Skip to content

Geotools

GeographicField

Bases: Field

geo_metadata = GeographicMetadata(cast('Metadata', field.metadata()), geography) instance-attribute

clone(*, values=None, metadata=None, **kwargs)

message()

Delegate message generation to the underlying field.

to_latlon(flatten=True, dtype=None, index=None)

Return the latitudes and longitudes for the field.

We take these from the geography as GeographicField might be a wrapper around a differently-sized field. This is done, for example, in ReprojectFilter.

to_numpy(flatten=False, dtype=None, index=None)

Delegate creation of numpy array to the underlying field.

It is important to do this rather than relying on the base class implementation since self._field might override the base class method.

GeographicGrid

Bases: Geography

resolution_ = resolution instance-attribute

native_crs = native_crs instance-attribute

latitudes_ = None instance-attribute

longitudes_ = None instance-attribute

latitudes(dtype=None)

longitudes(dtype=None)

x(dtype=None)

y(dtype=None)

bounding_box()

mars_area()

mars_grid()

resolution()

shape()

gridspec()

grid_spec()

projection()

regrid_forecast_run(predictions, target_latitudes, target_longitudes, first_valid_day, *, variable='ice_conc', output_variable='siconc', source_crs=None)

Regrid a single forecast run onto a regular lat/lon grid.

Selects the forecast run whose first lead time is valid on first_valid_day and bilinearly interpolates it from its native grid onto the target lat/lon grid, cropping to the target extent. The interpolation is done in the CRS in which the native grid is regular, which is EASE-Grid 2.0 unless source_crs is given. The output follows the CMEMS layout, with dimensions (time, latitude, longitude) and a daily (00:00) time axis.

If the predictions have a land mask, land cells are left out of the interpolation, so that they do not pull coastal values towards zero. Target points that are mostly land are set to NaN, and the land mask is regridded into the output.

Parameters:

Name Type Description Default
predictions Dataset

Predictions with dimensions (forecast_reference_time, lead_time, y, x), 2-D latitude/longitude coordinates and a valid_time coordinate, as written by the prediction writer. Unless source_crs is given, it must also have a "hemisphere" attribute of "north" or "south". It may have a (y, x) "land_mask" variable, with 1 for ocean and 0 for land.

 required
target_latitudes DataArray

1-D latitudes of the target grid.

 required
target_longitudes DataArray

1-D longitudes of the target grid.

 required
first_valid_day str | datetime64

Day on which the first lead time of the selected run is valid, e.g. "2024-01-14".

 required
variable str

Name of the variable to regrid in predictions.

 'ice_conc'
output_variable str

Name of the regridded variable in the output.

 'siconc'
source_crs str | None

CRS in which the prediction grid is regular, e.g. "EPSG:3413". Defaults to the EASE-Grid 2.0 CRS for the predictions' hemisphere.

 None

Returns:

Type Description
Dataset

Dataset containing output_variable with dimensions (time, latitude,
Dataset

longitude), where time holds the valid day of each lead time. The latitude
Dataset

and longitude coordinates have CF attributes, overridden by any attributes
Dataset

of target_latitudes and target_longitudes. The variable
Dataset

keeps the attributes of variable in predictions, plus the CF attributes
Dataset

in CF_VARIABLE_ATTRIBUTES for known variables, and is clipped to its valid
Dataset

range where one is defined. If the predictions have a land mask, the
Dataset

output also contains a (latitude, longitude) "land_mask" (1 for ocean, 0
Dataset

for land, missing outside the prediction grid), linked from
Dataset

output_variable as an ancillary variable.

Raises:

Type Description
ValueError

If the target latitudes or longitudes are not 1-D, if no run has its first lead time valid on first_valid_day, if source_crs is not given and the hemisphere attribute is missing or invalid, or if the prediction grid is not regular in the source CRS.

nearest_neighbour_indices(input_latlons, output_latlons)

Calculate the nearest neighbour input cell for each cell in the output grid.

Parameters:

Name Type Description Default
input_latlons ArrayHWV

Array of shape [input_h, input_w, 2] containing the latitudes and longitudes of each cell in the input grid.

 required
output_latlons ArrayHWV

Array of shape [output_h, output_w, 2] containing the latitudes and longitudes of each cell in the output grid.

 required

Returns:

Type Description
ArrayIndices2D

Tuple of (nn_indices_h, nn_indices_w) where each is a numpy array of shape
ArrayIndices2D

[output_height, output_width] containing, for each output cell, the index in
tuple[ArrayIndices2D, ArrayIndices2D]

the H and W dimensions of the nearest neighbour input cell that should be
tuple[ArrayIndices2D, ArrayIndices2D]

used as the source.