Speaker: Valery Ashu Ntui,  Lappeenranta-Lahti University of Technology.

Title: Neural ODE–Based Estimation of Reaction Rates in Atmospheric Chemical Systems with Uncertainty Quantification

Abstract: Air quality remains a critical global challenge, accounting for over 6 million premature deaths annually, according to the World Health Organization (2021) report, with atmospheric oxidation of volatile organic compounds (VOCs) driving the formation of secondary pollutants such as ozone and fine particulate matter. Accurately representing these processes is difficult due to the high dimensionality, stiffness, and partial observability of atmospheric chemical systems, which complicate the estimation of reliable reaction rate coefficients. In this work, I develop a Neural ODE–based framework that integrates a physics-informed chemical reaction neural network, enabling stable and interpretable estimation of rate coefficients in stiff systems. To quantify uncertainty in the inferred parameters, I employ a Bayesian Markov Chain Monte Carlo (MCMC) approach that explores the space of plausible rate coefficients and provides insight into parameter identifiability and variability.

Date & Time: [20th April 2026 | 5 PM (WAT)]

Venue: Online (Google Meet Link: meet.google.com/pjq-xson-qow )

• Rinel Foguen-Tchuendom, Bettermore Labs.