Identifying Atmospheric Nucleation Events Using Machine Learning

Abstract

[Abstract] Atmospheric nucleation (New Particle Formation) is not only a key process in aerosol dynamics, but it also assists in regulating the planet’s radiative balance. Accurate detection is essential to understand its implications for both climate and public health. However, manually identifying these events from particle size distributions is a slow and tedious process. This work studies the feasibility of a proposal based on machine learning and computer vision to classify nucleation events from images (surface plots) of particle distribution time series. To this end, different preprocessing configurations are explored and both classical models and deep neural networks are tested. Preliminary results show promising performance, highlighting the system’s ability to identify positive events with high sensitivity, suggesting a possible future integration into atmospheric monitoring platforms.