Barbosa, Sarah G. A., Estrela, Raissa, Filho, Paulo C. F. da Silva, Mugnai, Lorenzo V. ORCID: https://orcid.org/0000-0002-9007-9802 and de Freitas, Daniel B.
2026.
Towards the Habitable Worlds Observatory: retrieval of reflection spectra from evolving earth analogues using 1D CNNS.
RAS Techniques and Instruments
, rzag084.
10.1093/rasti/rzag084
|
|
PDF
- Accepted Post-Print Version
Available under License Creative Commons Attribution. Download (3MB) |
Abstract
The Habitable Worlds Observatory (HWO) will characterize Earth-like exoplanets from low-resolution reflected-light spectra. To assess how atmospheric evolution shapes what can be recovered, we trained a convolutional neural network on 109,945 spectra generated with the Planetary Spectrum Generator for Archean, Proterozoic, and Modern atmospheres under LUVOIR-B and HabEx/SS observing assumptions, separating the contributions of observational noise and model variability. We find that recovery depends jointly on abundance and spectral structure. Across eras, O2 recovery weakens in Proterozoic atmospheres and CH4 becomes reliable only when abundant, while an isolated UV band yields robust Modern O3 recovery (R2 ≃ 0.98) despite poor Modern CH4 recovery (R2 = 0.32-0.35). The wavelength analysis reinforces this interpretation. Without being supplied feature locations, the network assigns 58-71% of total attribution to visible wavelengths, while the ultraviolet remains decisive for O3 and the near-infrared for CH4 and CO2. Aggregate retrieval leverage declines beyond approximately 1.7 μm, supporting a practical long-wavelength boundary in this setup. These wavelength dependencies also shape detectability. At baseline noise, 83-85% of Modern O2 cases satisfy a 3σ criterion, while Proterozoic O2 remains below 50% when noise is halved and Modern CH4 remains below 3σ as observational noise approaches zero because model uncertainty caps its significance. Together, these results imply molecule-specific priorities for HWO trade studies. Low-abundance O2 retains considerable leverage from improved observational quality while motivating better inference, whereas at the 1,000-hour benchmark O3 and Modern CH4 show predominantly model-associated uncertainty in these networks, motivating broader training coverage and improved uncertainty calibration.
| Item Type: | Article |
|---|---|
| Date Type: | Published Online |
| Status: | In Press |
| Schools: | Schools > Physics and Astronomy Schools > Physical, Chemical & Environmental Sciences |
| Publisher: | Oxford University Press |
| Date of First Compliant Deposit: | 5 November 2026 |
| Date of Acceptance: | 14 October 2026 |
| Last Modified: | 05 Nov 2026 15:15 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/190702 |
Actions (repository staff only)
![]() |
Edit Item |





Altmetric
Altmetric