AI model enables fine-scale monitoring of ammonia emissions that contribute to fine particulate matter

New artificial intelligence technology (AI) now allows you to monitor ammonia (NH₃) – a key contributor to harmful fine dust particles – with unprecedented precision and spatial details, attacking longtime gaps in current observation methods.

Directed by Professor Jungho IM in the department of civil, urban, terrestrial and environmental genius of the UNIST, the research team has successfully developed a model of AI capable of estimating the daily concentrations of atmospheric ammonia with great precision.

The study was published in the Dangerous Journal.

Ammonia is issued by various sources, including agricultural fertilizers, livestock waste and fire incidents. Although relatively harmless in itself, ammonia reacts with atmospheric sulfuric and nitric acids to form fine particles (PM_2.5), which poses serious health and environmental risks. Precise monitoring of ammonia levels is therefore vital for precise air quality forecasts and the effective development of policies.

However, due to the short atmospheric lifespan of ammonia and the limited number of ground surveillance stations, existing data is generally limited to bihebdomedary intervals. The climatic models that consider ammonia on large regions often suffer from significant regional inaccuracies, which limits their usefulness for the localized management of air quality.

To overcome these challenges, the team has developed an AI model based on an advanced neural network which improves both temporal frequency and spatial resolution of ammonia monitoring.

By integrating the climatic data of the European center for medium -range weather forecasts (ERA5), AMMONAC column measures derived by satellite of the IASI instrument and the ground observations of the American ammonia surveillance network (AMON), the model effectively reduce Biweasly data in daily high -resolution estimates.

The AI ​​model has shown exceptional performance, reducing prediction errors up to 1.8 times compared to the climate model of the European surveillance and evaluation program (CAMS).

In particular, although trained mainly on American data, the model has successfully identified high -amplitude pollution events, such as generalized fire in Manchester, the United Kingdom, in 2019 – by lighting its high potential for a wider space application and a deployment of the real world.

This research was carried out by the first authors Saman Malik and Eunjin Kang. Professor IM pointed out that, unlike traditional climate models such as Cames or sparse soil stations, this AI approach can provide continuous and high resolution monitoring of ammonia.

“This technology can considerably improve air quality forecasts linked to nitrogen-based pollutants and support more efficient environmental policies,” he said.

He also added that “the application of this model at the national level could allow high -time high resolution monitoring of ammonia concentrations across the country, marking a crucial step towards more precise air quality management and public health protection.”