A Comparison of Feature-Based MLR and PLS Regression Techniques for the Prediction of Three Soil Constituents in a Degraded South African Ecosystem
Table 4
Influence () of the five most important spectral variables on the regression equation of each of the feature-based regression models of approach A.
Spectral dataset
Rank 1
Rank 2
Rank 3
Rank 4
Rank 5
[%]
Variable
[%]
Variable
[%]
Variable
[%]
Variable
[%]
Variable
ASD spectral resolution
(1) In situ field
36.6
−24.0
−10.9
−6.0
−5.9
(2) Bare soil field
30.5
−21.1
−8.4
8.0
−7.7
(3) Laboratory
27.6
16.9
−14.1
−11.8
−9.0
(1) In situ field
−23.7
21.8
−12.0
4.9
4.9
Iron oxides
(2) Bare soil field
−17.6
10.6
−10.4
9.9
9.8
(3) Laboratory
29.2
−14.7
−10.3
6.7
−5.6
(1) In situ field
29.3
−29.1
7.9
5.6
4.9
Clay
(2) Bare soil field
−26.4
23.2
7.8
7.7
7.5
(3) Laboratory
−28.3
18.8
−12.4
12.0
7.5
HyMap spectral resolution
(1) In situ field
34.5
−20.8
−13.1
8.1
5.7
ASAF2330
(2) Bare soil field
30.4
−18.6
−12.0
9.3
7.4
(3) Laboratory
21.3
−14.2
10.1
−7.6
−7.4
(1) In situ field
−14.0
12.3
−10.4
8.8
7.9
Iron oxides
(2) Bare soil field
22.0
−16.7
9.5
−7.8
−7.3
(3) Laboratory
18.9
−18.0
11.7
−8.3
−7.1
(1) In situ field
24.4
−22.2
11.2
−6.6
−6.5
Clay
(2) Bare soil field
24.8
−21.7
11.8
7.7
−5.3
(3) Laboratory
26.3
−21.2
−14.3
11.6
−6.2
Symbols: absorption features (AFs): : area, : maximum depth, : wavelength of , : depth at wavelength position given in literature, ASAF: asymmetry factor. Hull features (HFs): : mean reflectance in interval, : slope in interval.
