Exploring and Validating Exoplanet Atmospheric Retrievals with Solar System Analog Observations

Technical Summary: Exploring and Validating Exoplanet Atmospheric Retrievals with Solar System Analog Observations

Problem Statement
Atmospheric retrieval (inverse modeling) is essential for interpreting exoplanet observations, yet these tools are often validated only against synthetic data or theoretical benchmarks. A critical gap exists in validating these models against real-world data where the "truth" is known. Solar System worlds offer a unique opportunity to serve as analogs for exoplanets, providing high-quality ground- or orbiter-derived constraints that can validate retrieval models and test the simplifying parameterizations inherent in these tools. However, relatively few studies have systematically applied exoplanet retrieval techniques to Solar System observations to bridge the gap between planetary science and exoplanet characterization.

Methodology
The authors developed and validated a versatile, open-source atmospheric retrieval suite named rfast. Designed originally for exoplanet mission concept studies (specifically HabEx and LUVOIR), the tool is computationally efficient and written primarily in Python.

• Forward Models: rfast includes three primary forward modeling capabilities:
  1. Reflected Light: Supports both 1D (single plane-parallel scene) and 3D (pixelated globe) treatments. The 1D model uses a two-stream flux-adding approach, while the 3D model incorporates phase-angle dependence via Hapke scattering and Gauss-Chebyshev integration.
  2. Thermal Emission: Utilizes a similar flux-adding approach to the reflected light model, incorporating Planck function source terms and surface emissivity.
  3. Transit Spectroscopy: Computed in the geometric limit using a 1D path distribution approach, accounting for slant path optical depths and refractive floors.
• Radiative Transfer: Molecular opacities are derived from the HITRAN database (via LBLABC) or the Freedman et al. database, spanning 0.1–100 µm. The tool supports vertically varying gas and temperature profiles.
• Retrieval Engine: The suite employs the emcee Markov chain Monte Carlo sampler to explore posterior distributions for over 20 parameters, including gas mixing ratios, surface pressure, temperature profiles, cloud properties, and planetary radii.
• Validation Strategy: The authors validated rfast against:
  • Analytic solutions (Hunt 1973; Madhusudhan & Burrows 2012; Heng et al. 2021).
  • High-fidelity radiative transfer models (SMART and scaTran).
  • Previous retrieval results for exo-Earth scenarios (Feng et al. 2018).
• Application to Analog Data: Following validation, rfast was applied to three distinct Solar System datasets treated as exoplanet analogs:
  1. EPOXI Earth: Disk-integrated reflected light observations (UV to NIR) from the Deep Impact flyby.
  2. MGS-TES Earth: Disk-integrated thermal infrared spectrum (7–40 µm) from the Mars Global Surveyor.
  3. Cassini/VIMS Titan: Effective transit spectrum derived from stellar occultations (0.88–5 µm).

Key Contributions and Results

1. Model Validation:

   • rfast demonstrates strong agreement with high-fidelity models (SMART) for single-scene reflected light, thermal emission, and transit spectra, with runtimes 1,300–5,000 times faster.
   • Discrepancies arise in 3D reflected light models (up to ~10% bias in geometric albedo) due to simplifications in diffuse scattering treatments. However, the authors note that for mission concept studies using synthetic data generated by rfast, these systematic errors cancel out, making the tool suitable for feasibility studies.
   • Retrieval comparisons with Feng et al. (2018) show that 1D single-scene models can recover constraints comparable to 3D models for moderate Signal-to-Noise Ratios (SNR), suggesting 3D models are not strictly necessary for initial mission design trade-offs.

2. EPOXI Earth (Reflected Light):

   • At a V-band SNR of 20, the retrieval successfully constrained gas mixing ratios for O₂, H₂O, CO₂, O₃, and CH₄, as well as cloud properties.
   • Key Finding: Methane and CO₂ constraints relied heavily on spectral features beyond 1.8 µm.
   • SNR Sensitivity: At SNR 10, the retrieval failed to distinguish between a realistic Earth-like state and a "deep atmosphere" solution with near-global cloud coverage and artificially low gas mixing ratios. This suggests future exo-Earth missions may require SNR > 10 to avoid degenerate solutions.

3. MGS-TES Earth (Thermal Emission):

   • At SNR 20 (at 10 µm), the retrieval achieved strong constraints on biosignature gases (H₂O, CO₂, O₃, N₂O, CH₄), with 16/84 percentile ranges smaller than one order of magnitude.
   • Key Finding: The inferred surface temperature was constrained to be above the freezing point of water at 3-σ confidence.
   • Biases: Surface pressure was biased high (5x) and planetary radius biased low (10%). The inferred tropospheric lapse rate was unrealistically shallow. The authors attribute these biases to systematic calibration issues, the treatment of pressure-induced absorption, and the use of a single-cloud model. Introducing a multi-cloud model improved the fit to the thermal structure.

4. Cassini/VIMS Titan (Transit):

   • Applied to a transit spectrum derived from occultation data, the retrieval constrained gas mixing ratios (CO, CH₄, C₂H₂, C₃H₈) with uncertainties better than 0.5 dex, consistent with orbiter-derived results.
   • Key Finding: Including N₂-N₂ collision-induced absorption was necessary to reproduce the continuum near 4.3 µm, preventing erroneous haze opacity interpretations.
   • Haze Modeling: While a C-H stretch feature (3.4 µm) was modeled, it was not required to fit the data at the adopted noise level (0.1 ppm), suggesting that at current JWST-like noise levels, such features may not be strictly necessary for fitting Titan-like spectra.

Significance and Claims
The paper claims that Solar System analog observations provide a "pathway forward" for validating exoplanet retrieval tools. The authors assert that:

• Validation: Using Solar System data allows for the verification of exoplanet models and parameterizations against known "truth," a capability unavailable for true exoplanet observations.
• Mission Design: The rfast tool enables rapid exploration of how observational parameters (SNR, spectral coverage) map to atmospheric constraints, aiding the design of future missions like HabEx, LUVOIR, Roman, and Ariel.
• Cautionary Tales: The study highlights that lower SNR data can lead to degenerate solutions (e.g., deep atmosphere vs. shallow atmosphere) and that simplified cloud treatments can bias thermal structure inferences.
• Future Utility: The authors encourage future planetary science missions to acquire data specifically designed as exoplanet analogs to further refine these validation techniques.

The work concludes that while Solar System analogs are currently limited in number, they are indispensable for verifying the robustness of the inverse modeling tools that will interpret data from the James Webb Space Telescope (JWST) and future direct imaging missions.