Assessing robustness and bias in 1D retrievals of 3D Global Circulation Models at high spectral resolution: a WASP-76 b simulation case study in emission

TL;DR

This study evaluates the reliability of 1D atmospheric retrievals for exoplanets using high-resolution spectroscopy, highlighting biases introduced by 3D atmospheric effects and the importance of model parameterization.

Contribution

It demonstrates how 1D retrievals can approximate 3D atmospheric conditions but are sensitive to spatial variations and model choices, informing future analysis strategies.

Findings

Retrievals broadly recover atmospheric conditions but are biased by spatial thermal gradients.

The choice of P-T and chemical profile parameterization affects retrieval accuracy.

Doppler offsets among opacity sources influence the interpretation of high-resolution spectra.

Abstract

High-resolution spectroscopy (HRS) of exoplanet atmospheres has successfully detected many chemical species and is quickly moving toward detailed characterization of the chemical abundances and dynamics. HRS is highly sensitive to the line shape and position, thus, it can detect three-dimensional (3D) effects such as winds, rotation, and spatial variation of atmospheric conditions. At the same time, retrieval frameworks are increasingly deployed to constrain chemical abundances, pressure-temperature (P-T) structures, orbital parameters, and rotational broadening. To explore the multidimensional parameter space, they need computationally fast models that are consequently mostly one-dimensional (1D). However, this approach risks introducing interpretation bias since the planet's true nature is 3D. We investigate the robustness of this methodology at high spectral resolution by running 1D retrievals on simulated observations in emission within an observational framework using 3D Global Circulation Models of the quintessential HJ WASP-76 b. We find that the retrieval broadly recovers conditions present in the atmosphere, but that the retrieved P-T and chemical profiles are not a homogeneous average of all spatial and phase-dependent information. Instead, they are most sensitive to spatial regions with large thermal gradients, which do not necessarily coincide with the strongest emitting regions. Our results further suggest that the choice of parameterization for the P-T and chemical profiles, as well as Doppler offsets among opacity sources, impact retrieval results. These factors should be carefully considered in future retrieval analyses.