def geocentric_cartesian(
cube: Cube, x_coords: ndarray, y_coords: ndarray
) -> ndarray:
"""
A function to convert a global (lat/lon) coordinate system into a
geocentric (3D trignonometric) system. This function ignores orographic
height differences between coordinates, giving a 2D projected
neighbourhood akin to selecting a neighbourhood of grid points about a
point without considering their vertical displacement.
Args:
cube:
A cube from which is taken the globe for which the geocentric
coordinates are being calculated.
x_coords:
An array of x coordinates that will represent one axis of the
mesh of coordinates to be transformed.
y_coords:
An array of y coordinates that will represent one axis of the
mesh of coordinates to be transformed.
Returns:
An array of all the xyz combinations that describe the nodes of
the grid, now in 3D geocentric cartesian coordinates. The shape
of the array is (n_nodes, 3), order x[:, 0], y[:, 1], z[:, 2].
"""
coordinate_system = cube.coord_system().as_cartopy_crs()
cartesian_calculator = coordinate_system.as_geocentric()
z_coords = np.zeros_like(x_coords)
cartesian_nodes = cartesian_calculator.transform_points(
coordinate_system, x_coords, y_coords, z_coords
)
return cartesian_nodes
def transform_grid_to_lat_lon(cube: Cube) -> Tuple[ndarray, ndarray]:
"""
Calculate the latitudes and longitudes of each points in the cube.
Args:
cube:
Cube with points to transform
Returns
lats:
Array of cube.data.shape of Latitude values
lons:
Array of cube.data.shape of Longitude values
"""
trg_latlon = ccrs.PlateCarree()
trg_crs = cube.coord_system().as_cartopy_crs()
x_points = cube.coord(axis="x").points
y_points = cube.coord(axis="y").points
x_zeros = np.zeros_like(x_points)
y_zeros = np.zeros_like(y_points)
# Broadcast x points and y points onto grid
all_x_points = y_zeros.reshape(len(y_zeros), 1) + x_points
all_y_points = y_points.reshape(len(y_points), 1) + x_zeros
# Transform points
points = trg_latlon.transform_points(trg_crs, all_x_points, all_y_points)
lons = points[..., 0]
lats = points[..., 1]
return lats, lons
def _load_data(self, array_path, group_dims, grid_mapping,
attr_name=None, separator='__', handle_nan=None):
"""
Create an Iris cube from a TileDB array describing a data variable and
pre-loaded dimension-describing coordinates.
TODO not handled here: aux coords and dims, cell measures, aux factories.
"""
single_attr_name = 'dataset'
if attr_name is None:
attr_metadata, lazy_data = self._from_tdb_array(array_path, single_attr_name,
to_dask=True, handle_nan=handle_nan)
metadata = attr_metadata
attr_name = metadata.pop(single_attr_name)
else:
attr_metadata, lazy_data = self._from_tdb_array(array_path,
single_attr_name,
array_name=attr_name,
to_dask=True,
handle_nan=handle_nan)
metadata = {}
for key, value in attr_metadata.items():
# Varname-specific keys are of form `keyname__attrname`; we only want `keyname`.
# TODO pass the separator character to the method.
try:
key_name, key_attr = key.split(separator)
if key_attr == attr_name:
metadata[key_name] = value
except ValueError:
# Not all keys are varname-specific; we want all of these.
metadata[key] = value
cell_methods = parse_cell_methods(metadata.pop('cell_methods', None))
dim_names = metadata.pop('dimensions').split(',')
# Dim Coords And Dims (mapping of coords to cube axes).
dcad = [(group_dims[name], i) for i, name in enumerate(dim_names)]
safe_attrs = self._handle_attributes(metadata,
exclude_keys=['dataset', 'multiattr', 'grid_mapping'])
std_name = metadata.pop('standard_name', None)
long_name = metadata.pop('long_name', None)
var_name = metadata.pop('var_name', None)
if all(itm is None for itm in [std_name, long_name, var_name]):
long_name = attr_name
cube = Cube(lazy_data,
standard_name=std_name,
long_name=long_name,
var_name=var_name,
units=metadata.pop('units', '1'),
dim_coords_and_dims=dcad,
cell_methods=cell_methods,
attributes=safe_attrs)
cube.coord_system = grid_mapping
return cube
def transform_grid_to_lat_lon(cube: Cube) -> Tuple[ndarray, ndarray]:
"""
Calculate the latitudes and longitudes of each points in the cube.
Args:
cube:
Cube with points to transform
Returns
lats:
Array of cube.data.shape of Latitude values
lons:
Array of cube.data.shape of Longitude values
"""
trg_latlon = ccrs.PlateCarree()
trg_crs = cube.coord_system().as_cartopy_crs()
cube = cube.copy()
# TODO use the proj units that are accesible with later versions of proj
# to determine the default units to convert to for a given projection.
# Assuming proj units of metre for all projections not in degrees.
for axis in ["x", "y"]:
try:
cube.coord(axis=axis).convert_units("m")
except ValueError as err:
msg = (
"Cube passed to transform_grid_to_lat_lon does not have an "
f"{axis} coordinate with units that can be converted to metres. "
)
raise ValueError(msg + str(err))
x_points = cube.coord(axis="x").points
y_points = cube.coord(axis="y").points
x_zeros = np.zeros_like(x_points)
y_zeros = np.zeros_like(y_points)
# Broadcast x points and y points onto grid
all_x_points = y_zeros.reshape(len(y_zeros), 1) + x_points
all_y_points = y_points.reshape(len(y_points), 1) + x_zeros
# Transform points
points = trg_latlon.transform_points(trg_crs, all_x_points, all_y_points)
lons = points[..., 0]
lats = points[..., 1]
return lats, lons
def different_projection(self, method, ancillary_data, additional_data,
expected, **kwargs):
"""Test that the plugin copes with non-lat/lon grids."""
trg_crs = None
src_crs = ccrs.PlateCarree()
trg_crs = ccrs.LambertConformal(central_longitude=50,
central_latitude=10)
trg_crs_iris = coord_systems.LambertConformal(central_lon=50,
central_lat=10)
lons = self.cube.coord('longitude').points
lats = self.cube.coord('latitude').points
x, y = [], []
for lon, lat in zip(lons, lats):
x_trg, y_trg = trg_crs.transform_point(lon, lat, src_crs)
x.append(x_trg)
y.append(y_trg)
new_x = AuxCoord(x,
standard_name='projection_x_coordinate',
units='m',
coord_system=trg_crs_iris)
new_y = AuxCoord(y,
standard_name='projection_y_coordinate',
units='m',
coord_system=trg_crs_iris)
cube = Cube(self.cube.data,
long_name="air_temperature",
dim_coords_and_dims=[(self.cube.coord('time'), 0)],
aux_coords_and_dims=[(new_y, 1), (new_x, 2)],
units="K")
plugin = Plugin(method)
with iris.FUTURE.context(cell_datetime_objects=True):
cube = cube.extract(self.time_extract)
result = plugin.process(cube, self.sites, self.neighbour_list,
ancillary_data, additional_data, **kwargs)
self.assertEqual(cube.coord_system(), trg_crs_iris)
self.assertAlmostEqual(result.data, expected)
self.assertEqual(result.coord(axis='y').name(), 'latitude')
self.assertEqual(result.coord(axis='x').name(), 'longitude')
self.assertAlmostEqual(result.coord(axis='y').points, 4.74)
self.assertAlmostEqual(result.coord(axis='x').points, 9.47)
def lat_lon_determine(cube: Cube) -> Optional[CRS]:
"""
Test whether a diagnostic cube is on a latitude/longitude grid or uses an
alternative projection.
Args:
cube:
A diagnostic cube to examine for coordinate system.
Returns:
Coordinate system of the diagnostic cube in a cartopy format unless
it is already a latitude/longitude grid, in which case None is
returned.
"""
trg_crs = None
if (not cube.coord(axis="x").name() == "longitude"
or not cube.coord(axis="y").name() == "latitude"):
trg_crs = cube.coord_system().as_cartopy_crs()
return trg_crs
def transform_grid_to_lat_lon(cube: Cube) -> Tuple[ndarray, ndarray]:
"""Calculate the latitudes and longitudes of each points in the cube.
The result is defined over the spatial grid, of shape (ny, nx) where
ny is the length of the y-axis coordinate and nx the length of the
x-axis coordinate.
Args:
cube:
Cube with points to transform
Returns
- Array of shape (ny, nx) containing grid latitude values
- Array of shape (ny, nx) containing grid longitude values
"""
trg_latlon = ccrs.PlateCarree()
trg_crs = cube.coord_system().as_cartopy_crs()
cube = cube.copy()
# TODO use the proj units that are accesible with later versions of proj
# to determine the default units to convert to for a given projection.
# Assuming proj units of metre for all projections not in degrees.
for axis in ["x", "y"]:
try:
cube.coord(axis=axis).convert_units("m")
except ValueError as err:
msg = (
"Cube passed to transform_grid_to_lat_lon does not have an "
f"{axis} coordinate with units that can be converted to metres. "
)
raise ValueError(msg + str(err))
all_y_points, all_x_points = get_grid_y_x_values(cube)
# Transform points
points = trg_latlon.transform_points(trg_crs, all_x_points, all_y_points)
lons = points[..., 0]
lats = points[..., 1]
return lats, lons
def calc_true_north_offset(reference_cube: Cube) -> ndarray:
"""
Calculate the angles between grid North and true North, as a
matrix of values on the grid of the input reference cube.
Args:
reference_cube:
2D cube on grid for which "north" is required. Provides both
coordinate system (reference_cube.coord_system()) and template
spatial grid on which the angle adjustments should be provided.
Returns:
Angle in radians by which wind direction wrt true North at
each point must be rotated to be relative to grid North.
"""
reference_x_coord = reference_cube.coord(axis="x")
reference_y_coord = reference_cube.coord(axis="y")
# find corners of reference_cube grid in lat / lon coordinates
latlon = [
GLOBAL_CRS.as_cartopy_crs().transform_point(
reference_x_coord.points[i],
reference_y_coord.points[j],
reference_cube.coord_system().as_cartopy_crs(),
)
for i in [0, -1]
for j in [0, -1]
]
latlon = np.array(latlon).T.tolist()
# define lat / lon coordinates to cover the reference_cube grid at an
# equivalent resolution
lat_points = np.linspace(
np.floor(min(latlon[1])),
np.ceil(max(latlon[1])),
len(reference_y_coord.points),
)
lon_points = np.linspace(
np.floor(min(latlon[0])),
np.ceil(max(latlon[0])),
len(reference_x_coord.points),
)
lat_coord = DimCoord(
lat_points, "latitude", units="degrees", coord_system=GLOBAL_CRS
)
lon_coord = DimCoord(
lon_points, "longitude", units="degrees", coord_system=GLOBAL_CRS
)
# define a unit vector wind towards true North over the lat / lon grid
udata = np.zeros(reference_cube.shape, dtype=np.float32)
vdata = np.ones(reference_cube.shape, dtype=np.float32)
ucube_truenorth = Cube(
udata,
"grid_eastward_wind",
dim_coords_and_dims=[(lat_coord, 0), (lon_coord, 1)],
)
vcube_truenorth = Cube(
vdata,
"grid_northward_wind",
dim_coords_and_dims=[(lat_coord, 0), (lon_coord, 1)],
)
# rotate unit vector onto reference_cube coordinate system
ucube, vcube = rotate_winds(
ucube_truenorth, vcube_truenorth, reference_cube.coord_system()
)
# unmask and regrid rotated winds onto reference_cube grid
ucube.data = ucube.data.data
ucube = ucube.regrid(reference_cube, Linear())
vcube.data = vcube.data.data
vcube = vcube.regrid(reference_cube, Linear())
# ratio of u to v winds is the tangent of the angle which is the
# true North to grid North rotation
angle_adjustment = np.arctan2(ucube.data, vcube.data)
return angle_adjustment
def process(
self, sites: List[Dict[str, Any]], orography: Cube, land_mask: Cube
) -> Cube:
"""
Using the constraints provided, find the nearest grid point neighbours
to the given spot sites for the model/grid given by the input cubes.
Returned is a cube that contains the defining characteristics of the
spot sites (e.g. x coordinate, y coordinate, altitude) and the indices
of the selected grid point neighbour.
Args:
sites:
A list of dictionaries defining the spot sites for which
neighbours are to be found. e.g.:
[{'altitude': 11.0, 'latitude': 57.867000579833984,
'longitude': -5.632999897003174, 'wmo_id': 3034}]
orography:
A cube of orography, used to obtain the grid point altitudes.
land_mask:
A land mask cube for the model/grid from which grid point
neighbours are being selected, with land points set to one and
sea points set to zero.
Returns:
A cube containing both the spot site information and for each
the grid point indices of its nearest neighbour as per the
imposed constraints.
"""
# Check if we are dealing with a global grid.
self.global_coordinate_system = orography.coord(axis="x").circular
# Exclude regional grids with spatial dimensions other than metres.
if not self.global_coordinate_system:
if not orography.coord(axis="x").units == "metres":
msg = (
"Cube spatial coordinates for regional grids must be"
"in metres to match the defined search_radius."
)
raise ValueError(msg)
# Ensure land_mask and orography are on the same grid.
if not orography.dim_coords == land_mask.dim_coords:
msg = "Orography and land_mask cubes are not on the same " "grid."
raise ValueError(msg)
# Enforce x-y coordinate order for input cubes.
enforce_coordinate_ordering(
orography,
[orography.coord(axis="x").name(), orography.coord(axis="y").name()],
)
enforce_coordinate_ordering(
land_mask,
[land_mask.coord(axis="x").name(), land_mask.coord(axis="y").name()],
)
# Remap site coordinates on to coordinate system of the model grid.
site_x_coords = np.array([site[self.site_x_coordinate] for site in sites])
site_y_coords = np.array([site[self.site_y_coordinate] for site in sites])
site_coords = self._transform_sites_coordinate_system(
site_x_coords, site_y_coords, orography.coord_system().as_cartopy_crs()
)
# Exclude any sites falling outside the domain given by the cube and
# notify the user.
(
sites,
site_coords,
site_x_coords,
site_y_coords,
) = self.check_sites_are_within_domain(
sites, site_coords, site_x_coords, site_y_coords, orography
)
# Find nearest neighbour point using quick iris method.
nearest_indices = self.get_nearest_indices(site_coords, orography)
# Create an array containing site altitudes, using the nearest point
# orography height for any that are unset.
site_altitudes = np.array(
[site.get(self.site_altitude, None) for site in sites]
)
site_altitudes = np.where(
np.isnan(site_altitudes.astype(float)),
orography.data[tuple(nearest_indices.T)],
site_altitudes,
)
# If further constraints are being applied, build a KD Tree which
# includes points filtered by constraint.
if self.land_constraint or self.minimum_dz:
# Build the KDTree, an internal test for the land_constraint checks
# whether to exclude sea points from the tree.
tree, index_nodes = self.build_KDTree(land_mask)
# Site coordinates made cartesian for global coordinate system
if self.global_coordinate_system:
site_coords = self.geocentric_cartesian(
orography, site_coords[:, 0], site_coords[:, 1]
)
if not self.minimum_dz:
# Query the tree for the nearest neighbour, in this case a land
# neighbour is returned along with the distance to it.
distances, node_indices = tree.query([site_coords])
# Look up the grid coordinates that correspond to the tree node
(land_neighbour_indices,) = index_nodes[node_indices]
# Use the found land neighbour if it is within the
# search_radius, otherwise use the nearest neighbour.
distances = np.array([distances[0], distances[0]]).T
nearest_indices = np.where(
distances < self.search_radius,
land_neighbour_indices,
nearest_indices,
)
else:
# Query the tree for self.node_limit nearby neighbours.
distances, node_indices = tree.query(
[site_coords],
distance_upper_bound=self.search_radius,
k=self.node_limit,
)
# Loop over the sites and for each choose the returned
# neighbour with the minimum vertical displacement.
for index, (distance, indices) in enumerate(
zip(distances[0], node_indices[0])
):
grid_point = self.select_minimum_dz(
orography, site_altitudes[index], index_nodes, distance, indices
)
# None is returned if the tree query returned no neighbours
# within the search radius.
if grid_point is not None:
nearest_indices[index] = grid_point
# Calculate the vertical displacements between the chosen grid point
# and the spot site.
vertical_displacements = (
site_altitudes - orography.data[tuple(nearest_indices.T)]
)
# Create a list of WMO IDs if available. These are stored as strings
# to accommodate the use of 'None' for unset IDs.
wmo_ids = [str(site.get("wmo_id", None)) for site in sites]
# Construct a name to describe the neighbour finding method employed
method_name = self.neighbour_finding_method_name()
# Create an array of indices and displacements to return
data = np.stack(
(nearest_indices[:, 0], nearest_indices[:, 1], vertical_displacements),
axis=0,
)
data = np.expand_dims(data, 0).astype(np.float32)
# Regardless of input sitelist coordinate system, the site coordinates
# are stored as latitudes and longitudes in the neighbour cube.
if self.site_coordinate_system != ccrs.PlateCarree():
lon_lats = self._transform_sites_coordinate_system(
site_x_coords, site_y_coords, ccrs.PlateCarree()
)
longitudes = lon_lats[:, 0]
latitudes = lon_lats[:, 1]
else:
longitudes = site_x_coords
latitudes = site_y_coords
# Create a cube of neighbours
neighbour_cube = build_spotdata_cube(
data,
"grid_neighbours",
1,
site_altitudes.astype(np.float32),
latitudes.astype(np.float32),
longitudes.astype(np.float32),
wmo_ids,
neighbour_methods=[method_name],
grid_attributes=["x_index", "y_index", "vertical_displacement"],
)
# Add a hash attribute based on the model grid to ensure the neighbour
# cube is only used with a compatible grid.
grid_hash = create_coordinate_hash(orography)
neighbour_cube.attributes["model_grid_hash"] = grid_hash
return neighbour_cube
def _regrid_and_populate(
self,
temperature: Cube,
humidity: Cube,
pressure: Cube,
uwind: Cube,
vwind: Cube,
topography: Cube,
) -> None:
"""
Regrids input variables onto the high resolution orography field, then
populates the class instance with regridded variables before converting
to SI units. Also calculates V.gradZ as a class member.
Args:
temperature:
Temperature at top of boundary layer
humidity:
Relative humidity at top of boundary layer
pressure:
Pressure at top of boundary layer
uwind:
Positive eastward wind vector component at top of boundary
layer
vwind:
Positive northward wind vector component at top of boundary
layer
topography:
Height of topography above sea level on 1 km UKPP domain grid
"""
# convert topography grid, datatype and units
for axis in ["x", "y"]:
topography = sort_coord_in_cube(topography,
topography.coord(axis=axis))
enforce_coordinate_ordering(
topography,
[
topography.coord(axis="y").name(),
topography.coord(axis="x").name()
],
)
self.topography = topography.copy(
data=topography.data.astype(np.float32))
self.topography.convert_units("m")
# rotate winds
try:
uwind, vwind = rotate_winds(uwind, vwind,
topography.coord_system())
except ValueError as err:
if "Duplicate coordinates are not permitted" in str(err):
# ignore error raised if uwind and vwind do not need rotating
pass
else:
raise ValueError(str(err))
else:
# remove auxiliary spatial coordinates from rotated winds
for cube in [uwind, vwind]:
for axis in ["x", "y"]:
cube.remove_coord(cube.coord(axis=axis, dim_coords=False))
# regrid and convert input variables
self.temperature = self._regrid_variable(temperature, "kelvin")
self.humidity = self._regrid_variable(humidity, "1")
self.pressure = self._regrid_variable(pressure, "Pa")
self.uwind = self._regrid_variable(uwind, "m s-1")
self.vwind = self._regrid_variable(vwind, "m s-1")
# calculate orography gradients
gradx, grady = self._orography_gradients()
# calculate v.gradZ
self.vgradz = np.multiply(gradx.data, self.uwind.data) + np.multiply(
grady.data, self.vwind.data)
