Table 4 The ablation study compares the coupled model and the baseline model using the same evaluation metrics at the LHK and JP hydrological stations in the testing dataset.

From: Leveraging multi-source data and teleconnection indices for enhanced runoff prediction using coupled deep learning models

Evaluation indicators

LHK

JP

SVR

SRA-SVR

MLPR

SRA-MLPR

SVR

SRA-SVR

MLPR

SRA-MLPR

NSE

0.6792

0.7164±0.0006

0.7036

0.7598±0.0003

0.6369

0.6802±0.0005

0.6708

0.7094±0.0003

RMSE

341.8323

280.7276±2

303.6378

258.3484±2

601.3359

567.2251±2

590.3090

540.7824±2

MAE

238.3268

185.7391±1.6

218.7850

168.7936±1.5

383.5630

341.3531±1.6

369.3606

338.5220±1.5

MAPE

0.3479

0.2723±0.0005

0.2708

0.2290±0.0003

0.3153

0.2352±0.0005

0.2689

0.2257±0.0003

MSLE

0.1433

0.0984±0.0001

0.1208

0.0758±0.0006

0.1278

0.0767±0.0006

0.0894

0.0644±0.0006

  1. The units of RMSE and MAE in the table are both m3/s. Both SRA-SVR and SRA-MLPR were performed in 100 replicate trials due to the random initial weights of network, thus the metrics were expressed as the mean ± standard deviation.
Back to article page