Interpolations

class weatherbenchX.interpolations.Interpolation[source]

Interpolation base class.

class weatherbenchX.interpolations.MultipleInterpolation(interpolations: Sequence[Interpolation])[source]

Applies multiple interpolations to a dataset in sequence.

interpolations

List of interpolations to be applied in sequence.

Type:

Sequence[weatherbenchX.interpolations.Interpolation]

class weatherbenchX.interpolations.InterpolateToFixedCoords(method: str, coords: Mapping[str, DataArray | ndarray], wrap_longitude: bool = False, extrapolate_out_of_bounds: bool = True)[source]

Interpolate to a fixed set of coordinates.

Interplation is done using xarray’s built-in interp method: https://docs.xarray.dev/en/latest/generated/xarray.DataArray.interp.html

Init.

Parameters:

  • method – Interpolation method to be passed to xarray’s interpolation API.

  • coords – Dictionary of coordinate names and values to interpolate to.

  • wrap_longitude – If True, perform a wrapped interpolation in the longitude dimension. Default: False

  • extrapolate_out_of_bounds – If True, extrapolate to out of bounds values using the chosen interpolation method. Default: True

class weatherbenchX.interpolations.InterpolateToReferenceCoords(method: str, dims: Sequence[str] | None = None, wrap_longitude: bool = False, clip_reference_coords: Iterable[str] | None = None, extrapolate_out_of_bounds: bool = True)[source]

Interpolate to a reference dataset.

Interplation is done using xarray’s built-in interp method: https://docs.xarray.dev/en/latest/generated/xarray.DataArray.interp.html

Init.

Parameters:

  • method – Interpolation method to be passed to xarray’s interpolation API.

  • dims – (Optional) Dimensions over which to interpolate. If None (default), infer dimensions from intersect of DataArray dimensions and reference coordinates.

  • wrap_longitude – If True, perform a wrapped interpolation in the longitude dimension. Default: False

  • clip_reference_coords – Clip the reference dataset to the maximum extent of the data to be interpolated in the given dimensions, e.g. [‘latitude’, ‘longitude’]. Note that this can potentially lead to errors in the reference go unnoticed. It is preferred to use a fixed interpolation instead or ensure that the reference extent matches beforehand. Default: None.

  • extrapolate_out_of_bounds – If True, extrapolate to out of bounds values using the chosen interpolation method. Default: True

class weatherbenchX.interpolations.GridToSparseWithAltitudeAdjustment(method: str, grid_elevation: DataArray, dims: Sequence[str] | None = None, wrap_longitude: bool = False, extrapolate_out_of_bounds: bool = True, max_alititude_diff_in_m: float = 1500)[source]

Applies altitude adjustment to 2m_temperature and 10m_wind_speed.

Alititude adjustments are based on the difference of the grid elevation to the station elevation. Reference: https://rmets.onlinelibrary.wiley.com/doi/10.1002/qj.2372, Section 3.3.

Assumes that elevations are in meters and an ‘elevation’ coordinate exists on the reference dataset. Requires passing a DataArray with the grid elevation corresponding to the dataset to be interpolated. Variables must be named ‘2m_temperature’ and ‘10m_wind_speed’. Other variables will be left unchanged.

Init.

Parameters:

  • method – Interpolation method to be passed to xarray’s interpolation API.

  • grid_elevation – DataArray matching the dataset coordinates specifying the grid box elevation in m.

  • dims – (Optional) Dimensions over which to interpolate. If None (default), infer dimensions from intersect of DataArray dimensions and reference coordinates.

  • wrap_longitude – If True, perform a wrapped interpolation in the longitude dimension. Default: False

  • extrapolate_out_of_bounds – If True, extrapolate to out of bounds values using the chosen interpolation method. Default: True

  • max_alititude_diff_in_m – No adjustment is applied for elevation differences greater than this value. Large values can appear because of errors in the station dataset, e.g. elevation reported in ft instead of m. Default: 1500.

class weatherbenchX.interpolations.NeighborhoodThresholdProbabilities(neighborhood_sizes, thresholds, threshold_dim='threshold_value', wrap_longitude: bool = False)[source]

Converts a deterministic forecast to a probabilistic one by neighborhood averaging.

For a given threshold, the probability is devined as the fraction of the fraction of pixels in a square neighborhood that exceeds the threshold. This is the same computation as in the Fraction Skill Score.

Init.

Parameters:

  • neighborhood_sizes – List of neighborhood sizes to be used in pixels. Must be odd.

  • thresholds – List of thresholds to be used to binarize data.

  • threshold_dim – Dimension name of the thresholds. Default: ‘threshold_value’

  • wrap_longitude – If True, perform a wrapped convolution in the longitude dimension. Default: False