Author Affiliations
1College of Resources & Environmental Science, Xinjiang University, Urumqi, Xinjiang 830046, China;2Key Laboratory of Oasis Ecology, Ministry of Education, Xinjiang University, Urumqi, Xinjiang 830046, China3Key Laboratory of Smart City and Environment Modelling of Higher Education Institute, Xinjiang University, Urumqi, Xinjiang 830046, Chinashow less
Fig. 1. Distribution of the study area and sampling points
Fig. 2. PCA-Mahalanobis distance distribution
Fig. 3. Original spectra and the pretreated soil spectral reflectance curves. (a) Original spectral reflectance; (b) spectral reflectance after SG smoothing; (c) spectral reflectance corrected for multiple scattering; (d) spectral reflectance treated with first order differentiation
Fig. 4. Contribution diagram of the first 10 variables. (a) Original spectral reflectance; (b) spectral reflectance after SG-MSC treatments; (c) spectral reflectance after SG-MSC-FD treatments
Fig. 5. Comparison of soil organic matter prediction variables
Fig. 6. Correlation between different soil parameters (n=101), in which the curves are fitting curves
Fig. 7. Correlation between SOM, EC, Fe and pH and original spectral reflectance (n=101)
Fig. 8. Correlation between soil organic matter and the first five principal components for original spectrum and preprocessed spectra under two spectral treatments of SG-MSC and SG-MSC-FD
Fig. 9. Fitting scatter diagrams of PLSR model under three strategies. (a) Model 1; (b) model 2; (c) model 3; (d) model 4; (e) model 5; (f) model 6; (g) model 7
Fig. 10. VIP values of prediction variables in different PLSR models. (a) Model 3; (b) model 4; (c) model 5
Property | Dataset(CV/%) | n | Min | Mean | Max | Std |
---|
Content ofSOM /(g·kg-1) | Whole(57.91) | 101 | 0.60 | 8.94 | 23.00 | 5.18 | Calibration(57.33) | 68 | 0.60 | 8.90 | 23.00 | 5.10 | Validation(59.92) | 33 | 1.3 | 9.02 | 21.72 | 5.41 | EC /(dS·cm-1) | Whole(86.52) | 101 | 0.05 | 6.54 | 28.40 | 5.66 | Content of Fe /(g·kg-1) | Whole(56.93) | 101 | 0.10 | 13.18 | 25.91 | 7.67 | pH | Whole(4.12) | 101 | 8.22 | 8.87 | 9.93 | 0.37 |
|
Table 1. Statistical characteristics of soil properties
Strategy | Variable | Calibration set | Validation set |
---|
Model number | R2 | RMSE | R2 | RMSE | RPD |
---|
StrategyI | Original spectrum | 1 | 0.69 | 3.42 | 0.66 | 3.12 | 1.73 | SG-MSC | 2 | 0.69 | 3.35 | 0.67 | 3.10 | 1.76 | SG-MSC-FD | 3 | 0.84 | 2.14 | 0.82 | 2.51 | 2.15 | StrategyII | Soil auxiliary covariates | 4 | 0.44 | 4.19 | 0.40 | 4.46 | 1.21 | StrategyIII | Original spectrum combined withsoil auxiliary covariates | 5 | 0.75 | 3.09 | 0.67 | 3.10 | 1.74 | SG-MSC combined with soil auxiliary covariates | 6 | 0.86 | 1.91 | 0.83 | 2.54 | 2.13 | SG-MSC-FD combined with soil auxiliary covariates | 7 | 0.91 | 1.54 | 0.88 | 1.20 | 2.70 |
|
Table 2. PLSR modeling results under three strategies