Skip to main content
SearchLoginLogin or Signup

Methods for Incorporating Model Uncertainty into Exoplanet Atmospheric Analysis

Presentation #627.04 in the session Planetary Atmospheres - Theory.

Published onApr 03, 2024
Methods for Incorporating Model Uncertainty into Exoplanet Atmospheric Analysis

A key goal of exoplanet spectroscopy is to measure atmospheric properties, such as abundances of chemical species, in order to connect them to our understanding of atmospheric physics and planet formation. In this new era of high-quality JWST data, it is paramount that these measurement methods are robust. When comparing atmospheric models to observations, multiple candidate models may produce reasonable fits to the data. Typically, conclusions are reached by selecting the best-performing model according to some metric. This ignores model uncertainty in favour of specific model assumptions, potentially leading to measured atmospheric properties that are overconfident and/or incorrect. In this talk, we will present a case study highlighting these effects. The hot Jupiter HD 209458b is one of the most extensively studied exoplanets, boasting one of the highest-quality spectra obtained with HST. Despite this, past efforts to measure the abundance of water vapor in its atmosphere have led to a range of different conclusions, with some studies finding that water is depleted relative to equilibrium expectations and others disputing this. We finally resolved this apparent contradiction using a new ensemble approach to atmospheric retrieval, whereby results from multiple models are combined in a statistically sound manner to account for model uncertainty. We showed that when a range of reasonable models are considered, the uncertainty on the measured water abundance ranges from depleted to slightly enhanced relative to equilibrium. In this case, the results are strongly influenced by the chosen approach to modeling clouds and hazes. As we start to analyse JWST data, it is critical that we do not fall into the same trap, and that we consider effects of model uncertainty to avoid overconfident and biased inferences.

No comments here