pycif.plugins.transforms.complex.satellites — API reference

pycif.plugins.transforms.complex.satellites — API reference#

Configuration reference: satellites plugin

pycif.plugins.transforms.complex.satellites.adjoint.adjoint(transf, inout_datastore, controlvect, obsvect, mapper, di, df, mode, runsubdir, workdir, onlyinit=False, **kwargs)[source]#

De-aggregate total columns to the model level.

pycif.plugins.transforms.complex.satellites.apply_AK.apply_ak(sim_ak, aks, pwgt, drycols, qa0, chosen_level=-1, use_drycols=False, scale_factor=1, log_space=False, normalize_columns=False)[source]#

Apply averaging kernels to simulated concentrations (forward pass).

Computes the satellite-equivalent observation from the vertically interpolated model field sim_ak, the averaging kernels aks, pressure weights pwgt, dry-air columns drycols, and prior profile qa0. The exact formula depends on the keyword flags:

Normal space (log_space = False):

\[y = \sum_i \left[(sim\_ak_i \cdot drycols_i - qa0_i) \cdot aks_i \cdot pwgt_i\right] + y_0^{prior}\]

Log space (log_space = True):

\[y = \exp\left[\sum_i \left[(\log sim\_ak_i - \log qa0_i) \cdot aks_i \cdot pwgt_i\right] + \log y_0^{prior}\right]\]

Optional normalisations and unit scaling are applied afterwards.

Parameters:

  • sim_ak (np.ndarray) – simulated concentrations on satellite pressure levels, shape (nlevsat, nobs).

  • aks (np.ndarray) – averaging kernels, shape (nlevsat, nobs).

  • pwgt (np.ndarray) – pressure weights, shape (nlevsat, nobs).

  • drycols (np.ndarray) – dry-air column factors, shape (nlevsat, nobs).

  • qa0 (np.ndarray) – prior profiles, shape (nlevsat, nobs).

  • chosen_level (int) – level index for partial-column products (formula 3/8); -1 means use the pressure-weighted sum.

  • use_drycols (bool) – normalise by dry-air columns instead of pressure weights.

  • scale_factor (float) – multiplicative unit scaling factor.

  • log_space (bool) – apply AKs in log space (MOPITT-style).

  • normalize_columns (bool) – normalise the total column by the summed AK/pressure-weight product.

Returns:

satellite-equivalent observation, shape (nobs,).

Return type:

np.ndarray

pycif.plugins.transforms.complex.satellites.apply_AK.apply_ak_tl(sim_ak_tl, sim_ak, aks, pwgt, drycols, qa0, chosen_level=-1, use_drycols=False, scale_factor=1, log_space=False, normalize_columns=False)[source]#

Apply averaging kernels to concentration increments (tangent-linear pass).

The tangent-linear of apply_ak() with respect to the simulated concentrations sim_ak. Computes:

  • Normal space: \(y_{TL} = \sum_i sim\_ak^{TL}_i \cdot drycols_i \cdot aks_i \cdot pwgt_i\)

  • Log space: \(y_{TL} = y_{fwd} \sum_i (sim\_ak^{TL}_i / sim\_ak_i) \cdot aks_i \cdot pwgt_i\)

Parameters:

  • sim_ak_tl (np.ndarray) – TL of simulated concentrations on satellite pressure levels, shape (nlevsat, nobs).

  • sim_ak (np.ndarray) – forward simulated concentrations (needed in log space), shape (nlevsat, nobs).

  • aks

  • pwgt

  • drycols

  • qa0

  • chosen_level

  • use_drycols

  • scale_factor

:param : :param log_space: same as apply_ak(). :param normalize_columns: same as apply_ak().

Returns:

TL of the satellite-equivalent observation, shape (nobs,).

Return type:

np.ndarray

pycif.plugins.transforms.complex.satellites.apply_AK.apply_ak_ad(obs_incr, sim_ak, aks, pwgt, drycols, qa0, chosen_level=-1, use_drycols=False, scale_factor=1, log_space=False, normalize_columns=False)[source]#

Apply averaging kernels to propagate observation sensitivities (adjoint pass).

The adjoint of apply_ak() w.r.t. the simulated concentrations. Maps the scalar observation sensitivity obs_incr back to level-resolved concentration sensitivities:

  • Normal space: \(s_i = obs\_incr \cdot scale \cdot drycols_i \cdot aks_i \cdot pwgt_i / norm\)

  • Log space: \(s_i = obs\_incr \cdot scale \cdot aks_i \cdot pwgt_i / sim\_ak_i / norm \cdot y_{fwd}\)

Parameters:

  • obs_incr (np.ndarray) – observation-space sensitivity (adj_out), shape (nobs,).

  • sim_ak (np.ndarray) – forward simulated concentrations on satellite levels, shape (nlevsat, nobs).

  • aks

  • pwgt

  • drycols

  • qa0

  • chosen_level

  • use_drycols

  • scale_factor

:param : :param log_space: same as apply_ak(). :param normalize_columns: same as apply_ak().

Returns:

adjoint of simulated concentrations on satellite levels, shape (nlevsat, nobs).

Return type:

np.ndarray

pycif.plugins.transforms.complex.satellites.flushrun.flushrun(self, rundir, mode, transform_id, full_flush=True)[source]#

Cleaning the simulation directories to limit space usage

pycif.plugins.transforms.complex.satellites.forward.forward(transf, inout_datastore, controlvect, obsvect, mapper, di, df, mode, runsubdir, workdir, onlyinit=False, save_debug=False, **kwargs)[source]#

Aggregate simulations at the grid scale to total columns. Re-interpolate the model pressure levels to the satellite averaging kernel levels. Average using the averaging kernel formula

pycif.plugins.transforms.complex.satellites.vinterp.vertical_interp(pres_in, dpres_in, pres_out, dpres_out, cropstrato, vinterp_type='weight', weights_nsubsteps=20)[source]#

Compute the interpolation coefficients to apply a linear vertical interpolation