Improved streamflow simulations in hydrologically diverse basins using physically informed deep learning models DOI
Bhanu Magotra, Manabendra Saharia,

C. T. Dhanya

et al.

Hydrological Sciences Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 27, 2025

Physically informed deep learning models, especially Long Short-Term Memory (LSTM) networks, have shown promise in large-scale streamflow simulations. However, an in-depth understanding of the relative contribution physical information models has been missing. Using a large-sample testbed 220 catchments hydrologically diverse regions Indian subcontinent, we quantify impact incremental addition on model performance using multiple variants LSTM based various combinations static catchment attributes and simulated land surface states. We found that trained with geophysics as additional input outperformed base terms nationwide median Kling-Gupta Efficiency (KGE) in-sample catchments, increasing KGE from 0.32 to 0.60. Additionally, retained significant prediction skill out-of-sample demonstrating pre-trained can be powerful tool predict data-scarce regions.

Language: Английский

Improved streamflow simulations in hydrologically diverse basins using physically informed deep learning models DOI
Bhanu Magotra, Manabendra Saharia,

C. T. Dhanya

et al.

Hydrological Sciences Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 27, 2025

Physically informed deep learning models, especially Long Short-Term Memory (LSTM) networks, have shown promise in large-scale streamflow simulations. However, an in-depth understanding of the relative contribution physical information models has been missing. Using a large-sample testbed 220 catchments hydrologically diverse regions Indian subcontinent, we quantify impact incremental addition on model performance using multiple variants LSTM based various combinations static catchment attributes and simulated land surface states. We found that trained with geophysics as additional input outperformed base terms nationwide median Kling-Gupta Efficiency (KGE) in-sample catchments, increasing KGE from 0.32 to 0.60. Additionally, retained significant prediction skill out-of-sample demonstrating pre-trained can be powerful tool predict data-scarce regions.

Language: Английский

Citations

0