def sw_correction(mgcaprediction, age, drawsfun=None):
"""Apply Deep-Time seawater correction to Mg/Ca prediction.
Parameters
----------
mgcaprediction : baymag.predict.MgCaPrediction
age : sequence-like
Age of predictions in ``prediction``. Must be in units Ma. n-length
sequence where n == prediction.ensemble.shape[0].
drawsfun : None or function-like, optional
Optional function-like returning 2d array of MCMC parameter draws to
use for seawater correction. Used for testing and debugging only.
Default ``None`` uses ``baymag.modelparams.get_sw_draws()``.
Returns
-------
out : baymag.MgCaPrediction
Copy of mgcaprediction with correction to ensemble.
"""
mgsw_smooth = get_mgsw_smooth()
t = int(age * 2)
mgsw = np.divide(mgsw_smooth[t, :], mgsw_smooth[0, :])
out = MgCaPrediction(ensemble=np.array(mgcaprediction.ensemble * mgsw),
spp=str(mgcaprediction.spp))
return out
def predict_mgca_ln_dt(age,
seatemp,
cleaning,
salinity,
ph,
omega,
spp,
drawsfun=get_draws):
"""Predict Mg/Ca from sea temperature for deep time (sw correction)
Parameters
----------
seatemp : ndarray
n-length array of sea temperature observations (°C) from a single
location.
cleaning : ndarray
Binary n-length array indicating the cleaning method used for the
inferred Mg/Ca series. ``1`` for reductive, ``0`` for BCP (Barker).
salinity : scalar or ndarray
Sea water salinity (PSU).
ph : scalar or ndarray
Sea water pH.
omega : scalar or ndarray
Sea water calcite saturation state.
spp : str
Calibration model parameter options. Must be one of:
'all' : Pooled calibration using annual SSTs.
'all_sea' : Pooled calibration using seasonal SSTs.
'ruber' : Hierarchical calibration with G. ruber (white or pink).
'bulloides' : Hierarchical calibration with G. bulloides.
'sacculifer' : Hierarchical calibration with G. sacculifer.
'pachy' : Hierarchical calibration with N. pachyderma or N. incompta.
drawsfun : function-like, optional
For debugging and testing. Object to be called to get MCMC model
parameter draws. Don't mess with this.
Returns
-------
out : MgCaPrediction
See Also
--------
fetch_omega : Calculate modern insitu calcite saturation state (omega)
fetch_ph : Fetch modern seawater surface insitu pH
sw_correction : Apply Deep-Time seawater correction to Mg/Ca predictions
MatLab code created by Dr. Jessica Tierney, The University of Arizona (2019)
Python code by S. Brewster Malevich, The University of Arizona
Modified by Robert Tardif, University of Washington
Modified by Mingsong Li Penn State (2020)
Modified by Mingsong Li, Peking University (May 10, 2021)
sw correction
"""
seatemp = np.atleast_1d(seatemp)
cleaning = np.atleast_1d(cleaning)
salinity = np.atleast_1d(salinity)
ph = np.atleast_1d(ph)
nlens = np.size(seatemp)
# Invert omega for model.
omega = omega**-2
omega = np.atleast_1d(omega)
alpha, beta_temp, beta_salinity, beta_omega, beta_ph, beta_clean, sigma = drawsfun(
spp)
clean_term = (1 - beta_clean * cleaning[:, np.newaxis])
if spp in ['all', 'all_sea']:
sigma = np.swapaxes(np.tile(sigma, nlens), 0, 1)
alpha = np.swapaxes(np.tile(alpha, nlens), 0, 1)
mu = (alpha + beta_temp * seatemp[:, np.newaxis] +
beta_omega * omega[:, np.newaxis] +
beta_salinity * salinity[:, np.newaxis] + clean_term)
# if spp other than pachy or sacculifer, take sensitivity to pH into account
if spp not in ['pachy', 'sacculifer']:
mu += beta_ph * ph[:, np.newaxis]
# sw correction (Mingsong Li, Peking University)
mgsw_smooth = get_mgsw_smooth()
t = int(age * 2)
mgsw = np.log(np.divide(mgsw_smooth[t, :], mgsw_smooth[0, :]))
mu += mgsw
mgca = np.random.normal(mu, sigma)
out = MgCaPrediction(ensemble=mgca, spp=spp)
return out