GitHub

Cloud and aerosol optics

Cloud and aerosol optics are separate stages: they fill their own containers, which are then composed onto gas optical properties before the solver runs. This keeps gas optics, cloud optics, and solvers independently testable.

Containers

  • CloudOptics stores grid-mean layer cloud optical depths: longwave absorption, shortwave absorption, shortwave scattering, and the shortwave scattering asymmetry factor.
  • CloudyRegionCloudOptics stores the same channels for the cloudy region only, with cloud_fraction and overlap_parameter carried separately. This matches the all-sky solver convention: cloud cover must not be encoded by weakening in-cloud optical depth.
  • AerosolOptics mirrors the cloud container for aerosols.

Layer models

LayerCloudOpticsModel converts a condensed water path CWP (kg m⁻²; scalar, per-layer vector, or read from atmosphere.cloud_water_path) with mass absorption/extinction coefficients (m² kg⁻¹):

\[\tau^\mathrm{lw} = \kappa^\mathrm{lw}\,\mathrm{CWP}, \qquad \tau^\mathrm{sw}_\mathrm{absorption} = (1 - \omega)\,\kappa^\mathrm{sw}\,\mathrm{CWP}, \qquad \tau^\mathrm{sw}_\mathrm{scattering} = \omega\,\kappa^\mathrm{sw}\,\mathrm{CWP}.\]

LayerLiquidIceCloudOpticsModel keeps liquid water path, ice water path, and cloud fraction separate, with per-phase coefficients, albedos, and asymmetry factors. It fills either container:

  • cloud_optical_properties! returns grid-mean optical depth (scaled by cloud fraction raised to cloud_fraction_exponent) — the shortcut for homogeneous-column smoke tests;
  • cloudy_region_optical_properties! fills cloudy-region optical depths without cloud-fraction scaling, plus the fraction and overlap fields, for the all-sky solvers.

LayerAerosolOpticsModel is the aerosol analogue driven by aerosol_path.

Composition onto gas optics

add_cloud_optical_depths! and add_aerosol_optical_depths! add the layer absorption channels to the gas absorption arrays (broadcast across g-points when the gas optics are spectral) and mix the scattering channel into the shortwave scattering optical depth, updating the asymmetry factor as a scattering-optical-depth-weighted mean. After composition, the shortwave solver sees a single absorption/scattering/asymmetry triple per layer and g-point, and needs no cloud-specific logic.

Scattering tables

For spectrally resolved cloud optics, the package reads ecRad's droplet and ice scattering tables and maps them onto ecCKD g-point grids:

  • CloudScatteringTable stores mass extinction, single-scattering albedo, and asymmetry factor as functions of wavenumber (cm⁻¹) and effective radius (m), read by read_cloud_scattering_table (requires the NCDatasets extension). The reference files are the Mie liquid-droplet and Baum ice tables from the pinned ecRad data artifact.
  • EcCKDSpectralMapping stores the resolved spectral intervals and the gpoint_fraction matrix from a CKD-definition file — ecCKD's fixed wavenumber-to-g mapping — read by read_ecckd_spectral_mapping. Interval weights are the solar spectral irradiance (shortwave) or a Planck weight (longwave).

cloud_scattering_properties interpolates a table linearly in effective radius at one wavenumber index. cloud_scattering_gpoint_properties produces per-g-point mass extinction, single-scattering albedo, and asymmetry: extinction is weight-averaged, single-scattering albedo extinction-weighted, and asymmetry scattering-weighted. Two mapping methods are available: :midpoint samples the nearest table wavenumber at each interval midpoint, while :ecrad reproduces ecRad's spectral averaging matrix; optional delta-Eddington and optically-thick averaging follow ecRad's conventions.

add_mapped_cloud_scattering! then puts per-g-point liquid and ice scattering into (Ng, Nz) shortwave optical properties from layer liquid/ice water paths and cloud fraction, with optional delta-Eddington forward-scattering scaling.