MultiHub Forum

I'm an astronomy graduate student working on characterizing exoplanet atmospheres using transmission spectroscopy data from the James Webb Space Telescope, and I'm currently stuck on modeling the atmospheric retrieval for a specific hot Jupiter where the data suggests a possible cloud deck but the signal-to-noise ratio makes it difficult to distinguish between cloud composition and a high metallicity atmosphere. My models keep hitting degeneracies where different combinations of parameters fit the observed spectrum equally well, which is frustrating when trying to draw concrete conclusions about the planet's formation history. For others in the field, what retrieval frameworks or statistical approaches have you found most effective for breaking these parameter degeneracies in low to moderate signal-to-noise data? How do you prioritize which molecular features to focus on when the spectrum is noisy, and what are the best practices for quantifying and reporting the uncertainties in your derived atmospheric properties?

Plan of attack for breaking degeneracies in JWST transmission spectra: choose a flexible forward model (e.g., TauREx3, petitRADTRANS) and couple it with robust Bayesian inference (nested sampling with dynesty or UltraNest). Treat clouds with a flexible parameterization: cloud top pressure, optical depth, and optional patchy coverage; keep chemistry as either free (free-chemistry retrieval with individual molecule abundances) or chemically consistent (thermochemical equilibrium with C/O, metallicity priors). Run two parallel retrievals (free vs chemically constrained) and compare via Bayesian evidence to assess degeneracy strength. Use joint retrieval across all available bands to maximize constraints, and include a simple, physically motivated priors (e.g., mass-based metallicity expectations).