Abstract Ammonia (NH3) is a gaseous pollutant with significant environmental and health implications. Over recent decades, its increasing concentrations, driven by industrialization and agriculture, have necessitated high-resolution monitoring. However, limited daily ground-based observations hinder comprehensive analysis. This study developed machine learning-based frameworks—deep neural network (DNN), random forest, and light gradient boosting machine—to predict biweekly NH3 concentrations and downscale them to daily estimates across the United States during 2017–2022. The models integrate NH3 column concentrations, meteorological variables, land cover data, livestock information, and ground-based measurements. Among the models, DNN showed superior performance in both spatial cross-validation and independent testing, achieving a correlation coefficient of 0.79, a root mean square error of 0.98 µg/m³ , and an index of agreement of 0.83— effectively capturing fine-scale spatial variations at a 9 km resolution. Shapley additive explanations analysis identified temporal dynamic factors—such as day of year and meteorological variables—as key predictors, along with land cover and cattle density, highlighting the model’s ability to support the temporal downscaling of NH3 from biweekly to daily scale. When applied to the UK, the model demonstrated its potential for spatial transferability via the leave-one station-out and leave-one year-out cross validations. These findings highlight the ability of machine learning in bridging temporal gaps and generating high-resolution daily NH3 estimates.

A novel artificial intelligence (AI) technology now makes it possible to monitor ammonia (NH₃)—a key contributor to harmful fine dust particles—with unprecedented precision and spatial detail, addressing longstanding gaps in current observation methods.

Led by Professor Jungho Im in the Department of Civil, Urban, Earth, and Environmental Engineering at UNIST, the research team has successfully developed an AI model capable of estimating daily atmospheric ammonia concentrations with high accuracy.

Ammonia is emitted from various sources, including agricultural fertilizers, livestock waste, and fire incidents. While relatively harmless on its own, ammonia reacts with atmospheric sulfuric and nitric acids to form fine particulate matter (PM2.5), which poses serious health and environmental risks. Precise monitoring of ammonia levels is thus vital for accurate air quality forecasts and effective policymaking.

However, due to ammonia’s short atmospheric lifespan and the limited number of ground-based monitoring stations, existing data are typically restricted to biweekly intervals. Climate models that estimate ammonia over large regions often suffer from significant regional inaccuracies, limiting their usefulness for localized air quality management.

To overcome these challenges, the team developed an advanced deep neural network-based AI model that enhances both the temporal frequency and spatial resolution of ammonia monitoring. By integrating climate data from the European Centre for Medium-Range Weather Forecasts (ERA5), satellite-derived ammonia column measurements from the IASI instrument, and ground-based observations from the U.S. Ammonia Monitoring Network (AMoN), the model effectively downscales biweekly data into high-resolution daily estimates.

The AI model demonstrated outstanding performance, reducing prediction errors by up to 1.8 times compared to the European Monitoring and Evaluation Programme (CAMS) climate model. Notably, although trained primarily on U.S. data, the model successfully identified high-amplitude pollution events, such as the widespread fire in Manchester, UK, in 2019—highlighting its strong potential for broader spatial application and real-world deployment.

This research was led by first authors Saman Malik and Eunjin Kang. Professor Im emphasized that, unlike traditional climate models like CAMS or sparse ground stations, this AI approach can deliver continuous, high-resolution ammonia monitoring. "This technology can significantly improve air quality forecasts related to nitrogen-based pollutants and support more effective environmental policies," he stated.

He further added, "Applying this model domestically could enable real-time, high-resolution monitoring of ammonia concentrations across the country, marking a crucial step toward more precise air quality management and public health protection."

The study was published in the esteemed journal Journal of Hazardous Materials (Impact Factor = 11.3) on September 15, 2025. Funding support was provided by the Korea Environment Industry & Technology Institute (KEITI) and the National Research Foundation of Korea (NRF), under the Ministry of Environment (ME) and the Ministry of Science and ICT (MSIT).

Journal Reference
Saman Malik, Eunjin Kang, Yoojin Kang, et al., "Bridging temporal gaps: AI-based temporal downscaling of biweekly NH3 to daily scale with spatial transferability," J. Hazard. Mater., (2025).