Predicting the Water Level Fluctuation in an Alpine Lake Using Physically Based, Artificial Neural Network, and Time Series Forecasting Models
Table 1
Assessment of water level prediction performance for three-dimensional hydrodynamic model (EFDC), artificial neural network model (ANN), time series forecasting model (ARMAX), and the combination model (EFDC + ANN).
Statistical parameter
MAE (cm)
RMSE (cm)
SS
Calibration
EFDC model
2.56
3.70
0.990
0.974
ANN case 1
0.87
1.41
0.998
0.996
ANN case 2
0.81
1.21
0.999
0.997
ANN case 3
0.77
1.17
0.999
0.997
ARMAX case 4
0.98
1.71
0.997
0.994
ARMAX case 5
0.96
1.64
0.997
0.995
ARMAX case 6
0.96
1.64
0.997
0.995
EFDC + ANN case 7
0.81
1.26
0.998
0.997
EFDC + ANN case 8
0.74
1.08
0.999
0.998
EFDC + ANN case 9
0.72
1.05
0.999
0.998
Validation
EFDC model
2.75
3.53
0.923
0.769
ANN case 1
0.84
1.27
0.985
0.970
ANN case 2
0.78
1.15
0.988
0.976
ANN case 3
0.77
1.12
0.988
0.977
ARMAX case 4
0.83
1.18
0.987
0.974
ARMAX case 5
0.83
1.18
0.987
0.974
ARMAX case 6
0.83
1.18
0.987
0.974
EFDC + ANN case 7
0.85
1.29
0.986
0.970
EFDC + ANN case 8
0.82
1.19
0.988
0.974
EFDC + ANN case 9
0.80
1.17
0.989
0.975
Case 1: the input nodes include and ; case 2: the input nodes include , , and ; and case 3: the input nodes include , , , and . Case 4: ARMAX (1, 1, 1); case 5: ARMAX (2, 1, 1); case 6: ARMAX (3, 1, 1). Case 7: the input nodes include , , and ; case 8: the input nodes include , , , and ; case 9: the input nodes include , , , , and .
