle component analysis (PCA) and optimal band combination algorithm, to derive the spectral parameters. We compared three modeling strategies for soil Cd estimation as follows: strategy I with spectral data as input variables, strategy II with soil covariates as predictors, and strategy III with the combination of strategies I and II as predictors. Results indicated that the optimal band combination algorithm outperformed PCA regardless of the spectral transformations used. The inclusion of soil covariates into the spectral data improved the model prediction for soil Cd (validation r(2), 0.70-0.83) compared with the individual application of either spectral data (validation r(2), 0.55-0.68) or soil covariates (validation r(2) = 0.62) for modeling. The most successful prediction for soil Cd was achieved (validation r(2) = 0.83) using the combination of the optimal band combination algorithm on the basis of the FDR and additional soil information, for which the sum index, difference index, product index, and SOM were identified as important predictor variables. In summary, the proposed strategy of combining Vis-NIR spectroscopy with soil auxiliary information will increase the model performance for soil Cd and can be extended to other study sites.

WOS: 000486133300020

Language: English

Document Type: Article

Key words of author: Visible and near-infrared spectroscopy; Soil Cd; Soil auxiliary information; Optimal band combination algorithm; Random forest

Keywords Plus: ORGANIC-MATTER CONTENT; INTEGRATED FIELD SPECTROSCOPY; NEAR-INFRARED SPECTROSCOPY; HEAVY-METAL CONCENTRATIONS; REMOTE-SENSING TECHNIQUES; ENHANCED ASSESSMENT; SPATIAL-ANALYSIS; RAPID IDENTIFICATION; POLLUTION RISK; CONTAMINATION

Addresses: [Hong, Yongsheng; Cheng, Hang; Chen, Yiyun; He, Jianhua; Liu, Yaolin; Liu, Yi] Wuhan Univ, Sch Resource & Environm Sci, Wuhan 430079, Hubei, Peoples R China.

[Shen, Ruili] Hubei Acad Environm Sci, Wuhan 430072, Hubei, Peoples R China.

[Chen, Songchao] INRA, Unite InfoSol, F-45075 Orleans, France.

[Chen, Songchao] Agrocampus Ouest, INRA, UMR SAS, F-35042 Rennes, France.

[Chen, Yiyun] Chinese Acad Sci, State Key Lab Soil & Sustainable Agr, Nanjing 210008, Jiangsu, Peoples R China.

[Guo, Long] Huazhong Agr Univ, Coll Resources & Environm, Wuhan 430070, Hubei, Peoples R China.

[Yu, Lei] Cent China Normal Univ, Sch Urban & Environm Sci, Wuhan 430079, Hubei, Peoples R China.

[Yu, Lei] Cent China Normal Univ, Key Lab Geog Proc Anal & Simulat Hubei Prov, Wuhan 430079, Hubei, Peoples R China.

Addresses of reprint authors: Chen, YY; Liu, YL ，Wuhan Univ, Sch Resource & Environm Sci, Wuhan 430079, Hubei, Peoples R China.

Email: chenyy@whu.edu.cn; yaolin6100@sina.com

Impact factor：4.336