Particles (Prt)#
Prt-Core v1 [prt_core_v1]#
This data format is derived from libRadtran’s aerosol data format. It features an adaptive scattering angle (θ) grid that allows for an optimal positioning of samples depending on the wavelength.
The number of angular samples can vary across wavelengths up to a maximum equal
to the size of the iangle dimension. The per-wavelength count is stored in
nangles; entries beyond nangles[iw] in theta, mu, and phase
are NaN-padded.
Similarly, the number of Legendre coefficients can vary across wavelengths up to
a maximum equal to the size of the imom dimension. When pmom is present,
nmom stores the per-wavelength count; entries beyond nmom[iw] in
pmom are NaN-padded.
The format can be extended with state dimensions (see Extensions).
- Format
xarray.Dataset(in-memory), NetCDF (storage)- Dimensions
w: radiation wavelengthphamat: nonzero coefficients in the phase matrixiangle: angular data pointsimom: Legendre coefficients for the (1,1) phase matrix component
- Coordinates
When relevant, units are required and specified in the “units” metadata field.
w(w)float [length]: wavelengthphamat(phamat)str [—]: row and column indices of the phase matrix coefficient (e.g. “11”, “12”, etc.)theta(w, iangle)float [angle]: scattering angle θ; NaN-padded beyondnangles[iw]mu(w, iangle)float [—]: value of cos θ; NaN-padded beyondnangles[iw]
- Data variables
When relevant, units are required and specified in the “units” metadata field.
ext(w)float [any]: extinction coefficient per unit amount (see Extinction units)ssa(w)float [—]: single-scattering albedophase(phamat, w, iangle)float [1 / sr]: value of the phase matrix; NaN-padded beyondnangles[iw]nangles(w)int [—]: number of valid angular samples per wavelength; must be consistent with the NaN padding oftheta,mu, andphasepmom(w, imom)float [—], optional: Legendre expansion coefficients of the (1,1) phase matrix element; NaN-padded beyondnmom[iw]nmom(w)int [—], required ifpmomis present: number of valid Legendre coefficients per wavelength; must be consistent with the NaN padding ofpmom
Note
pmomstores only the Legendre expansion of the (1,1) element \(p_{11}\). Legendre expansions of the other phase matrix components are not part of this format.- Normalization conventions
The (1,1) element of the phase matrix \(p_{11} (\mu)\) follows the 4π-normalized convention, normalizing the µ-parametrized phase function to 2:
\[\frac{1}{4\pi} \int_{4\pi} p_{11}(\Omega)\, \mathrm{d}\Omega = 1 \Longleftrightarrow \int_{-1}^{1} p_{11}(\mu)\, \mathrm{d}\mu = 2,\]where \(p_{11} (\mu)\) is in sr⁻¹.
The
pmomvariable stores the Legendre expansion coefficients \(p_l = (2l + 1) \xi_l\), where\[\xi_l = \frac{1}{2} \int_{-1}^{1} p_{11}(\mu)\, P_l(\mu)\, \mathrm{d}\mu,\]and \(P_l\) is the Legendre polynomial of degree \(l\). The phase function is thus recovered as
\[p_{11}(\mu) = \sum_{l=0}^{L} \mathtt{pmom}[l]\, P_l(\mu).\]For a normalized phase function, \(\mathtt{pmom}[0] = 1\) and \(\mathtt{pmom}[1] = 3g\), where \(g\) is the asymmetry parameter.
Note
Data are sorted in ascending order of
wandmuValid
phamatvalues are (in that order):["11", "12", "33", "34", "22", "44"]The number of
phamatvalues implicitly defines whether the data describes light polarization: 1 means without polarization, 2+ means with polarizationThe number of
phamatvalues implicitly defines the scattering particle shape: 1 or 4 means spherical particles, 6 means spheroidal particles
Eradiate |
libRadtran |
|
|---|---|---|
Spectral dimension |
|
|
Phase matrix coefficients |
|
|
Angular data points (maximum) |
|
|
Angular data points (per wavelength) |
|
|
Legendre coefficients |
|
|
See also
The full specification of the aer_core_v2 format:
Aer-Core v2 (legacy) [aer_core_v2]
Extinction units#
The ext variable holds the extinction coefficient per unit amount of
particles. Its units are not fixed by the format, but must be specified in the
“units” metadata field. The component that consumes the dataset sets the
constraint:
ParticleOTComponentonly uses the spectral ratio \(\mathtt{ext}(\lambda) / \mathtt{ext}(\lambda_\mathrm{ref})\), so any units are accepted.ParticleStateComponentmultipliesextby a profile amount variable (e.g. mass concentration in g/m³); their product must have the dimensionality of an inverse length (e.g.extin m²/g).
Extensions#
Prt-Core v1 can be extended with one or several state dimensions on which
optical properties are indexed (e.g. particle size distribution parameters,
relative humidity). State dimensions are inserted after w: every variable
and coordinate indexed by w (ext, ssa, phase, nangles,
pmom, nmom, mu, theta) gains the state dimensions, in the same
order, right after w.
Each state dimension d comes with a coordinate d(d) that must be
strictly increasing, but need not be regularly spaced. Its units are free, but
must be specified in the “units” metadata field (dimensionless if
relevant). Data are sorted in ascending order of each state coordinate.
The Prt-MS v1 [prt_ms_v1] and Prt-Hum v1 [prt_hum_v1]
formats are named extensions that fix the name, meaning and units of their
state dimensions. Datasets with other state dimensions are valid
extensions; they are created with make_prt_v1().
ParticleProperties loads datasets with any number of state
dimensions. Components set further constraints:
ParticleOTComponent requires a dataset without state dimension, and
ParticleStateComponent requires 1 or 2 state dimensions, which is the
number of state variables the particlephase kernel plugin interpolates
over.
Prt-MS v1 [prt_ms_v1]#
Named extension of prt_core_v1 for particle properties. This data format
indexes properties defined by prt_core_v1 on two state dimensions,
reff and veff, the effective radius and effective variance of the
particle size distribution, respectively. It shares the same normalization
conventions as prt_core_v1.
- Format
xarray.Dataset(in-memory), NetCDF (storage)- Dimensions
w: radiation wavelengthreff: effective radius of the particle size distributionveff: effective variance of the particle size distributionphamat: nonzero coefficients in the phase matrixiangle: angular data pointsimom: Legendre coefficients for the (1,1) phase matrix component
- Coordinates
When relevant, units are required and specified in the “units” metadata field.
w(w): float [length]: wavelengthreff(reff): float [length]: effective radius of the particle size distributionveff(veff): float [—]: effective variance of the particle size distributionphamat(phamat): str [—]: row and column indices of the phase matrix coefficient (e.g. “11”, “12”, etc.)theta(w, reff, veff, iangle): float [angle]: scattering angle θ; NaN-padded beyondnangles[iw, ireff, iveff]mu(w, reff, veff, iangle): float [—]: value of cos θ; NaN-padded beyondnangles[iw, ireff, iveff]
- Data variables
When relevant, units are required and specified in the “units” metadata field.
ext(w, reff, veff)float [any]: extinction coefficient per unit amount (see Extinction units)ssa(w, reff, veff)float [—]: single-scattering albedophase(phamat, w, reff, veff, iangle)float [1 / sr]: value of the phase matrix; NaN-padded beyondnangles[iw, ireff, iveff]nangles(w, reff, veff)int [—]: number of valid angular samples per wavelength, reff and veff; must be consistent with the NaN padding oftheta,mu, andphasepmom(w, reff, veff, imom)float [—], optional: Legendre expansion coefficients of the (1,1) phase matrix element; NaN-padded beyondnmom[iw, ireff, iveff]nmom(w, reff, veff)int [—], required ifpmomis present: number of valid Legendre coefficients per wavelength; must be consistent with the NaN padding ofpmom
Note
Data are sorted in ascending order of
w,reff,veffandmureffandveffmust be strictly increasing, but need not be regularly spacedphamatconventions (valid values, and their implication for polarization and particle shape) are the same asprt_core_v1
Prt-Hum v1 [prt_hum_v1]#
Named extension of prt_core_v1 for hygroscopic particles. This data format
indexes properties defined by prt_core_v1 on one state dimension,
hum, the ambient relative humidity. It is typically used for aerosol
components whose optical properties depend on humidity through water uptake
(e.g. OPAC components). It shares the same normalization conventions as
prt_core_v1.
- Format
xarray.Dataset(in-memory), NetCDF (storage)- Dimensions
w: radiation wavelengthhum: relative humidityphamat: nonzero coefficients in the phase matrixiangle: angular data pointsimom: Legendre coefficients for the (1,1) phase matrix component
- Coordinates
When relevant, units are required and specified in the “units” metadata field.
w(w): float [length]: wavelengthhum(hum): float [—]: relative humidity (e.g. in percent)phamat(phamat): str [—]: row and column indices of the phase matrix coefficient (e.g. “11”, “12”, etc.)theta(w, hum, iangle): float [angle]: scattering angle θ; NaN-padded beyondnangles[iw, ihum]mu(w, hum, iangle): float [—]: value of cos θ; NaN-padded beyondnangles[iw, ihum]
- Data variables
When relevant, units are required and specified in the “units” metadata field.
ext(w, hum)float [any]: extinction coefficient per unit amount (see Extinction units)ssa(w, hum)float [—]: single-scattering albedophase(phamat, w, hum, iangle)float [1 / sr]: value of the phase matrix; NaN-padded beyondnangles[iw, ihum]nangles(w, hum)int [—]: number of valid angular samples per wavelength and humidity; must be consistent with the NaN padding oftheta,mu, andphasepmom(w, hum, imom)float [—], optional: Legendre expansion coefficients of the (1,1) phase matrix element; NaN-padded beyondnmom[iw, ihum]nmom(w, hum)int [—], required ifpmomis present: number of valid Legendre coefficients per wavelength and humidity; must be consistent with the NaN padding ofpmom
Note
Data are sorted in ascending order of
w,humandmuhummust be strictly increasing, but need not be regularly spacedphamatconventions (valid values, and their implication for polarization and particle shape) are the same asprt_core_v1
Creation#
Datasets in all Prt variants are created with make_prt_v1(), which
inserts optional state dimensions (e.g. reff and veff, hum)
between w and the angular and moment dimensions. make_prt_core_v1(),
make_prt_ms_v1() and make_prt_hum_v1() are shortcuts for each
variant.
Conversion#
The following conversion components are provided:
Function |
Source format |
Target format |
|---|---|---|
libRadtran / MOPSMAP |
Prt-Core v1 [prt_core_v1], or
Prt-MS v1 [prt_ms_v1] / Prt-Hum v1 [prt_hum_v1]
when the output varies over |
|
Prt-MS v1 [prt_ms_v1] (several) |