Presentation #1010 in the session “Open Engagement Session A”.
Crossing into the next frontier in exoplanet science will require development of quantitative methods that permit reliably characterizing exoplanet atmospheres despite the unavoidable limitations of the low resolution and low s/n data available to current and near-future technology. Exoplanet atmospheres include hundreds of molecules interacting in complex reaction networks, where most of the details of specific molecular species and reaction rates are not known — a problem made more challenging when considering the influence of unknown biology. This combination of observational and model uncertainty creates a unique challenge for exoplanet science, particularly when attempting to infer planetary properties such as degree of disequilibria or the presence of life. Here we demonstrate how the combination of complex systems and atmospheric science can provide a new synthesis for statistically assessing features of exoplanetary atmospheres. We generated ensembles of hot jupiter atmospheres simulated over a wide range of temperatures and metallicities, to produce likelihood distributions of possible atmospheric models centered around a specific observable such as T or metallicity. We then calculated measures of a variety of inferred properties, including abundances of gases, network structure of the atmospheric chemistry and thermodynamic properties and determined which sets of properties most reliably predicted the state of thermochemical disequilibria via machine learning. We discuss implications for the longer-term goal of inferring the presence of life as a driver of atmospheric disequilibrium on terrestrial worlds.