What does this research mean for the field?

A hybrid FEDformer-LSTM deep learning model, particularly when integrated with CEEMDAN preprocessing, significantly outperforms baseline deep learning architectures in multi-step ahead river streamflow forecasting. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To evaluate deep learning architectures, including a novel hybrid FEDformer-LSTM framework with CEEMDAN preprocessing, for multi-step ahead daily river streamflow forecasting.

September 10, 2026Engineering Applications of Computational Fluid MechanicsOpen Access

A hybrid FEDformer-LSTM model for accurate multi-step ahead river streamflow forecasting

View Full Paper

Ask AI

Bookmark

Authors

AYAtefeh Yarahmadi RARasoul Ameri AGAmin Gharehbaghi

Discussion

Member takes

Overview

Comparative modeling study demonstrates superior accuracy of hybrid deep learning architectures in river streamflow forecasting, highlighting their utility for flood control and water allocation.

Key Points

To evaluate deep learning architectures, including a novel hybrid FEDformer-LSTM framework with CEEMDAN preprocessing, for multi-step ahead daily river streamflow forecasting.
Assessed five baseline architectures (FEDformer, Informer, LSTM, Transformer, iTransformer) alongside a proposed hybrid FEDformer-LSTM scheme.
Evaluated models across two hydrometric monitoring stations (10AA001 and 10ED002) on the Liard River, Canada.
Implemented Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) as a data preprocessing method to improve prediction accuracy.
The proposed FEDformer-LSTM hybrid architecture outperformed the baseline FEDformer and all other standalone deep learning models in multi-step ahead streamflow forecasting.
Integrating CEEMDAN preprocessing enhanced performance across models, with CEEMDAN-LSTM achieving optimal forecasts at lead time t + 1 for station 10AA001 (MAE = 6.4435 m³/s, RMSE = 15.8598 m³/s, R² = 0.9989, NSE = 0.9988).
At station 10ED002, CEEMDAN-LSTM attained superior accuracy at lead time t + 1 (MAE = 55.7703 m³/s, RMSE = 96.1726 m³/s, R² = 0.9994, NSE = 0.999).

Cite This Study

Yarahmadi et al. (2026) studied this question.

synapsesocial.com/papers/6aa27bfe58559d80afc75811 https://doi.org/10.1080/19942060.2026.2728742

View Full Paper

Ask AI

Bookmark

September 10, 2026Engineering Applications of Computational Fluid MechanicsOpen Access

A hybrid FEDformer-LSTM model for accurate multi-step ahead river streamflow forecasting

View Full Paper

Ask AI

Bookmark

Authors

AYAtefeh Yarahmadi RARasoul Ameri AGAmin Gharehbaghi

Discussion

Member takes

Overview

Comparative modeling study demonstrates superior accuracy of hybrid deep learning architectures in river streamflow forecasting, highlighting their utility for flood control and water allocation.

Key Points

To evaluate deep learning architectures, including a novel hybrid FEDformer-LSTM framework with CEEMDAN preprocessing, for multi-step ahead daily river streamflow forecasting.
Assessed five baseline architectures (FEDformer, Informer, LSTM, Transformer, iTransformer) alongside a proposed hybrid FEDformer-LSTM scheme.
Evaluated models across two hydrometric monitoring stations (10AA001 and 10ED002) on the Liard River, Canada.
Implemented Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) as a data preprocessing method to improve prediction accuracy.
The proposed FEDformer-LSTM hybrid architecture outperformed the baseline FEDformer and all other standalone deep learning models in multi-step ahead streamflow forecasting.
Integrating CEEMDAN preprocessing enhanced performance across models, with CEEMDAN-LSTM achieving optimal forecasts at lead time t + 1 for station 10AA001 (MAE = 6.4435 m³/s, RMSE = 15.8598 m³/s, R² = 0.9989, NSE = 0.9988).
At station 10ED002, CEEMDAN-LSTM attained superior accuracy at lead time t + 1 (MAE = 55.7703 m³/s, RMSE = 96.1726 m³/s, R² = 0.9994, NSE = 0.999).

Cite This Study

Yarahmadi et al. (2026) studied this question.

synapsesocial.com/papers/6aa27bfe58559d80afc75811 https://doi.org/10.1080/19942060.2026.2728742

View Full Paper

Ask AI

Bookmark