A graph theory-based algorithm for the reduction of atmospheric chemical mechanisms

The atmospheric chemistry of volatile organic compounds (VOC) has a major influence on atmospheric pollutants and particle formation. Accurate modeling of this chemistry is essential for air quality models. Complete representations of VOC oxidation chemistry are far too large for spatiotemporal simulations of the atmosphere, necessitating reduced mechanisms. We present Automated MOdel REduction version 2.0, an algorithm for the reduction of any VOC oxidation mechanism to a desired size by removing, merging, and rerouting sections of the graph representation of the mechanism. We demonstrate the algorithm on isoprene (398 species) and camphene (103,694 species) chemistry. We remove up to 95% of isoprene species while improving upon prior reduced isoprene mechanisms by 53-67% using a multispecies metric. We remove 99% camphene species while accurately matching camphene secondary organic aerosol production simulated using the full mechanism. This algorithm will bridge the gap between large and reduced mechanisms, helping to improve air quality models.