Large language models: Tools for new environmental decision-making.

Area of Science:

• Environmental Engineering
• Artificial Intelligence
• Water Resource Management

Background:

• Environmental decision-making faces challenges in complexity and optimization.
• Large Language Models (LLMs) offer potential for augmenting traditional environmental workflows.
• Integrating AI with environmental models requires careful framework design.

Purpose of the Study:

• To explore the benefits and limitations of LLMs in environmental decision-making.
• To propose and compare two frameworks: LLM-assisted and LLM-driven.
• To assess framework performance using a water engineering case study (PFAS control).

Main Methods:

• Development of two generalizable frameworks: LLM-assisted and LLM-driven.
• Application of frameworks in a water engineering case study using Environmental Fluid Dynamics Code (EFDC).
• Evaluation of framework performance in optimizing environmental decisions and managing PFAS contamination.

Main Results:

• Both LLM frameworks contribute to environmental decision optimization, with differing applicability in complex scenarios.
• The LLM-assisted framework enhanced human decision-making for PFAS interception and flow rate regulation.
• The LLM-driven framework encountered limitations in complex parameter optimization due to LLM constraints.

Conclusions:

• LLMs can enhance, not replace, human expertise in environmental decision-making.
• Responsible human-AI collaboration is crucial for optimizing environmental management.
• The study establishes a foundation for integrating AI with conventional environmental modeling.