Related Concept Videos
Modeling and Similitude
Typical Model Studies
Clearance Models: Physiological Models
The organ's clearance rate depends on the blood flow to the organ and the extraction ratio (E). The extraction ratio describes the organ's...
Models, Theories, and Laws
Mechanistic Models: Compartment Models in Algorithms for Numerical Problem Solving
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...
Language and Cognition
You might also read
Related Articles
Articles linked to this work by shared authors, journal, and citation graph.
Large language models: Tools for new environmental decision-making.
An Electrochemical Sensor Based on Electropolymerization of β-Cyclodextrin on Glassy Carbon Electrode for the Determination of Fenitrothion.
Correlating microbial community compositions with environmental factors in activated sludge from four full-scale municipal wastewater treatment plants in Shanghai, China.
Related Experiment Video
Updated: Jan 7, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
Large language models for environmental modeling: Framework, capabilities, constraints.
1Graduate School of Environmental Science, Hokkaido University, Sapporo, 060-0810, Japan.
Large Language Models (LLMs) offer new pathways for environmental modeling. A human-AI Copilot framework excelled in parameter calibration and real-time correction, while Autopilot faced limitations.
Area of Science:
- Environmental Science
- Artificial Intelligence
- Computational Modeling
Background:
- Environmental modeling complexity is increasing.
- Integrating Large Language Models (LLMs) into these workflows presents challenges.
- Practical frameworks for LLM integration in environmental modeling are needed.
Purpose of the Study:
- To introduce and evaluate two frameworks for embedding LLMs into environmental modeling workflows: a human-AI collaborative Copilot and an LLM-driven Autopilot.
- To assess the performance of these frameworks in parameter calibration and real-time correction using the Rainfall-Runoff-Inundation (RRI) model.
- To provide guidance for the generalizable deployment of LLMs in environmental modeling.
Main Methods:
- Development of two LLM integration frameworks: Copilot (human-AI collaboration) and Autopilot (LLM-driven automation).
- Application of frameworks to the Rainfall-Runoff-Inundation (RRI) model in Japan's Kuzuryu River basin.
- Evaluation of performance in parameter calibration and real-time correction tasks, utilizing prompt engineering and physics constraints.
Main Results:
- The Copilot framework demonstrated robust performance, achieving high accuracy in parameter calibration (NSE 0.91/0.81) and stable real-time correction.
- The Autopilot framework showed competence in physics-constrained calibration but failed in long-sequence real-time correction due to "attention decay".
- LLMs are effective as knowledge engines and coding assistants under human supervision (Copilot), but full automation (Autopilot) is limited by context window constraints.
Conclusions:
- Human-AI collaborative frameworks (Copilot) are currently more effective for complex environmental modeling tasks than fully automated ones (Autopilot).
- Strategic task design, human oversight, and addressing LLM limitations like "attention decay" are crucial for successful LLM integration.
- The study offers a methodological framework and design principles for deploying LLMs in environmental modeling, highlighting future research directions.

