argopy.extensions.CanyonB.predict

CanyonB.predict(params: str | List[str] = None, epres: float | None = None, etemp: float | None = None, epsal: float | None = None, edoxy: float | ndarray | None = None, include_uncertainties: bool | None = False, n_jobs: int | None = -1) → Dataset

Make predictions using the CANYON-B method.

This method implements the CANYON-B Bayesian neural network ensemble to estimate oceanic parameters from hydrographic data.

Parameters:

• params (str, list of str, or None, optional) –

  Parameter(s) to predict. Valid options:

  • ’AT’: Total alkalinity (umol/kg)

  • ’DIC’: Dissolved inorganic carbon (umol/kg)

  • ’pHT’: Total pH

  • ’pCO2’: Partial pressure of CO2 (uatm)

  • ’NO3’: Nitrate concentration (umol/kg)

  • ’PO4’: Phosphate concentration (umol/kg)

  • ’SiOH4’: Silicate concentration (umol/kg)

  If None (default), all seven parameters are predicted.

• epres (float, optional) – Pressure measurement uncertainty in dbar (default: 0.5 dbar)

• etemp (float, optional) – Temperature measurement uncertainty in degC (default: 0.005 degC)

• epsal (float, optional) – Salinity measurement uncertainty in PSU (default: 0.005)

• edoxy (float or np.ndarray, optional) – Oxygen measurement uncertainty in umol/kg. If not provided, defaults to 1% of measured oxygen values. Can be a scalar applied to all points or an array matching data dimensions.

• include_uncertainties (bool, optional) – If True, include uncertainty estimates for each predicted parameter

• n_jobs (int, optional) – Number of parallel jobs used for prediction (only used when there is more than one parameter to predict). Default is -1 (use all available CPUs). This option is directly passed to joblib.Parallel.

Returns:

Input dataset augmented with predicted parameters.

Return type:

xr.Dataset