def test_specify_realization_numbers(self):
"""Use the ensemble_realization_numbers optional argument to specify
particular values for the ensemble realization numbers."""
ensemble_realization_numbers = [12, 13, 14]
result = Plugin().process(self.cube, ensemble_realization_numbers)
self.assertArrayEqual(
result.coord("realization").points, ensemble_realization_numbers)
def test_raises_exception_if_realization_already_exists(self):
"""Check that we raise an exception if a realization coordinate already
exists."""
cube = self.current_temperature_cube
cube.add_aux_coord(AuxCoord(0, 'realization'))
plugin = Plugin()
msg = r"Cannot rebadge percentile coordinate to realization.*"
with self.assertRaisesRegex(InvalidCubeError, msg):
plugin.process(cube)
def test_basic(self):
"""Test that a cube is produced is produced and the realization
coordinate is a dimension coordinate, after the percentile coordinate
is rebadged."""
cube = self.current_temperature_cube
plugin = Plugin()
result = plugin.process(cube)
self.assertIsInstance(result, Cube)
self.assertIsInstance(result.coord("realization"), DimCoord)
def test_number_of_realizations(self):
"""Check the values for the realization coordinate generated without
specifying the ensemble_realization_numbers argument."""
cube = self.current_temperature_cube
plen = len(cube.coord("percentile_over_realization").points)
plugin = Plugin()
result = plugin.process(cube)
self.assertEqual(len(result.coord("realization").points), plen)
self.assertArrayAlmostEqual(
result.coord("realization").points, np.array([0, 1, 2]))
def test_still_works_if_percentile_coord_is_different(self):
"""Check this still works if a different percentile coord used."""
cube = self.current_temperature_cube
cube.coord("percentile_over_realization").rename(
"percentile_over_nbhood")
plen = len(cube.coord("percentile_over_nbhood").points)
plugin = Plugin()
result = plugin.process(cube)
self.assertEqual(len(result.coord("realization").points), plen)
self.assertArrayAlmostEqual(
result.coord("realization").points, np.array([0, 1, 2]))
def test_specify_realization_numbers(self):
"""Use the ensemble_realization_numbers optional argument to specify
particular values for the ensemble realization numbers."""
cube = self.current_temperature_cube
plen = len(cube.coord("percentile_over_realization").points)
ensemble_realization_numbers = np.arange(plen) + 12
plugin = Plugin()
result = plugin.process(cube, ensemble_realization_numbers)
self.assertEqual(len(result.coord("realization").points), plen)
self.assertArrayAlmostEqual(
result.coord("realization").points, np.array([12, 13, 14]))
def process(self, cube: Cube) -> Cube:
"""Expected value calculation and metadata updates.
Args:
cube:
Probabilistic data with a realization, threshold or percentile
representation.
Returns:
Expected value of probability distribution. Same shape as input cube
but with realization/threshold/percentile coordinate removed.
"""
if is_probability(cube):
# TODO: replace this with direct calculation of the integral over
# probability thresholds. The current approach works and has the
# correct interface, but some accuracy will be lost in the conversion
# and memory usage is very high.
#
# 19 percentiles corresponds to 5, 10, 15...95%.
cube = ConvertProbabilitiesToPercentiles().process(
cube, no_of_percentiles=19)
if is_percentile(cube):
cube = RebadgePercentilesAsRealizations().process(cube)
mean_cube = collapse_realizations(cube)
mean_cube.add_cell_method(CellMethod("mean", coords="realization"))
return mean_cube
def _combine_subensembles(
self, forecast_subensembles: Cube,
output_realizations_count: Optional[int]) -> Cube:
"""Combine the forecast sub-ensembles into a single ensemble. This is done by
first stacking the sub-ensembles into a single super-ensemble and then resampling
the super-ensemble to produce a subset of output realizations.
Args:
forecast_subensembles:
Cube containing a series of forecast sub-ensembles.
output_realizations_count:
The number of ensemble realizations that will be extracted from the
super-ensemble. If realizations_count is None, all realizations will
be returned.
Returns:
Cube containing single realization dimension.
"""
superensemble_cube = self._stack_subensembles(forecast_subensembles)
if output_realizations_count is None:
warnings.warn(
Warning(
"output_realizations_count not specified. Returning all realizations from the "
"full super-ensemble."))
return superensemble_cube
output_percentiles = choose_set_of_percentiles(
output_realizations_count,
sampling="quantile",
)
percentile_cube = generate_percentiles.process(
superensemble_cube,
coordinates="realization",
percentiles=output_percentiles,
)
reduced_ensemble = RebadgePercentilesAsRealizations()(percentile_cube)
return reduced_ensemble
def process(self, cube: Cube) -> Cube:
"""Expected value calculation and metadata updates.
Args:
cube:
Probabilistic data with a realization, threshold or percentile
representation.
Returns:
Expected value of probability distribution. Same shape as input cube
but with realization/threshold/percentile coordinate removed.
"""
if is_probability(cube):
ev_cube = self.integrate_over_thresholds(cube)
elif is_percentile(cube):
ev_cube = collapse_realizations(
RebadgePercentilesAsRealizations().process(cube)
)
else:
ev_cube = collapse_realizations(cube)
ev_cube.add_cell_method(CellMethod("mean", coords="realization"))
return ev_cube
def process(
cube: cli.inputcube,
raw_cube: cli.inputcube = None,
*,
realizations_count: int = None,
random_seed: int = None,
ignore_ecc_bounds=False,
):
"""Converts an incoming cube into one containing realizations.
Args:
cube (iris.cube.Cube):
A cube to be processed.
raw_cube (iris.cube.Cube):
Cube of raw (not post processed) weather data.
If this argument is given ensemble realizations will be created
from percentiles by reshuffling them in correspondence to the rank
order of the raw ensemble. Otherwise, the percentiles are rebadged
as realizations.
realizations_count (int):
The number of ensemble realizations in the output.
random_seed (int):
Option to specify a value for the random seed when reordering percentiles.
This value is for testing purposes only, to ensure reproduceable outputs.
It should not be used in real time operations as it may introduce a bias
into the reordered forecasts.
ignore_ecc_bounds (bool):
If True where percentiles (calculated as an intermediate output
before realization) exceed the ECC bounds range, raises a
warning rather than an exception.
Returns:
iris.cube.Cube:
The processed cube.
"""
from improver.ensemble_copula_coupling.ensemble_copula_coupling import (
ConvertProbabilitiesToPercentiles,
EnsembleReordering,
RebadgePercentilesAsRealizations,
ResamplePercentiles,
)
from improver.metadata.probabilistic import is_probability
if cube.coords("realization"):
return cube
if not cube.coords("percentile") and not is_probability(cube):
raise ValueError("Unable to convert to realizations:\n" + str(cube))
if realizations_count is None:
try:
realizations_count = len(raw_cube.coord("realization").points)
except AttributeError:
# raised if raw_cube is None, hence has no attribute "coord"
msg = "Either realizations_count or raw_cube must be provided"
raise ValueError(msg)
if cube.coords("percentile"):
percentiles = ResamplePercentiles(
ecc_bounds_warning=ignore_ecc_bounds)(
cube, no_of_percentiles=realizations_count)
else:
percentiles = ConvertProbabilitiesToPercentiles(
ecc_bounds_warning=ignore_ecc_bounds)(
cube, no_of_percentiles=realizations_count)
if raw_cube:
result = EnsembleReordering()(percentiles,
raw_cube,
random_seed=random_seed)
else:
result = RebadgePercentilesAsRealizations()(percentiles)
return result
def main(argv=None):
"""
Load in the arguments and apply the requested variant of Ensemble
Copula Coupling for converting percentile data to realizations.
"""
parser = ArgParser(description='Convert a dataset containing '
'probabilities into one containing '
'ensemble realizations using Ensemble Copula Coupling.')
# General options:
parser.add_argument('input_filepath',
metavar='INPUT_FILE',
help='A path to an input NetCDF file to be processed.'
' Must contain a percentile dimension.')
parser.add_argument('output_filepath',
metavar='OUTPUT_FILE',
help='The output path for the processed NetCDF.')
parser.add_argument('--no_of_percentiles',
default=None,
type=int,
metavar='NUMBER_OF_PERCENTILES',
help='The number of percentiles to be generated. '
'This is also equal to the number of ensemble '
'realizations that will be generated.')
parser.add_argument('--sampling_method',
default='quantile',
const='quantile',
nargs='?',
choices=['quantile', 'random'],
metavar='PERCENTILE_SAMPLING_METHOD',
help='Method to be used for generating the list of '
'percentiles with forecasts generated at each '
'percentile. The options are "quantile" and '
'"random". "quantile" is the default option. '
'The "quantile" option produces equally spaced '
'percentiles which is the preferred '
'option for full Ensemble Copula Coupling with '
'reordering enabled.')
parser.add_argument(
'--ecc_bounds_warning',
default=False,
action='store_true',
help='If True, where percentiles (calculated as an intermediate '
'output before realizations) exceed the ECC bounds range, raise '
'a warning rather than an exception.')
# Different use cases:
# (We can either reorder OR rebadge)
group = parser.add_mutually_exclusive_group(required=True)
group.add_argument('--reordering',
default=False,
action='store_true',
help='The option used to create ensemble realizations '
'from percentiles by reordering the input '
'percentiles based on the order of the '
'raw ensemble forecast.')
group.add_argument('--rebadging',
default=False,
action='store_true',
help='The option used to create ensemble realizations '
'from percentiles by rebadging the input '
'percentiles.')
# If reordering, can do so either based on original realizations,
# or randomly.
reordering = parser.add_argument_group(
'Reordering options', 'Options for reordering the input percentiles '
'using the raw ensemble forecast as required to create ensemble '
'realizations.')
reordering.add_argument('--raw_forecast_filepath',
metavar='RAW_FORECAST_FILE',
help='A path to an raw forecast NetCDF file to be '
'processed. This option is compulsory, if the '
'reordering option is selected.')
reordering.add_argument('--random_ordering',
default=False,
action='store_true',
help='Decide whether or not to use random '
'ordering within the ensemble reordering step.')
reordering.add_argument(
'--random_seed',
default=None,
help='Option to specify a value for the random seed for testing '
'purposes, otherwise, the default random seed behaviour is '
'utilised. The random seed is used in the generation of the '
'random numbers used for either the random_ordering option to '
'order the input percentiles randomly, rather than use the '
'ordering from the raw ensemble, or for splitting tied values '
'within the raw ensemble, so that the values from the input '
'percentiles can be ordered to match the raw ensemble.')
rebadging = parser.add_argument_group(
'Rebadging options', 'Options for rebadging the input percentiles '
'as ensemble realizations.')
rebadging.add_argument('--realization_numbers',
default=None,
metavar='REALIZATION_NUMBERS',
nargs="+",
help='A list of ensemble realization numbers to '
'use when rebadging the percentiles '
'into realizations.')
args = parser.parse_args(args=argv)
# CLI argument checking:
# Can only do one of reordering or rebadging: if options are passed which
# correspond to the opposite method, raise an exception.
# Note: Shouldn't need to check that both/none are set, since they are
# defined as mandatory, but mutually exclusive, options.
if args.reordering:
if args.realization_numbers is not None:
parser.wrong_args_error('realization_numbers', 'reordering')
if args.rebadging:
if ((args.raw_forecast_filepath is not None)
or (args.random_ordering is not False)):
parser.wrong_args_error('raw_forecast_filepath, random_ordering',
'rebadging')
# Safe to now actually do the work...
cube = load_cube(args.input_filepath)
result_cube = ResamplePercentiles(
ecc_bounds_warning=args.ecc_bounds_warning).process(
cube,
no_of_percentiles=args.no_of_percentiles,
sampling=args.sampling_method)
if args.reordering:
raw_forecast = load_cube(args.raw_forecast_filepath)
result_cube = EnsembleReordering().process(
result_cube,
raw_forecast,
random_ordering=args.random_ordering,
random_seed=args.random_seed)
elif args.rebadging:
if args.realization_numbers is not None:
args.realization_numbers = ([
int(num) for num in args.realization_numbers
])
result_cube = RebadgePercentilesAsRealizations().process(
result_cube, ensemble_realization_numbers=args.realization_numbers)
save_netcdf(result_cube, args.output_filepath)
def process(cube: cli.inputcube,
coefficients: cli.inputcube = None,
land_sea_mask: cli.inputcube = None,
*,
distribution,
realizations_count: int = None,
randomise=False,
random_seed: int = None,
ignore_ecc_bounds=False,
predictor='mean',
shape_parameters: cli.comma_separated_list = None):
"""Applying coefficients for Ensemble Model Output Statistics.
Load in arguments for applying coefficients for Ensemble Model Output
Statistics (EMOS), otherwise known as Non-homogeneous Gaussian
Regression (NGR). The coefficients are applied to the forecast
that is supplied, so as to calibrate the forecast. The calibrated
forecast is written to a cube. If no coefficients are provided the input
forecast is returned unchanged.
Args:
cube (iris.cube.Cube):
A Cube containing the forecast to be calibrated. The input format
could be either realizations, probabilities or percentiles.
coefficients (iris.cube.Cube):
A cube containing the coefficients used for calibration or None.
If none then then input is returned unchanged.
land_sea_mask (iris.cube.Cube):
A cube containing the land-sea mask on the same domain as the
forecast that is to be calibrated. Land points are "
"specified by ones and sea points are specified by zeros. "
"If not None this argument will enable land-only calibration, in "
"which sea points are returned without the application of "
"calibration."
distribution (str):
The distribution for constructing realizations, percentiles or
probabilities. This should typically match the distribution used
for minimising the Continuous Ranked Probability Score when
estimating the EMOS coefficients. The distributions available are
those supported by :data:`scipy.stats`.
realizations_count (int):
Option to specify the number of ensemble realizations that will be
created from probabilities or percentiles for input into EMOS.
randomise (bool):
Option to reorder the post-processed forecasts randomly. If not
set, the ordering of the raw ensemble is used. This option is
only valid when the input format is realizations.
random_seed (int):
Option to specify a value for the random seed for testing
purposes, otherwise the default random seen behaviour is utilised.
The random seed is used in the generation of the random numbers
used for either the randomise option to order the input
percentiles randomly, rather than use the ordering from the raw
ensemble, or for splitting tied values within the raw ensemble,
so that the values from the input percentiles can be ordered to
match the raw ensemble.
ignore_ecc_bounds (bool):
If True, where the percentiles exceed the ECC bounds range,
raises a warning rather than an exception. This occurs when the
current forecasts is in the form of probabilities and is
converted to percentiles, as part of converting the input
probabilities into realizations.
predictor (str):
String to specify the form of the predictor used to calculate
the location parameter when estimating the EMOS coefficients.
Currently the ensemble mean ("mean") and the ensemble
realizations ("realizations") are supported as the predictors.
shape_parameters (float or str):
The shape parameters required for defining the distribution
specified by the distribution argument. The shape parameters
should either be a number or 'inf' or '-inf' to represent
infinity. Further details about appropriate shape parameters
are available in scipy.stats. For the truncated normal
distribution with a lower bound of zero, as available when
estimating EMOS coefficients, the appropriate shape parameters
are 0 and inf.
Returns:
iris.cube.Cube:
The calibrated forecast cube.
Raises:
ValueError:
If the current forecast is a coefficients cube.
ValueError:
If the coefficients cube does not have the right name of
"emos_coefficients".
ValueError:
If the forecast type is 'percentiles' or 'probabilities' and the
realizations_count argument is not provided.
"""
import warnings
import numpy as np
from iris.exceptions import CoordinateNotFoundError
from improver.calibration.ensemble_calibration import (
ApplyCoefficientsFromEnsembleCalibration)
from improver.ensemble_copula_coupling.ensemble_copula_coupling import (
EnsembleReordering, ConvertLocationAndScaleParametersToPercentiles,
ConvertLocationAndScaleParametersToProbabilities,
ConvertProbabilitiesToPercentiles, RebadgePercentilesAsRealizations,
ResamplePercentiles)
from improver.calibration.utilities import merge_land_and_sea
from improver.metadata.probabilistic import find_percentile_coordinate
current_forecast = cube
if current_forecast.name() in ['emos_coefficients', 'land_binary_mask']:
msg = "The current forecast cube has the name {}"
raise ValueError(msg.format(current_forecast.name()))
if coefficients is None:
msg = ("There are no coefficients provided for calibration. The "
"uncalibrated forecast will be returned.")
warnings.warn(msg)
return current_forecast
if coefficients.name() != 'emos_coefficients':
msg = ("The current coefficients cube does not have the "
"name 'emos_coefficients'")
raise ValueError(msg)
if land_sea_mask and land_sea_mask.name() != 'land_binary_mask':
msg = ("The land_sea_mask cube does not have the "
"name 'land_binary_mask'")
raise ValueError(msg)
original_current_forecast = current_forecast.copy()
try:
find_percentile_coordinate(current_forecast)
input_forecast_type = "percentiles"
except CoordinateNotFoundError:
input_forecast_type = "realizations"
if current_forecast.name().startswith("probability_of"):
input_forecast_type = "probabilities"
conversion_plugin = ConvertProbabilitiesToPercentiles(
ecc_bounds_warning=ignore_ecc_bounds)
elif input_forecast_type == "percentiles":
# Initialise plugin to resample percentiles so that the percentiles are
# evenly spaced.
conversion_plugin = ResamplePercentiles(
ecc_bounds_warning=ignore_ecc_bounds)
if input_forecast_type in ["percentiles", "probabilities"]:
if not realizations_count:
raise ValueError(
"The current forecast has been provided as {0}. "
"These {0} need to be converted to realizations "
"for ensemble calibration. The realizations_count "
"argument is used to define the number of realizations "
"to construct from the input {0}, so if the "
"current forecast is provided as {0} then "
"realizations_count must be defined.".format(
input_forecast_type))
current_forecast = conversion_plugin.process(
current_forecast, no_of_percentiles=realizations_count)
current_forecast = (
RebadgePercentilesAsRealizations().process(current_forecast))
# Apply coefficients as part of Ensemble Model Output Statistics (EMOS).
ac = ApplyCoefficientsFromEnsembleCalibration(predictor=predictor)
location_parameter, scale_parameter = ac.process(
current_forecast, coefficients, landsea_mask=land_sea_mask)
if shape_parameters:
shape_parameters = [np.float32(x) for x in shape_parameters]
# Convert the output forecast type (i.e. realizations, percentiles,
# probabilities) to match the input forecast type.
if input_forecast_type == "probabilities":
result = ConvertLocationAndScaleParametersToProbabilities(
distribution=distribution,
shape_parameters=shape_parameters).process(
location_parameter, scale_parameter, original_current_forecast)
elif input_forecast_type == "percentiles":
perc_coord = find_percentile_coordinate(original_current_forecast)
result = ConvertLocationAndScaleParametersToPercentiles(
distribution=distribution,
shape_parameters=shape_parameters).process(
location_parameter,
scale_parameter,
original_current_forecast,
percentiles=perc_coord.points)
elif input_forecast_type == "realizations":
# Ensemble Copula Coupling to generate realizations
# from the location and scale parameter.
no_of_percentiles = len(current_forecast.coord('realization').points)
percentiles = ConvertLocationAndScaleParametersToPercentiles(
distribution=distribution,
shape_parameters=shape_parameters).process(
location_parameter,
scale_parameter,
original_current_forecast,
no_of_percentiles=no_of_percentiles)
result = EnsembleReordering().process(percentiles,
current_forecast,
random_ordering=randomise,
random_seed=random_seed)
if land_sea_mask:
# Fill in masked sea points with uncalibrated data.
merge_land_and_sea(result, original_current_forecast)
return result
def main(argv=None):
"""Load in arguments for applying coefficients for Ensemble Model Output
Statistics (EMOS), otherwise known as Non-homogeneous Gaussian
Regression (NGR). The coefficients are applied to the forecast
that is supplied, so as to calibrate the forecast. The calibrated
forecast is written to a netCDF file.
"""
parser = ArgParser(
description='Apply coefficients for Ensemble Model Output '
'Statistics (EMOS), otherwise known as Non-homogeneous '
'Gaussian Regression (NGR). The supported input formats '
'are realizations, probabilities and percentiles. '
'The forecast will be converted to realizations before '
'applying the coefficients and then converted back to '
'match the input format.')
# Filepaths for the forecast, EMOS coefficients and the output.
parser.add_argument(
'forecast_filepath',
metavar='FORECAST_FILEPATH',
help='A path to an input NetCDF file containing the forecast to be '
'calibrated. The input format could be either realizations, '
'probabilities or percentiles.')
parser.add_argument('coefficients_filepath',
metavar='COEFFICIENTS_FILEPATH',
help='A path to an input NetCDF file containing the '
'coefficients used for calibration.')
parser.add_argument('output_filepath',
metavar='OUTPUT_FILEPATH',
help='The output path for the processed NetCDF')
# Optional arguments.
parser.add_argument(
'--num_realizations',
metavar='NUMBER_OF_REALIZATIONS',
default=None,
type=np.int32,
help='Optional argument to specify the number of '
'ensemble realizations to produce. '
'If the current forecast is input as probabilities or '
'percentiles then this argument is used to create the requested '
'number of realizations. In addition, this argument is used to '
'construct the requested number of realizations from the mean '
'and variance output after applying the EMOS coefficients.'
'Default will be the number of realizations in the raw input '
'file, if realizations are provided as input, otherwise if the '
'input format is probabilities or percentiles, then an error '
'will be raised if no value is provided.')
parser.add_argument(
'--random_ordering',
default=False,
action='store_true',
help='Option to reorder the post-processed forecasts randomly. If not '
'set, the ordering of the raw ensemble is used. This option is '
'only valid when the input format is realizations.')
parser.add_argument(
'--random_seed',
metavar='RANDOM_SEED',
default=None,
help='Option to specify a value for the random seed for testing '
'purposes, otherwise, the default random seed behaviour is '
'utilised. The random seed is used in the generation of the '
'random numbers used for either the random_ordering option to '
'order the input percentiles randomly, rather than use the '
'ordering from the raw ensemble, or for splitting tied values '
'within the raw ensemble, so that the values from the input '
'percentiles can be ordered to match the raw ensemble.')
parser.add_argument(
'--ecc_bounds_warning',
default=False,
action='store_true',
help='If True, where the percentiles exceed the ECC bounds range, '
'raise a warning rather than an exception. This occurs when the '
'current forecast is in the form of probabilities and is '
'converted to percentiles, as part of converting the input '
'probabilities into realizations.')
parser.add_argument(
'--predictor_of_mean',
metavar='PREDICTOR_OF_MEAN',
choices=['mean', 'realizations'],
default='mean',
help='String to specify the predictor used to calibrate the forecast '
'mean. Currently the ensemble mean ("mean") and the ensemble '
'realizations ("realizations") are supported as options. '
'Default: "mean".')
args = parser.parse_args(args=argv)
current_forecast = load_cube(args.forecast_filepath)
coeffs = load_cube(args.coefficients_filepath)
original_current_forecast = current_forecast.copy()
msg = ("The current forecast has been provided as {0}. "
"These {0} need to be converted to realizations "
"for ensemble calibration. The args.num_realizations "
"argument is used to define the number of realizations "
"to construct from the input {0}, so if the "
"current forecast is provided as {0} then "
"args.num_realizations must be defined.")
try:
find_percentile_coordinate(current_forecast)
input_forecast_type = "percentiles"
except CoordinateNotFoundError:
input_forecast_type = "realizations"
if current_forecast.name().startswith("probability_of"):
input_forecast_type = "probabilities"
# If probabilities, convert to percentiles.
conversion_plugin = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=args.ecc_bounds_warning)
elif input_forecast_type == "percentiles":
# If percentiles, resample percentiles so that the percentiles are
# evenly spaced.
conversion_plugin = ResamplePercentiles(
ecc_bounds_warning=args.ecc_bounds_warning)
# If percentiles, resample percentiles and then rebadge.
# If probabilities, generate percentiles and then rebadge.
if input_forecast_type in ["percentiles", "probabilities"]:
if not args.num_realizations:
raise ValueError(msg.format(input_forecast_type))
current_forecast = conversion_plugin.process(
current_forecast, no_of_percentiles=args.num_realizations)
current_forecast = (
RebadgePercentilesAsRealizations().process(current_forecast))
# Default number of ensemble realizations is the number in
# the raw forecast.
if not args.num_realizations:
args.num_realizations = len(
current_forecast.coord('realization').points)
# Apply coefficients as part of Ensemble Model Output Statistics (EMOS).
ac = ApplyCoefficientsFromEnsembleCalibration(
current_forecast,
coeffs,
predictor_of_mean_flag=args.predictor_of_mean)
calibrated_predictor, calibrated_variance = ac.process()
# If input forecast is probabilities, convert output into probabilities.
# If input forecast is percentiles, convert output into percentiles.
# If input forecast is realizations, convert output into realizations.
if input_forecast_type == "probabilities":
result = GenerateProbabilitiesFromMeanAndVariance().process(
calibrated_predictor, calibrated_variance,
original_current_forecast)
elif input_forecast_type == "percentiles":
perc_coord = find_percentile_coordinate(original_current_forecast)
result = GeneratePercentilesFromMeanAndVariance().process(
calibrated_predictor,
calibrated_variance,
percentiles=perc_coord.points)
elif input_forecast_type == "realizations":
# Ensemble Copula Coupling to generate realizations
# from mean and variance.
percentiles = GeneratePercentilesFromMeanAndVariance().process(
calibrated_predictor,
calibrated_variance,
no_of_percentiles=args.num_realizations)
result = EnsembleReordering().process(
percentiles,
current_forecast,
random_ordering=args.random_ordering,
random_seed=args.random_seed)
save_netcdf(result, args.output_filepath)
def process(cube,
raw_forecast=None,
no_of_realizations=None,
reordering=False,
rebadging=False,
random_seed=None,
ecc_bounds_warning=False):
"""Convert from probabilities to ensemble realizations.
Args:
cube (iris.cube.Cube):
Cube to be processed.
raw_forecast (iris.cube.Cube):
Cube of raw (not post processed) weather data.
This option is compulsory, if the reordering option is selected.
no_of_realizations (int):
Optional definition of the number of ensemble realizations to
be generated. These are generated though an intermediate
percentile representation. Theses percentiles will be
distributed regularly with the aim of dividing into blocks
of equal probability. If the reordering option is specified
and the number of realization is not given the number
of realizations is taken from the number of realizations
in the raw forecast cube.
Default is None.
reordering (bool):
The option used to create ensemble realizations from percentiles
by reordering the input percentiles based on the order of the
raw ensemble.
Default is False.
rebadging (bool):
Th option used to create ensemble realizations from percentiles
by rebadging the input percentiles.
Default is False.
random_seed (int):
Option to specify a value for the random seed for testing
purposes, otherwise the default random seed behaviours is
utilised. The random seed is used in the generation of the
random numbers used for splitting tied values within the raw
ensemble, so that the values from the input percentiles can
be ordered to match the raw ensemble.
Default is None.
ecc_bounds_warning (bool):
If True, where percentiles (calculated as an intermediate output
before realization) exceed to ECC bounds range, raises a warning
rather than an exception.
Default is False.
Returns:
iris.cube.Cube:
Processed result Cube.
Raises:
TypeError:
If rebadging is used with raw_forecast.
TypeError:
If rebadging is used with random_seed.
ValueError:
If raw_forecast isn't supplied when using reordering.
"""
if rebadging:
if raw_forecast is not None:
raise TypeError('rebadging cannot be used with raw_forecast.')
if random_seed is not None:
raise TypeError('rebadging cannot be used with random_seed.')
if reordering:
no_of_realizations = no_of_realizations
# If no_of_realizations is not given, take the number from the raw
# ensemble cube.
if no_of_realizations is None:
no_of_realizations = len(raw_forecast.coord("realization").points)
if raw_forecast is None:
message = ("You must supply a raw forecast cube if using the "
"reordering option.")
raise ValueError(message)
cube = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=ecc_bounds_warning).process(
cube, no_of_percentiles=no_of_realizations)
result = EnsembleReordering().process(cube,
raw_forecast,
random_ordering=False,
random_seed=random_seed)
elif rebadging:
cube = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=ecc_bounds_warning).process(
cube, no_of_percentiles=no_of_realizations)
result = RebadgePercentilesAsRealizations().process(cube)
return result
def forecast_dataframe_to_cube(
df: DataFrame,
training_dates: DatetimeIndex,
forecast_period: int,
) -> Cube:
"""Convert a forecast DataFrame into an iris Cube. The percentiles
within the forecast DataFrame are rebadged as realizations.
Args:
df:
DataFrame expected to contain the following columns: forecast,
blend_time, forecast_period, forecast_reference_time, time,
wmo_id, percentile, diagnostic, latitude, longitude, period,
height, cf_name, units. Any other columns are ignored.
training_dates:
Datetimes spanning the training period.
forecast_period:
Forecast period in seconds as an integer.
Returns:
Cube containing the forecasts from the training period.
"""
fp_point = pd.Timedelta(int(forecast_period), unit="seconds")
cubelist = CubeList()
for adate in training_dates:
time_df = df.loc[(df["time"] == adate)
& (df["forecast_period"] == fp_point)]
time_df = _preprocess_temporal_columns(time_df)
if time_df.empty:
continue
# The following columns are expected to contain one unique value
# per column.
for col in ["period", "height", "cf_name", "units", "diagnostic"]:
_unique_check(time_df, col)
if time_df["period"].isna().all():
time_bounds = None
fp_bounds = None
else:
period = time_df["period"].values[0]
time_bounds = [adate - period, adate]
fp_bounds = [fp_point - period, fp_point]
time_coord = _define_time_coord(adate, time_bounds)
height_coord = _define_height_coord(time_df["height"].values[0])
fp_coord = AuxCoord(
np.array(fp_point.total_seconds(),
dtype=TIME_COORDS["forecast_period"].dtype),
"forecast_period",
bounds=fp_bounds if fp_bounds is None else [
np.array(f.total_seconds(),
dtype=TIME_COORDS["forecast_period"].dtype)
for f in fp_bounds
],
units=TIME_COORDS["forecast_period"].units,
)
frt_coord = AuxCoord(
np.array(
time_df["forecast_reference_time"].values[0].timestamp(),
dtype=TIME_COORDS["forecast_reference_time"].dtype,
),
"forecast_reference_time",
units=TIME_COORDS["forecast_reference_time"].units,
)
for percentile in sorted(df["percentile"].unique()):
perc_coord = DimCoord(np.float32(percentile),
long_name="percentile",
units="%")
perc_df = time_df.loc[time_df["percentile"] == percentile]
cube = build_spotdata_cube(
perc_df["forecast"].astype(np.float32),
perc_df["cf_name"].values[0],
perc_df["units"].values[0],
perc_df["altitude"].astype(np.float32),
perc_df["latitude"].astype(np.float32),
perc_df["longitude"].astype(np.float32),
perc_df["wmo_id"].values.astype("U5"),
scalar_coords=[
time_coord,
frt_coord,
fp_coord,
perc_coord,
height_coord,
],
)
cubelist.append(cube)
if not cubelist:
return
cube = cubelist.merge_cube()
return RebadgePercentilesAsRealizations()(cube)
def process(cube: cli.inputcube,
raw_cube: cli.inputcube = None,
*,
realizations_count: int = None,
sampling_method='quantile',
ignore_ecc_bounds=False,
randomise=False,
random_seed: int = None,
realizations: cli.comma_separated_list = None):
"""Convert percentiles to ensemble realizations using Ensemble Coupla
Coupling.
Percentiles are either rebadged as realizations or reordered if the
raw_cube argument is given.
Args:
cube (iris.cube.Cube):
Cube expected to contain a percentiles coordinate.
raw_cube (iris.cube.Cube):
Cube of raw (not post processed) weather data.
If this argument is given ensemble realizations will be created
from percentiles by reshuffling them in correspondance to the rank
order of the raw ensemble. Otherwise, the percentiles are rebadged
as realizations.
realizations_count (int):
The number of percentiles to be generated. This is also equal to
the number of ensemble realizations that will be generated.
sampling_method (str):
Method to be used for generating the list of percentiles with
forecasts generated at each percentile. The options are "quantile"
and "random".
The quantile option produces equally spaced percentiles which is
the preferred option for full ensemble couple coupling with
reorder enabled.
ignore_ecc_bounds (bool):
If True where percentiles (calculated as an intermediate output
before realization) exceed the ECC bounds range, raises a
warning rather than an exception.
randomise (bool):
Reorder randomly, rather than using the rank order of the raw_cube.
random_seed (int):
Option to specify a value for the random seed for testing purposes,
otherwise, the default random seed behaviour is utilised.
The random seed is used in the generation of the random numbers
used for either the randomise option to order the input
percentiles randomly, rather than use the ordering from the
raw ensemble, or for splitting tied values within the raw ensemble
so that the values from the input percentiles can be ordered to
match the raw ensemble.
realizations (list of ints):
A list of ensemble realization numbers to use when rebadging the
percentiles into realizations.
Returns:
iris.cube.Cube:
The processed Cube.
"""
from improver.ensemble_copula_coupling.ensemble_copula_coupling import (
RebadgePercentilesAsRealizations, ResamplePercentiles,
EnsembleReordering)
if realizations:
realizations = [int(x) for x in realizations]
result = ResamplePercentiles(
ecc_bounds_warning=ignore_ecc_bounds).process(
cube, no_of_percentiles=realizations_count,
sampling=sampling_method)
if raw_cube:
result = EnsembleReordering().process(
result, raw_cube, random_ordering=randomise,
random_seed=random_seed)
else:
result = RebadgePercentilesAsRealizations().process(
result, ensemble_realization_numbers=realizations)
return result
def process(current_forecast, coeffs, landsea_mask, num_realizations=None,
random_ordering=False, random_seed=None,
ecc_bounds_warning=False, predictor_of_mean='mean'):
"""Applying coefficients for Ensemble Model Output Statistics.
Load in arguments for applying coefficients for Ensemble Model Output
Statistics (EMOS), otherwise known as Non-homogeneous Gaussian
Regression (NGR). The coefficients are applied to the forecast
that is supplied, so as to calibrate the forecast. The calibrated
forecast is written to a cube. If no coefficients are provided the input
forecast is returned unchanged.
Args:
current_forecast (iris.cube.Cube):
A Cube containing the forecast to be calibrated. The input format
could be either realizations, probabilities or percentiles.
coeffs (iris.cube.Cube or None):
A cube containing the coefficients used for calibration or None.
If none then then current_forecast is returned unchanged.
landsea_mask (iris.cube.Cube or None):
A cube containing the land-sea mask on the same domain as the
forecast that is to be calibrated. Land points are "
"specified by ones and sea points are specified by zeros. "
"If not None this argument will enable land-only calibration, in "
"which sea points are returned without the application of "
"calibration."
num_realizations (numpy.int32):
Optional argument to specify the number of ensemble realizations
to produce. If the current forecast is input as probabilities or
percentiles then this argument is used to create the requested
number of realizations. In addition, this argument is used to
construct the requested number of realizations from the mean and
variance output after applying the EMOS coefficients.
Default is None.
random_ordering (bool):
Option to reorder the post-processed forecasts randomly. If not
set, the ordering of the raw ensemble is used. This option is
only valid when the input format is realizations.
Default is False.
random_seed (int):
Option to specify a value for the random seed for testing
purposes, otherwise the default random seen behaviour is utilised.
The random seed is used in the generation of the random numbers
used for either the random_ordering option to order the input
percentiles randomly, rather than use the ordering from the raw
ensemble, or for splitting tied values within the raw ensemble,
so that the values from the input percentiles can be ordered to
match the raw ensemble.
Default is None.
ecc_bounds_warning (bool):
If True, where the percentiles exceed the ECC bounds range,
raises a warning rather than an exception. This occurs when the
current forecasts is in the form of probabilities and is
converted to percentiles, as part of converting the input
probabilities into realizations.
Default is False.
predictor_of_mean (str):
String to specify the predictor used to calibrate the forecast
mean. Currently the ensemble mean "mean" as the ensemble
realization "realization" are supported as options.
Default is 'mean'
Returns:
iris.cube.Cube:
The calibrated forecast cube.
Raises:
ValueError:
If the current forecast is a coefficients cube.
ValueError:
If the coefficients cube does not have the right name of
"emos_coefficients".
ValueError:
If the forecast type is 'percentiles' or 'probabilities' while no
num_realizations are given.
"""
if coeffs is None:
msg = ("There are no coefficients provided for calibration. The "
"uncalibrated forecast will be returned.")
warnings.warn(msg)
return current_forecast
elif coeffs.name() != 'emos_coefficients':
msg = ("The current coefficients cube does not have the "
"name 'emos_coefficients'")
raise ValueError(msg)
if current_forecast.name() == 'emos_coefficients':
msg = "The current forecast cube has the name 'emos_coefficients'"
raise ValueError(msg)
original_current_forecast = current_forecast.copy()
try:
find_percentile_coordinate(current_forecast)
input_forecast_type = "percentiles"
except CoordinateNotFoundError:
input_forecast_type = "realizations"
if current_forecast.name().startswith("probability_of"):
input_forecast_type = "probabilities"
# If probabilities, convert to percentiles.
conversion_plugin = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=ecc_bounds_warning)
elif input_forecast_type == "percentiles":
# If percentiles, resample percentiles so that the percentiles are
# evenly spaced.
conversion_plugin = ResamplePercentiles(
ecc_bounds_warning=ecc_bounds_warning)
# If percentiles, re-sample percentiles and then re-badge.
# If probabilities, generate percentiles and then re-badge.
if input_forecast_type in ["percentiles", "probabilities"]:
if not num_realizations:
raise ValueError(
"The current forecast has been provided as {0}. "
"These {0} need to be converted to realizations "
"for ensemble calibration. The num_realizations "
"argument is used to define the number of realizations "
"to construct from the input {0}, so if the "
"current forecast is provided as {0} then "
"num_realizations must be defined.".format(
input_forecast_type))
current_forecast = conversion_plugin.process(
current_forecast, no_of_percentiles=num_realizations)
current_forecast = (
RebadgePercentilesAsRealizations().process(current_forecast))
# Default number of ensemble realizations is the number in
# the raw forecast.
if not num_realizations:
num_realizations = len(
current_forecast.coord('realization').points)
# Apply coefficients as part of Ensemble Model Output Statistics (EMOS).
ac = ApplyCoefficientsFromEnsembleCalibration(
predictor_of_mean_flag=predictor_of_mean)
calibrated_predictor, calibrated_variance = ac.process(
current_forecast, coeffs, landsea_mask=landsea_mask)
# If input forecast is probabilities, convert output into probabilities.
# If input forecast is percentiles, convert output into percentiles.
# If input forecast is realizations, convert output into realizations.
if input_forecast_type == "probabilities":
result = GenerateProbabilitiesFromMeanAndVariance().process(
calibrated_predictor, calibrated_variance,
original_current_forecast)
elif input_forecast_type == "percentiles":
perc_coord = find_percentile_coordinate(original_current_forecast)
result = GeneratePercentilesFromMeanAndVariance().process(
calibrated_predictor, calibrated_variance,
percentiles=perc_coord.points)
elif input_forecast_type == "realizations":
# Ensemble Copula Coupling to generate realizations
# from mean and variance.
percentiles = GeneratePercentilesFromMeanAndVariance().process(
calibrated_predictor, calibrated_variance,
no_of_percentiles=num_realizations)
result = EnsembleReordering().process(
percentiles, current_forecast,
random_ordering=random_ordering, random_seed=random_seed)
return result
def process(cube: cli.inputcube,
raw_cube: cli.inputcube = None,
*,
realizations_count: int = None,
random_seed: int = None,
ignore_ecc_bounds=False):
"""Convert probabilities to ensemble realizations using Ensemble Copula
Coupling.
Probabilities are first converted to percentiles, which are then either
rebadged as realizations or reordered if the raw_cube argument is given.
Args:
cube (iris.cube.Cube):
Cube to be processed.
raw_cube (iris.cube.Cube):
Cube of raw (not post processed) weather data.
If this argument is given ensemble realizations will be created
from percentiles by reshuffling them in correspondence to the rank
order of the raw ensemble. Otherwise, the percentiles are rebadged
as realizations.
realizations_count (int):
Optional definition of the number of ensemble realizations to
be generated. These are generated though an intermediate
percentile representation. Theses percentiles will be
distributed regularly with the aim of dividing into blocks
of equal probability. If the raw_cube is given
and the number of realization is not given the number
of realizations is taken from the number of realizations
in the raw_cube.
random_seed (int):
Option to specify a value for the random seed for testing
purposes, otherwise the default random seed behaviours is
utilised. The random seed is used in the generation of the
random numbers used for splitting tied values within the raw
ensemble, so that the values from the input percentiles can
be ordered to match the raw ensemble.
ignore_ecc_bounds (bool):
If True, where percentiles (calculated as an intermediate output
before realization) exceed to ECC bounds range, raises a warning
rather than an exception.
Returns:
iris.cube.Cube:
Processed result Cube.
"""
from improver.ensemble_copula_coupling.ensemble_copula_coupling import (
ConvertProbabilitiesToPercentiles, RebadgePercentilesAsRealizations,
EnsembleReordering)
if realizations_count is None and raw_cube:
# If realizations_count is not given, take the number from the raw
# ensemble cube.
realizations_count = len(raw_cube.coord("realization").points)
result = ConvertProbabilitiesToPercentiles(
ecc_bounds_warning=ignore_ecc_bounds)(
cube, no_of_percentiles=realizations_count)
if raw_cube:
result = EnsembleReordering()(result,
raw_cube,
random_ordering=False,
random_seed=random_seed)
else:
result = RebadgePercentilesAsRealizations()(result)
return result
def main(argv=None):
"""Convert from probabilities to ensemble realizations via a CLI."""
cli_specific_arguments = [(['--no_of_realizations'], {
'metavar':
'NUMBER_OF_REALIZATIONS',
'default':
None,
'type':
int,
'help':
("Optional definition of the number of ensemble realizations to "
"be generated. These are generated through an intermediate "
"percentile representation. These percentiles will be "
"distributed regularly with the aim of dividing into blocks of "
"equal probability. If the reordering option is specified and "
"the number of realizations is not given then the number of "
"realizations is taken from the number of realizations in the "
"raw forecast NetCDF file.")
})]
cli_definition = {
'central_arguments': ('input_file', 'output_file'),
'specific_arguments':
cli_specific_arguments,
'description': ('Convert a dataset containing '
'probabilities into one containing '
'ensemble realizations.')
}
parser = ArgParser(**cli_definition)
# add mutually exculsive options rebadge and reorder.
# If reordering add option for raw ensemble - raise error if
# raw ens missing.
group = parser.add_mutually_exclusive_group(required=True)
group.add_argument('--reordering',
default=False,
action='store_true',
help='The option used to create ensemble realizations '
'from percentiles by reordering the input '
'percentiles based on the order of the '
'raw ensemble forecast.')
group.add_argument('--rebadging',
default=False,
action='store_true',
help='The option used to create ensemble realizations '
'from percentiles by rebadging the input '
'percentiles.')
# If reordering, we need a raw ensemble forecast.
reordering = parser.add_argument_group(
'Reordering options', 'Options for reordering the input percentiles '
'using the raw ensemble forecast as required to create ensemble '
'realizations.')
reordering.add_argument('--raw_forecast_filepath',
metavar='RAW_FORECAST_FILE',
help='A path to an raw forecast NetCDF file to be '
'processed. This option is compulsory, if the '
'reordering option is selected.')
reordering.add_argument(
'--random_seed',
default=None,
help='Option to specify a value for the random seed for testing '
'purposes, otherwise, the default random seed behaviour is '
'utilised. The random seed is used in the generation of the '
'random numbers used for splitting tied values '
'within the raw ensemble, so that the values from the input '
'percentiles can be ordered to match the raw ensemble.')
reordering.add_argument(
'--ecc_bounds_warning',
default=False,
action='store_true',
help='If True, where percentiles (calculated as an intermediate '
'output before realizations) exceed the ECC bounds range, raise '
'a warning rather than an exception.')
args = parser.parse_args(args=argv)
# CLI argument checking:
# Can only do one of reordering or rebadging: if options are passed which
# correspond to the opposite method, raise an exception.
# Note: Shouldn't need to check that both/none are set, since they are
# defined as mandatory, but mutually exclusive, options.
if args.rebadging:
if ((args.raw_forecast_filepath is not None)
or (args.random_seed is not None)):
parser.wrong_args_error('raw_forecast_filepath, random_seed',
'rebadging')
# Process the data
cube = load_cube(args.input_filepath)
if args.reordering:
if args.raw_forecast_filepath is None:
message = ("You must supply a raw forecast filepath if using the "
"reordering option.")
raise ValueError(message)
else:
raw_forecast = load_cube(args.raw_forecast_filepath)
try:
raw_forecast.coord("realization")
except CoordinateNotFoundError:
message = ("The netCDF file from the raw_forecast_filepath "
"must have a realization coordinate.")
raise ValueError(message)
no_of_realizations = args.no_of_realizations
# If no_of_realizations is not given, take the number from the raw
# ensemble cube.
if args.no_of_realizations is None:
no_of_realizations = len(raw_forecast.coord("realization").points)
cube = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=args.ecc_bounds_warning).process(
cube, no_of_percentiles=no_of_realizations)
cube = EnsembleReordering().process(cube,
raw_forecast,
random_ordering=False,
random_seed=args.random_seed)
elif args.rebadging:
cube = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=args.ecc_bounds_warning).process(
cube, no_of_percentiles=args.no_of_realizations)
cube = RebadgePercentilesAsRealizations().process(cube)
save_netcdf(cube, args.output_filepath)
def test_basic(self):
"""Test that a cube is produced with a realization dimension"""
result = Plugin().process(self.cube)
self.assertIsInstance(result, Cube)
self.assertIsInstance(result.coord("realization"), DimCoord)
self.assertEqual(result.coord("realization").units, "1")
def process(cube, raw_forecast=None, no_of_percentiles=None,
sampling_method='quantile', ecc_bounds_warning=False,
reordering=False, rebadging=False, random_ordering=False,
random_seed=None, realization_numbers=None):
"""Runs Ensemble Copula Coupling processing.
Converts a dataset containing percentiles into one containing ensemble
realizations using Ensemble Coupla Coupling.
Args:
cube (iris.cube.Cube):
Cube expected to contain a percentiles coordinate.
raw_forecast (iris.cube.Cube):
Cube of raw (not post processed) weather data.
This option is compulsory, if the reordering option is selected.
no_of_percentiles (int):
The number of percentiles to be generated. This is also equal to
the number of ensemble realizations that will be generated.
Default is None.
sampling_method (str):
Method to be used for generating the list of percentiles with
forecasts generated at each percentile. The options are "quantile"
and "random".
The quantile option produces equally spaced percentiles which is
the preferred option for full ensemble couple coupling with
reordering enabled.
Default is 'quantile'.
ecc_bounds_warning (bool):
If True where percentiles (calculated as an intermediate output
before realization) exceed the ECC bounds range, raises a
warning rather than an exception.
Default is False.
reordering (bool):
The option used to create ensemble realizations from percentiles
by reordering the input percentiles based on the order of the
raw ensemble forecast.
Default is False.
rebadging (bool):
The option used to create ensemble realizations from percentiles
by rebadging the input percentiles.
Default is False.
random_ordering (bool):
If random_ordering is True, the post-processed forecasts are
reordered randomly, rather than using the ordering of the
raw ensemble.
Default is False.
random_seed (int):
Option to specify a value for the random seed for testing purposes,
otherwise, the default random seed behaviour is utilised.
The random seed is used in the generation of the random numbers
used for either the random_ordering option to order the input
percentiles randomly, rather than use the ordering from the
raw ensemble, or for splitting tied values within the raw ensemble
so that the values from the input percentiles can be ordered to
match the raw ensemble.
Default is None.
realization_numbers (list of ints):
A list of ensemble realization numbers to use when rebadging the
percentiles into realizations.
Default is None.
Returns:
iris.cube.Cube:
The processed Cube.
"""
if reordering:
if realization_numbers is not None:
raise TypeError('realization_numbers cannot be used with '
'reordering.')
if rebadging:
if raw_forecast is not None:
raise TypeError('rebadging cannot be used with raw_forecast.')
if rebadging:
if random_ordering is not False:
raise TypeError('rebadging cannot be used with random_ordering.')
result = ResamplePercentiles(
ecc_bounds_warning=ecc_bounds_warning).process(
cube, no_of_percentiles=no_of_percentiles,
sampling=sampling_method)
if reordering:
result = EnsembleReordering().process(
result, raw_forecast, random_ordering=random_ordering,
random_seed=random_seed)
elif rebadging:
result = RebadgePercentilesAsRealizations().process(
result, ensemble_realization_numbers=realization_numbers)
return result
def test_number_of_realizations(self):
"""Check the values for the realization coordinate generated without
specifying the ensemble_realization_numbers argument."""
result = Plugin().process(self.cube)
self.assertArrayAlmostEqual(
result.coord("realization").points, np.array([0, 1, 2]))
def main(argv=None):
"""Do ensemble calibration using the EnsembleCalibration plugin.
"""
parser = ArgParser(
description='Apply the requested ensemble calibration method using '
'the current forecast (to be calibrated) in the form of '
'realizations, probabilities, or percentiles, historical '
'forecasts in the form of realizations and historical truth data '
'(to use in calibration). The mean and variance output from the '
'EnsembleCalibration plugin can be written to an output file '
'if required. If the current forecast is supplied in the form of '
'probabilities or percentiles, these are converted to realizations '
'prior to calibration. After calibration, the mean and variance '
'computed in the calibration are converted to match the format of the '
'current forecast i.e. if realizations are input, realizations '
'are output, if probabilities are input, probabilities are output, '
'and if percentiles are input, percentiles are output.'
'If realizations are input, realizations are regenerated using '
'Ensemble Copula Coupling.')
# Arguments for EnsembleCalibration
parser.add_argument(
'units',
metavar='UNITS_TO_CALIBRATE_IN',
help='The unit that calibration should be undertaken in. The current '
'forecast, historical forecast and truth will be converted as '
'required.')
parser.add_argument(
'distribution',
metavar='DISTRIBUTION',
choices=['gaussian', 'truncated gaussian'],
help='The distribution that will be used for calibration. This will '
'be dependent upon the input phenomenon. This has to be '
'supported by the minimisation functions in '
'ContinuousRankedProbabilityScoreMinimisers.')
# Filepaths for current, historic and truth data.
parser.add_argument(
'input_filepath',
metavar='INPUT_FILE',
help='A path to an input NetCDF file containing the current forecast '
'to be processed. The file provided could be in the form of '
'realizations, probabilities or percentiles.')
parser.add_argument(
'historic_filepath',
metavar='HISTORIC_DATA_FILE',
help='A path to an input NetCDF file containing the historic '
'forecast(s) used for calibration. The file provided must be in '
'the form of realizations.')
parser.add_argument(
'truth_filepath',
metavar='TRUTH_DATA_FILE',
help='A path to an input NetCDF file containing the historic truth '
'analyses used for calibration.')
parser.add_argument('output_filepath',
metavar='OUTPUT_FILE',
help='The output path for the processed NetCDF')
# Optional arguments.
parser.add_argument(
'--predictor_of_mean',
metavar='CALIBRATE_MEAN_FLAG',
choices=['mean', 'realizations'],
default='mean',
help='String to specify the input to calculate the calibrated mean. '
'Currently the ensemble mean ("mean") and the ensemble '
'realizations ("realizations") are supported as the predictors. '
'Default: "mean".')
parser.add_argument(
'--save_mean',
metavar='MEAN_FILE',
default=False,
help='Option to save the mean output from EnsembleCalibration plugin. '
'If used, a path to save the output to must be provided.')
parser.add_argument(
'--save_variance',
metavar='VARIANCE_FILE',
default=False,
help='Option to save the variance output from EnsembleCalibration '
'plugin. If used, a path to save the output to must be provided.')
parser.add_argument(
'--num_realizations',
metavar='NUMBER_OF_REALIZATIONS',
default=None,
type=np.int32,
help='Optional argument to specify the number of '
'ensemble realizations to produce. '
'If the current forecast is input as probabilities or '
'percentiles then this argument is used to create the requested '
'number of realizations. In addition, this argument is used to '
'construct the requested number of realizations from the mean '
'and variance output after applying the EMOS coefficients.'
'Default will be the number of realizations in the raw input '
'file, if realizations are provided as input, otherwise if the '
'input format is probabilities or percentiles, then an error '
'will be raised if no value is provided.')
parser.add_argument(
'--random_ordering',
default=False,
action='store_true',
help='Option to reorder the post-processed forecasts randomly. If not '
'set, the ordering of the raw ensemble is used. This option is '
'only valid when the input format is realizations.')
parser.add_argument(
'--random_seed',
metavar='RANDOM_SEED',
default=None,
help='Option to specify a value for the random seed for testing '
'purposes, otherwise, the default random seed behaviour is '
'utilised. The random seed is used in the generation of the '
'random numbers used for either the random_ordering option to '
'order the input percentiles randomly, rather than use the '
'ordering from the raw ensemble, or for splitting tied values '
'within the raw ensemble, so that the values from the input '
'percentiles can be ordered to match the raw ensemble.')
parser.add_argument(
'--ecc_bounds_warning',
default=False,
action='store_true',
help='If True, where the percentiles exceed the ECC bounds range, '
'raise a warning rather than an exception. This occurs when the '
'current forecast is in the form of probabilities and is '
'converted to percentiles, as part of converting the input '
'probabilities into realizations.')
parser.add_argument(
'--max_iterations',
metavar='MAX_ITERATIONS',
type=np.int32,
default=1000,
help='The maximum number of iterations allowed until the minimisation '
'has converged to a stable solution. If the maximum number of '
'iterations is reached, but the minimisation has not yet '
'converged to a stable solution, then the available solution is '
'used anyway, and a warning is raised. This may be modified for '
'testing purposes but otherwise kept fixed. If the '
'predictor_of_mean is "realizations", then the number of '
'iterations may require increasing, as there will be more '
'coefficients to solve for.')
args = parser.parse_args(args=argv)
current_forecast = load_cube(args.input_filepath)
historic_forecast = load_cube(args.historic_filepath)
truth = load_cube(args.truth_filepath)
original_current_forecast = current_forecast.copy()
msg = ("The current forecast has been provided as {0}. "
"These {0} need to be converted to realizations "
"for ensemble calibration. The args.num_realizations "
"argument is used to define the number of realizations "
"to construct from the input {0}, so if the "
"current forecast is provided as {0} then "
"args.num_realizations must be defined.")
try:
find_percentile_coordinate(current_forecast)
input_forecast_type = "percentiles"
except CoordinateNotFoundError:
input_forecast_type = "realizations"
if current_forecast.name().startswith("probability_of"):
input_forecast_type = "probabilities"
# If probabilities, convert to percentiles.
conversion_plugin = GeneratePercentilesFromProbabilities(
ecc_bounds_warning=args.ecc_bounds_warning)
elif input_forecast_type == "percentiles":
# If percentiles, resample percentiles so that the percentiles are
# evenly spaced.
conversion_plugin = ResamplePercentiles(
ecc_bounds_warning=args.ecc_bounds_warning)
# If percentiles, resample percentiles and then rebadge.
# If probabilities, generate percentiles and then rebadge.
if input_forecast_type in ["percentiles", "probabilities"]:
if not args.num_realizations:
raise ValueError(msg.format(input_forecast_type))
current_forecast = conversion_plugin.process(
current_forecast, no_of_percentiles=args.num_realizations)
current_forecast = (
RebadgePercentilesAsRealizations().process(current_forecast))
# Default number of ensemble realizations is the number in
# the raw forecast.
if not args.num_realizations:
args.num_realizations = len(
current_forecast.coord('realization').points)
# Ensemble-Calibration to calculate the mean and variance.
forecast_predictor, forecast_variance = EnsembleCalibration(
args.distribution,
args.units,
predictor_of_mean_flag=args.predictor_of_mean,
max_iterations=args.max_iterations).process(current_forecast,
historic_forecast, truth)
# If required, save the mean and variance.
if args.save_mean:
save_netcdf(forecast_predictor, args.save_mean)
if args.save_variance:
save_netcdf(forecast_variance, args.save_variance)
# If input forecast is probabilities, convert output into probabilities.
# If input forecast is percentiles, convert output into percentiles.
# If input forecast is realizations, convert output into realizations.
if input_forecast_type == "probabilities":
result = GenerateProbabilitiesFromMeanAndVariance().process(
forecast_predictor, forecast_variance, original_current_forecast)
elif input_forecast_type == "percentiles":
perc_coord = find_percentile_coordinate(original_current_forecast)
result = GeneratePercentilesFromMeanAndVariance().process(
forecast_predictor,
forecast_variance,
percentiles=perc_coord.points)
elif input_forecast_type == "realizations":
# Ensemble Copula Coupling to generate realizations
# from mean and variance.
percentiles = GeneratePercentilesFromMeanAndVariance().process(
forecast_predictor,
forecast_variance,
no_of_percentiles=args.num_realizations)
result = EnsembleReordering().process(
percentiles,
current_forecast,
random_ordering=args.random_ordering,
random_seed=args.random_seed)
save_netcdf(result, args.output_filepath)