Table 4 Mean validation metrics across 100 bootstrap iterations.

From: CLSoilMaps: A national soil gridded database of physical and hydraulic soil properties for Chile

Soil Attribute

Model

Depth

RMSE

nRMSE

R2

PBIAS

Bulk Density

RF-U

0–5

0.18

8.96

0.82

0.84

RF-U

5–15

0.14

6.93

0.88

−0.04

RF-U

15–30

0.15

7.79

0.86

−0.81

RF-U

30–60

0.19

9.62

0.79

−0.90

RF-L

60–100

0.19

10.13

0.78

−0.89

RF-L

100–200

0.20

10.69

0.76

0.20

Clay

RF-U

0–5

7.23

11.74

0.68

2.75

RF-U

5–15

6.11

9.95

0.76

2.97

RF-L

15–30

7.29

10.21

0.75

0.27

RF-L

30–60

9.29

12.05

0.72

−7.17

RF-L

60–100

11.36

13.52

0.60

−5.93

RF-U

100–200

12.99

15.70

0.50

3.90

Sand

RF-U

0–5

10.83

11.16

0.75

0.25

RF-U

5–15

9.25

10.06

0.83

0.84

RF-U

15–30

9.16

9.78

0.83

0.04

RF-U

30–60

9.73

10.09

0.84

2.47

RF-U

60–100

13.78

13.89

0.67

−0.03

RF-U

100–200

13.91

14.63

0.71

−3.82

  1. RF-U: random forest model using the Upper set of covariates. RF-L: random forest model trained using the Lower set of covariates.