Use of Airborne Hyperspectral Imagery to Map Soil Properties in Tilled Agricultural Fields
Table 3
Partial least squares (PLS) model accuracy in predicting soil analyte concentrations1 for (a) the 269 calibration samples using repeated leave-one-out cross validation and (b) 46 validation samples from the field (Temple 1S) that had been left apart from the calibration. Results were derived using a first derivative gap 2 math pretreatment, using spectra derived from unsmoothed imagery (2.5 m2 pixel size) or from spatially smoothed imagery to which a low pass filter had been applied. See Section 2 for analyte descriptions. Units refer to analyte residual mean squared error (RMSE) values. Bold indicates sd/sepred >1.20.
Factors no.
C%
Sand%
Silt%
Clay%
pH
OM%
K mg/kg
Ca mg/kg
Mg mg/kg
Mn mg/kg
Zn mg/kg
Fe mg/kg
Al mg/kg
8
9
10
9
10
10
4
8
9
10
8
10
10
(a) 269 calibration samples collected from five agricultural fields
using unsmoothed imagery
0.65
0.79
0.79
0.66
0.51
0.75
0.59
0.69
0.69
0.62
0.64
0.75
0.76
RMSE
0.19
7.9
6.9
2.2
0.4
0.4
89.5
166.1
50.3
19.6
1.4
49.3
104.7
using smoothed imagery
0.64
0.80
0.80
0.68
0.58
0.77
0.61
0.71
0.71
0.67
0.67
0.78
0.81
RMSE
0.18
7.2
6.2
2.0
0.4
0.3
86.1
151.8
45.5
17.8
1.2
43.8
89.0
(b) 46 validation samples collected from the remaining agricultural field (Temple 1S)
using unsmoothed imagery
2se pred
0.20
9.0
7.3
2.4
0.5
0.4
147.1
178.1
51.4
27.5
1.2
52.9
123.9
3bias
0.10
−2.88
1.42
1.00
0.03
0.34
24.92
40.45
31.84
3.23
0.45
2.12
107.33
bias/mean (pred)
0.08
−0.06
0.03
0.11
0.01
0.15
0.10
0.05
0.22
0.09
0.18
0.01
0.21
4sd/se pred
0.74
0.81
0.86
0.84
0.89
0.63
1.05
0.96
0.76
1.15
0.92
1.41
0.98
using smoothed imagery
se pred
0.15
9.5
7.9
2.9
0.4
0.3
154.9
233.6
58.5
38.3
1.3
98.0
170.4
bias
−0.10
−0.87
−0.66
0.95
−0.25
−0.13
−23.06
−2.46
−2.07
−0.38
0.10
−5.26
77.66
bias/mean (pred)
−0.08
−0.02
−0.02
0.10
−0.05
−0.06
−0.10
0.00
−0.02
−0.01
0.04
−0.03
0.14
sd/se pred
0.86
1.29
1.26
1.07
0.97
0.74
1.04
1.34
1.41
1.25
0.98
1.91
1.29
1Analytes that predicted poorly ( < 0.5) in the set of 315 samples (N, P, Cu, B, S, and P saturation) are not included here 2se pred: sd(observed-predicted) 3bias: mean(pred)-mean(obs) 4sd: standard deviation.