Tradeoff between cost and accuracy in large-scale surface water dynamic modeling

Area of Science:

• Hydrology
• Computational Fluid Dynamics
• Environmental Modeling

Background:

• The kinematic wave equation (KINE) is widely used for surface water dynamics but can lack accuracy.
• The local inertia formulation (INER) offers a more accurate yet cost-efficient alternative.
• Evaluating these formulations over large basins like the Amazon is crucial for understanding their performance.

Purpose of the Study:

• To compare the accuracy and computational costs of INER and KINE for simulating Amazon basin streamflows and water levels.
• To assess the performance of hybrid models (HYBR) that combine INER and KINE.
• To identify optimal configurations for hybrid models to balance accuracy and efficiency.

Main Methods:

• Synthetic experiments and comparison against ground-based observations in the Amazon basin.
• Evaluation of different time steps for INER and KINE simulations.
• Development and testing of five hybrid (HYBR) models with varying spatial distributions of INER and KINE.

Main Results:

• KINE simulations showed significant deterioration in accuracy along the Amazon river, with high RMSE values for streamflow and water level.
• KINE was found to be at least 25% more computationally efficient than INER.
• Hybrid models (HYBR) achieved substantial accuracy improvements, reducing maximum RMSE values and decreasing computational costs by 6-16%.

Conclusions:

• Hybrid models integrating INER and KINE offer a superior approach for simulating surface water dynamics in large river basins.
• Optimal performance was achieved when INER was applied to approximately one-third of the Amazon basin, balancing accuracy and computational efficiency.
• The optimal spatial distribution of INER and KINE in hybrid models may vary depending on regional hydrodynamics and geomorphological characteristics.