r2ogs5

r2ogs5 calibrates numerical groundwater flow models by integrating OpenGeoSys 5 with R and applying Bayesian optimization with surrogate modeling to efficiently explore model parameter spaces.

Key Features:

• Integration with OpenGeoSys v5: Incorporates OpenGeoSys 5 into R to run numerical groundwater flow simulations.
• Bayesian optimization and surrogate modeling: Employs a sequential model-based optimization approach combining Bayesian optimization with surrogate models to handle multi-modal objective functions.
• Calibration efficiency: Reduces computational demand relative to gradient search and Latin hypercube sampling by requiring fewer model runs, enabling calibration of high-dimensional parameter sets (e.g., 12 parameters).
• Statistical analysis and visualization in R: Leverages R's statistical tools for analyzing calibration results and visualizing data.

Scientific Applications:

• Climate impact studies: Assess regional impacts of climate trends and extreme weather events on groundwater balance using high-resolution and large-scale models.
• Parameterization of groundwater models: Support detailed parameterization to accommodate increasing complexity in modern groundwater modeling.

Methodology:

r2ogs5 uses a sequential model-based optimization approach that integrates Bayesian optimization with surrogate modeling to explore and exploit groundwater model parameter spaces, demonstrated on 4-parameter and 12-parameter calibration examples.