Improvements on Scalable and Reproducible Cloud Implementation of Numerical Groundwater Modeling

Area of Science:

• Environmental science
• Computational hydrogeology
• Cloud computing applications

Background:

• Groundwater modeling increasingly relies on computationally intensive methods requiring parallel processing.
• Existing cloud solutions for MODFLOW and PEST lack comprehensive reviews, especially concerning open-source and scalable implementations.
• Computational burden and resource accessibility limit complex groundwater simulations.

Purpose of the Study:

• To develop and evaluate an infrastructure-as-code architecture for parallel groundwater model execution on the cloud.
• To demonstrate a repeatable and efficient cloud-based solution for MODFLOW and PEST++ using open-source tools.
• To assess the scalability and performance of cloud-implemented groundwater modeling.

Main Methods:

• Utilized Docker containers and open-source software for cloud-based parallel PEST++ execution.
• Employed Amazon Web Services and Terraform for cloud infrastructure deployment and monitoring.
• Evaluated parallel performance using a publicly available MODFLOW-6 model, comparing local and cloud execution.

Main Results:

• Achieved near-perfect scaling for cloud-based groundwater model runs, with minimal time increase per model (0.02s) compared to local runs (12s per agent).
• Demonstrated efficient parallel execution of up to 200 concurrent model runs in the cloud.
• Successfully calibrated a consulting groundwater model using the developed cloud infrastructure, accelerating project completion.

Conclusions:

• The proposed cloud architecture offers a scalable, cost-effective, and efficient solution for complex groundwater modeling tasks.
• Infrastructure-as-code principles enable simple and repeatable deployment of parallel groundwater simulations in the cloud.
• Cloud computing provides a powerful platform for overcoming computational limitations in modern groundwater modeling.