Exploratory ensemble designs for environmental models using k-extended Latin Hypercubes.

Area of Science:

• Environmental modeling
• Computational science
• Statistical analysis

Background:

• Computer models are crucial for environmental studies but require efficient parameter space exploration.
• Existing methods like Latin Hypercube (LHC) sampling have limitations in complex model scenarios.
• Quantifying uncertainty in parameters, initial conditions, and boundary conditions is essential for model reliability.

Purpose of the Study:

• To present a novel class of designs for exploring computer model parameter spaces.
• To enhance the efficiency and coverage of initial ensemble designs.
• To facilitate simultaneous quantification of parametric and initial condition uncertainties in environmental models.

Main Methods:

• Expanding existing Latin Hypercube (LHC) technology to create designs composed of smaller, orthogonal LHCs.
• Developing a new emulator diagnostic to assess statistical model weaknesses.
• Applying the method to a 400-member ensemble of the Nucleus for European Modelling of the Ocean (NEMO) model.

Main Results:

• The proposed designs maximize parameter space coverage and orthogonality.
• The method effectively quantifies parametric uncertainty and uncertainties from initial/boundary conditions.
• A new emulator diagnostic successfully identified structural weaknesses in statistical modeling.

Conclusions:

• The novel design class offers a flexible and multi-purpose approach for initial computer model exploration.
• These designs are particularly beneficial for complex environmental and climate models.
• The method provides a robust framework for guiding further model development and calibration.