Evaluating Lower Computational Burden Approaches for Calibration of Large Environmental Models

Area of Science:

• Environmental modeling
• Computational science
• Geosciences

Background:

• Environmental models for decision-making are often highly parameterized, making calibration computationally intensive.
• Traditional parameter estimation relies on Jacobian matrices, which are sensitive to numerical noise and require numerous forward runs.

Purpose of the Study:

• To compare the efficiency and effectiveness of different parameter estimation algorithms for environmental models.
• To identify methods that reduce computational burden and improve model-to-measurement fit.

Main Methods:

• Comparison of a traditional full Jacobian matrix approach with simultaneous increments, iterative ensemble smoother (IES), and randomized Jacobian methods.
• Application to a decision-making model for the Mississippi Alluvial Plain, USA.

Main Results:

• The IES and randomized Jacobian approaches achieved desirable model fits with significantly fewer forward runs than the traditional method.
• Both efficient methods obtained a good fit in fewer runs than the number of adjustable parameters.
• The simultaneous increments approach was less effective due to issues with parameter sensitivities.

Conclusions:

• Highly efficient algorithms can substantially accelerate environmental model parameter estimation.
• Faster calibration enhances model vetting and the utility of realistic environmental models for decision-making.