Zero-shot learning of aerosol optical properties with graph neural networks.

Black carbon (BC), a strongly absorbing aerosol sourced from combustion, is an important short-lived climate forcer. BC's complex morphology contributes to uncertainty in its direct climate radiative effects, as current methods to accurately calculate the optical properties of these aerosols are too computationally expensive to be used online in models or for observational retrievals. Here we demonstrate that a Graph Neural Network (GNN) trained to predict the optical properties of numerically-generated BC fractal aggregates can accurately generalize to arbitrarily shaped particles, including much larger ([Formula: see text]) aggregates than in the training dataset. This zero-shot learning approach could be used to estimate single particle optical properties of realistically-shaped aerosol and cloud particles for inclusion in radiative transfer codes for atmospheric models and remote sensing inversions. In addition, GNN's can be used to gain physical intuition on the relationship between small-scale interactions (here of the spheres' positions and interactions) and large-scale properties (here of the radiative properties of aerosols).

Pyrocumulonimbus affect average stratospheric aerosol composition.

Pyrocumulonimbus (pyroCb) are wildfire-generated convective clouds that can inject smoke directly into the stratosphere. PyroCb have been tracked for years, yet their apparent rarity and episodic nature lead to highly uncertain climate impacts. In situ measurements of pyroCb smoke reveal its distinctive and exceptionally stable aerosol properties and define the long-term influence of pyroCb activity on the stratospheric aerosol budget. Analysis of 13 years of airborne observations shows that pyroCb are responsible for 10 to 25% of the black carbon and organic aerosols in the "present-day" lower stratosphere, with similar impacts in both the North and South Hemispheres. These results suggest that, should pyroCb increase in frequency and/or magnitude in future climates, they could generate dominant trends in stratospheric aerosol.

Nonperturbative retrieval of the scattering strength in one-dimensional media.

We examine several approaches on how to use the transmission and reflection amplitudes as functions of the modulation frequency of the laser's intensity to reconstruct the position-dependent scattering coefficient for a simple turbid medium. We explore the region where the contrast between the coefficient and its spatially averaged value is large enough such that perturbative methods fail. We show that in the case of a transillumination geometry, the knowledge of the transmission profile alone is not sufficient for unique image reconstruction, whereas the reflection spectrum allows for a complete inversion. We demonstrate the invertibility for media sampled at only a few positions.