Efficient Calibration/Uncertainty Analysis Using Paired Complex/Surrogate Models.

Area of Science:

• Environmental science
• Hydrogeology
• Computational modeling

Background:

• Detailed groundwater models face challenges with long run-times and convergence issues.
• These limitations hinder uncertainty quantification and decision-making in environmental management.
• Existing methods struggle with highly parameterized and complex simulation scenarios.

Purpose of the Study:

• To present a novel methodology for calibrating complex groundwater models.
• To enable efficient exploration of posterior predictive uncertainty.
• To overcome computational limitations in environmental process simulation.

Main Methods:

• Combines a complex groundwater model with simplified surrogate models.
• Utilizes gradient-based subspace analysis for parameter estimation.
• Employs surrogate models for Jacobian matrix computation and the complex model for parameter updates and predictions.

Main Results:

• The methodology effectively addresses long run-times and convergence problems.
• Demonstrated success using a density-dependent seawater intrusion model.
• Facilitates robust uncertainty quantification in complex hydrogeological settings.

Conclusions:

• The novel approach enhances the usability of detailed groundwater models for decision-making.
• It offers a computationally efficient solution for calibrating complex environmental models.
• This method improves the quantification and reduction of predictive uncertainty in hydrogeological studies.