Table 1 Model performance across different depth intervals.

From: Geospatial modeling of near subsurface temperatures of the contiguous United States for assessment of materials degradation

Model type

Depth interval

Validation set size

RMSE

MAE

Outside 1 degree (%)

Outside 2 degrees (%)

Outside 5 degrees (%)

Outside 10 degrees (%)

Neural Net

0–4000 m

595,315

29.99

19.08

95.12

89.33

75.08

54.22

Neural Net

50–350 m

258,227

13.18

8.29

89.95

79.97

55.42

27.83

Linear Interpolation

0–4000 m

334,028

6.77

3.11

65.26

48.57

20.29

2.70

Linear Interpolation

50–350 m

210,618

3.18

2.12

56.25

38.11

11.53

0.70

LightGBM

0–4000 m

605,139

5.88

2.42

38.76

28.12

13.45

5.77

LightGBM

50–350 m

260,645

2.44

0.56

10.67

6.26

1.77

0.76

LightGBM (Hybrid)

0–4000 m

486,124

8.68

3.40

48.58

36.48

18.94

8.03

LightGBM (Hybrid)

50–350 m

221,785

2.61

0.68

12.05

7.46

2.12

0.87

Back to article page