Abstract Description: Modeling complex operating scenarios and/or statewide air quality management strategies results in the generation of large AERMOD datasets. The evaluation of these large datasets can be challenging and time-consuming. The integration of Artificial Intelligence (AI) offers a transformative approach to unravel this complexity. This study explores the innovative application of AI algorithms in analyzing and interpreting AERMOD modeling results to provide an understanding of air dispersion patterns, trends, and the range of potential source impacts.
Leveraging AI, we develop a framework that automates the analysis of simulated data from thousands of AERMOD runs for various pollutants and averaging periods, including NO2, PM10, and PM2.5, generated by iterating through different stack exhaust parameters, emission rates, and meteorological datasets. Our objective is to utilize machine learning (e.g., Random Forest) and deep learning (e.g., Neural Networks) to evaluate the potential for AI to analyze and predict impacts from stationary sources. We utilize a variety of AI tools (e.g., ChatGPT, Claude) to assess their performance in reading and understanding AERMOD results.
The AI’s performance is benchmarked against a set of success criteria, including precision in data interpretation and relevance and clarity in responding to user queries. Results indicate that the AI tools that we evaluated are able to read AERMOD dispersion results, respond to user queries, and interpolate concentrations for various release scenarios at an exceptional level. In our presentation, we will include specific cases demonstrating the AI model's capability in handling complex datasets, its accuracy in identifying dispersion patterns and trends, and its potential in providing more robust environmental assessments. Additionally, we will explore the AI model's implications for policy makers, environmental consultants, and stakeholders interested in transformative methods for air dispersion modeling.
This study utilizes a novel approach where AI serves not as a replacement but as a synergistic tool augmenting traditional environmental modeling techniques, setting a new benchmark for modern air quality management practices.
