Comparison of different heuristic and decomposition techniques for river stage modeling

Area of Science:

• Hydrology and Water Resources
• Computational Intelligence
• Environmental Modeling

Background:

• Accurate daily river stage modeling is crucial for water resource management and flood forecasting.
• Traditional soft computing models often face limitations in capturing complex hydrological dynamics.
• Hybrid approaches are explored to enhance the predictive capabilities of existing models.

Purpose of the Study:

• To propose and evaluate hybrid soft computing models for daily river stage prediction.
• To compare the performance of models based on variational mode decomposition (VMD) with single and ensemble empirical mode decomposition (EEMD) based models.
• To identify the most effective hybrid model for reliable river stage forecasting.

Main Methods:

• Developed hybrid models by combining variational mode decomposition (VMD) with artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and random forest (RF).
• Assessed model performance using efficiency indices and graphical comparisons.
• Compared VMD-based hybrid models (VMD-ANN, VMD-ANFIS, VMD-RF) against single models (ANN, ANFIS, RF) and EEMD-based models (EEMD-ANN, EEMD-ANFIS, EEMD-RF).

Main Results:

• VMD-based hybrid models (VMD-ANN, VMD-ANFIS, VMD-RF) significantly outperformed their single-model counterparts (ANN, ANFIS, RF).
• VMD-based models showed slightly superior performance compared to EEMD-based models.
• The VMD-ANFIS model achieved the highest efficiency and accuracy, followed by EEMD-ANFIS, VMD-ANN, VMD-RF, and ANFIS.

Conclusions:

• Variational mode decomposition (VMD) effectively enhances the performance of conventional soft computing models for river stage modeling.
• VMD proves more effective than EEMD in developing high-performing hybrid models.
• Hybrid modeling using VMD, particularly with ANFIS, offers a superior and reliable approach for daily river stage prediction.