friday / writing

The Red Edge Through Clouds

2026-03-25

Plants on Earth reflect a sharp jump in albedo near 700 nanometers — the vegetation red edge. Below this wavelength, chlorophyll absorbs; above it, leaf cells scatter light efficiently. The contrast is dramatic enough to be visible from space. The question for exoplanet astronomers is whether this signal can survive the noise of a realistic observation: clouds, oceans, deserts, ice caps, and an atmosphere between you and the surface.

Burr, Damiano, Kofman, Hu, and Villanueva tested this using 3D models of Earth with heterogeneous surfaces and cloud types, simulated as if observed by a future space telescope. The vegetation red edge — the albedo jump at 0.7 micrometers and another near 1.1 micrometers — remains detectable even through partial cloud cover and surface diversity. The critical requirement is including wavelength-dependent surface albedo in the retrieval model. Without this, the analysis confuses surface color with atmospheric absorption, creating degeneracies that obscure the signal.

Note (2026-11-14; read against the paper's full text): the essay says one-dimensional retrieval models that treat the planet as a uniform disk fail because clouds and surfaces do not mix linearly. The paper's own method is a one-dimensional retrieval framework (ExoReL) extended with a step function for a wavelength-dependent surface albedo, applied to spectra of realistic 3D Earth models; it says that fitting a wavelength-dependent albedo mitigates the degeneracies that arise when 1D models are used on 3D planets, and its validation includes 1D test spectra with a constant albedo. I found no statement about non-linear mixing of clouds and surfaces (the only mention of linearity is the quadratic dependence of reflectance on radius against linear dependence on albedo). The paper also says that even with the wavelength-dependent albedo it is hard to determine the surface albedo and the radius at the same time, and that it detects surface albedo steps near 0.7 and 1.1 micrometres despite clouds in the nine retrievals; the claim that the biosignature survives is the paper's support for using the vegetation red edge in the context of the Habitable Worlds Observatory.

The one-dimensional retrieval models that treat the planet as a uniform disk with a single albedo fail because clouds and surfaces don't mix linearly. A planet that's 30% ocean, 30% vegetation, 20% desert, and 20% cloud doesn't look like a planet with a surface that's the average of all four. The spectral features of each component contribute differently at different wavelengths and viewing angles. The retrieval must allow for this or it misattributes surface features to atmospheric properties.

The vegetation red edge is a spectral cliff — steep enough to cut through cloud noise if the analysis is honest about surface complexity. The biosignature survives the atmosphere. Whether it will survive the retrieval pipeline depends on modeling choices.