dpyverification.scores.probabilistic#

Probabilistic verification scores.

For verification of probabilistic and ensemble forecasts, and probabilistic historical simulations of continuous variables.

For reference, see: https://scores.readthedocs.io/en/stable/included.html#probability

Classes

CrpsCDF(config)

Implementation for CRPS for probabilistic forecasts, expressed as cdf.

CrpsCDFConfig(*[, reduce_dims, ...])

Configuration for CRPS for CDF.

CrpsForEnsemble(config)

Implementation for CRPS for an ensemble.

CrpsForEnsembleConfig(*[, reduce_dims, ...])

Configuration for CRPS for ensemble.

RankHistogram(config)

Compute the rank histogram (Talagrand diagram) over the specified dimensions.

RankHistogramConfig(*[, reduce_dims, ...])

A rank histogram config element.
class dpyverification.scores.probabilistic.CrpsCDF(config)[source]#

Implementation for CRPS for probabilistic forecasts, expressed as cdf.

Parameters:

config (CrpsCDFConfig)

kind: str = 'crps_cdf'#
config_class#

alias of CrpsCDFConfig

supported_data_types: ClassVar[set[DataType]] = {DataType.simulated_forecast_probabilistic}#
compute(obs, sim)[source]#

Compute the CRPS for an ensemble of forecasts and observations.

Parameters:
Return type:

DataArray | Dataset

class dpyverification.scores.probabilistic.CrpsCDFConfig(*, reduce_dims=<factory>, score_adapter, general, verification_pair_ids=[], integration_method='exact', **extra_data)[source]#

Configuration for CRPS for CDF.

For reference, see: https://scores.readthedocs.io/en/stable/api.html#scores.probability.crps_cdf

Parameters:

  • reduce_dims (list[Literal[StandardDim.station, StandardDim.forecast_reference_time, StandardDim.forecast_period]])

  • score_adapter (Literal[ScoreKind.crps_cdf])

  • general (Annotated[GeneralInfoConfig, SkipJsonSchema()])

  • verification_pair_ids (list[str])

  • integration_method (Literal['exact', 'trapz'])

  • extra_data (Any)

score_adapter: Literal[ScoreKind.crps_cdf]#
integration_method: Annotated[Literal['exact', 'trapz'], FieldInfo(annotation=NoneType, required=True, description="The method of integration. 'exact' computes the exact integral, 'trapz' uses a trapezoidal rule and is an approximation of the CRPS.")]#
class dpyverification.scores.probabilistic.CrpsForEnsemble(config)[source]#

Implementation for CRPS for an ensemble.

Parameters:

config (CrpsForEnsembleConfig)

kind: str = 'crps_for_ensemble'#
config_class#

alias of CrpsForEnsembleConfig

supported_data_types: ClassVar[set[DataType]] = {DataType.simulated_forecast_ensemble}#
config: CrpsForEnsembleConfig#
compute(obs, sim)[source]#

Compute the CRPS for an ensemble of forecasts and observations.

Parameters:
Return type:

Dataset | DataArray

class dpyverification.scores.probabilistic.CrpsForEnsembleConfig(*, reduce_dims=<factory>, score_adapter, general, verification_pair_ids=[], method='ecdf', **extra_data)[source]#

Configuration for CRPS for ensemble.

For reference, see: See: https://scores.readthedocs.io/en/stable/api.html#scores.probability.crps_for_ensemble

Parameters:

  • reduce_dims (list[Literal[StandardDim.station, StandardDim.forecast_reference_time, StandardDim.forecast_period]])

  • score_adapter (Literal[ScoreKind.crps_for_ensemble])

  • general (Annotated[GeneralInfoConfig, SkipJsonSchema()])

  • verification_pair_ids (list[str])

  • method (Literal['ecdf', 'fair'])

  • extra_data (Any)

score_adapter: Literal[ScoreKind.crps_for_ensemble]#
method: Annotated[Literal['ecdf', 'fair'], FieldInfo(annotation=NoneType, required=False, default='ecdf', description='Method to compute the cumulative distribution function from an ensemble.')]#
class dpyverification.scores.probabilistic.RankHistogram(config)[source]#

Compute the rank histogram (Talagrand diagram) over the specified dimensions.

For external documentation, see below: https://xskillscore.readthedocs.io/en/stable/api/xskillscore.rank_histogram.html?highlight=rank%20histogram#xskillscore.rank_histogram

Parameters:

config (RankHistogramConfig)

kind: str = 'rank_histogram'#
config_class#

alias of RankHistogramConfig

supported_data_types: ClassVar[set[DataType]] = {DataType.simulated_forecast_ensemble}#
config: RankHistogramConfig#
compute(obs, sim)[source]#

Compute the histogram of ranks over the specified dimensions.

Parameters:
Return type:

DataArray | Dataset

class dpyverification.scores.probabilistic.RankHistogramConfig(*, reduce_dims=<factory>, score_adapter, general, verification_pair_ids=[], **extra_data)[source]#

A rank histogram config element.

Parameters:

  • reduce_dims (list[Literal[StandardDim.station, StandardDim.forecast_reference_time, StandardDim.forecast_period]])

  • score_adapter (Literal[ScoreKind.rank_histogram])

  • general (Annotated[GeneralInfoConfig, SkipJsonSchema()])

  • verification_pair_ids (list[str])

  • extra_data (Any)

score_adapter: Literal[ScoreKind.rank_histogram]#