Metrics¶
All high-level functions like verify()
and
bootstrap()
(for both HindcastEnsemble
and PerfectModelEnsemble
objects) have a metric
argument
that has to be called to determine which metric is used in computing predictability.
Note
We use the phrase ‘observations’ o
here to refer to the ‘truth’ data to which
we compare the forecast f
. These metrics can also be applied relative
to a control simulation, reconstruction, observations, etc. This would just
change the resulting score from quantifying skill to quantifying potential
predictability.
Internally, all metric functions require forecast
and observations
as inputs.
The dimension dim
has to be set to specify over which dimensions
the metric
is applied and are hence reduced.
See Comparisons for more on the dim
argument.
Deterministic¶
Deterministic metrics assess the forecast as a definite prediction of the future, rather than in terms of probabilities. Another way to look at deterministic metrics is that they are a special case of probabilistic metrics where a value of one is assigned to one category and zero to all others [Jolliffe2011].
Correlation Metrics¶
The below metrics rely fundamentally on correlations in their computation. In the
literature, correlation metrics are typically referred to as the Anomaly Correlation
Coefficient (ACC). This implies that anomalies in the forecast and observations
are being correlated. Typically, this is computed using the linear
Pearson Product-Moment Correlation.
However, climpred
also offers the
Spearman’s Rank Correlation.
Note that the p value associated with these correlations is computed via a separate
metric. Use pearson_r_p_value
or spearman_r_p_value
to compute p values assuming
that all samples in the correlated time series are independent. Use
pearson_r_eff_p_value
or spearman_r_eff_p_value
to account for autocorrelation
in the time series by calculating the effective_sample_size
.
Pearson Product-Moment Correlation Coefficient¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [1]: print(f"\n\nKeywords: {metric_aliases['pearson_r']}")
Keywords: ['pearson_r', 'pr', 'acc', 'pacc']
- climpred.metrics._pearson_r(forecast, verif, dim=None, **metric_kwargs)[source]¶
Pearson product-moment correlation coefficient.
A measure of the linear association between the forecast and verification data that is independent of the mean and variance of the individual distributions. This is also known as the Anomaly Correlation Coefficient (ACC) when correlating anomalies.
where and represent the standard deviation of the forecast and verification data over the experimental period, respectively.
Note
Use metric
pearson_r_p_value
orpearson_r_eff_p_value
to get the corresponding p value.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
pearson_r()
- Details:
minimum
-1.0
maximum
1.0
perfect
1.0
orientation
positive
Example
>>> HindcastEnsemble.verify(metric='pearson_r', comparison='e2o', ... alignment='same_verifs', dim=['init']) <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.9272 0.9145 0.9127 0.9319 ... 0.9315 0.9185 0.9112
Pearson Correlation p value¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [2]: print(f"\n\nKeywords: {metric_aliases['pearson_r_p_value']}")
Keywords: ['pearson_r_p_value', 'p_pval', 'pvalue', 'pval']
- climpred.metrics._pearson_r_p_value(forecast, verif, dim=None, **metric_kwargs)[source]¶
Probability that forecast and verification data are linearly uncorrelated.
Two-tailed p value associated with the Pearson product-moment correlation coefficient (
pearson_r
), assuming that all samples are independent. Usepearson_r_eff_p_value
to account for autocorrelation in the forecast and verification data.- Parameters
- Details:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
negative
Example
>>> HindcastEnsemble.verify(metric='pearson_r_p_value', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 5.779e-23 2.753e-21 4.477e-21 ... 8.7e-22 6.781e-21
Effective Sample Size¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [3]: print(f"\n\nKeywords: {metric_aliases['effective_sample_size']}")
Keywords: ['effective_sample_size', 'n_eff', 'eff_n']
- climpred.metrics._effective_sample_size(forecast, verif, dim=None, **metric_kwargs)[source]¶
Effective sample size for temporally correlated data.
Note
Weights are not included here due to the dependence on temporal autocorrelation.
Note
This metric can only be used for hindcast-type simulations.
The effective sample size extracts the number of independent samples between two time series being correlated. This is derived by assessing the magnitude of the lag-1 autocorrelation coefficient in each of the time series being correlated. A higher autocorrelation induces a lower effective sample size which raises the correlation coefficient for a given p value.
The effective sample size is used in computing the effective p value. See
pearson_r_eff_p_value
andspearman_r_eff_p_value
.where and are the lag-1 autocorrelation coefficients for the forecast and verification data.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
effective_sample_size()
- Details:
minimum
0.0
maximum
∞
perfect
N/A
orientation
positive
- Reference:
Bretherton, Christopher S., et al. “The effective number of spatial degrees of freedom of a time-varying field.” Journal of climate 12.7 (1999): 1990-2009.
Example
>>> HindcastEnsemble.verify(metric='effective_sample_size', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 5.0 4.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0
Pearson Correlation Effective p value¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [4]: print(f"\n\nKeywords: {metric_aliases['pearson_r_eff_p_value']}")
Keywords: ['pearson_r_eff_p_value', 'p_pval_eff', 'pvalue_eff', 'pval_eff']
- climpred.metrics._pearson_r_eff_p_value(forecast, verif, dim=None, **metric_kwargs)[source]¶
Probability that forecast and verification data are linearly uncorrelated, accounting for autocorrelation.
Note
Weights are not included here due to the dependence on temporal autocorrelation.
Note
This metric can only be used for hindcast-type simulations.
The effective p value is computed by replacing the sample size in the t-statistic with the effective sample size, . The same Pearson product-moment correlation coefficient is used as when computing the standard p value.
where is computed via the autocorrelation in the forecast and verification data.
where and are the lag-1 autocorrelation coefficients for the forecast and verification data.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
pearson_r_eff_p_value()
- Details:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
negative
Example
>>> HindcastEnsemble.verify(metric='pearson_r_eff_p_value', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.02333 0.08552 0.2679 ... 0.2369 0.2588 0.2703
- Reference:
Bretherton, Christopher S., et al. “The effective number of spatial degrees of freedom of a time-varying field.” Journal of climate 12.7 (1999): 1990-2009.
Spearman’s Rank Correlation Coefficient¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [5]: print(f"\n\nKeywords: {metric_aliases['spearman_r']}")
Keywords: ['spearman_r', 'sacc', 'sr']
- climpred.metrics._spearman_r(forecast, verif, dim=None, **metric_kwargs)[source]¶
Spearman’s rank correlation coefficient.
This correlation coefficient is nonparametric and assesses how well the relationship between the forecast and verification data can be described using a monotonic function. It is computed by first ranking the forecasts and verification data, and then correlating those ranks using the
pearson_r
correlation.This is also known as the anomaly correlation coefficient (ACC) when comparing anomalies, although the Pearson product-moment correlation coefficient (
pearson_r
) is typically used when computing the ACC.Note
Use metric
spearman_r_p_value
orspearman_r_eff_p_value
to get the corresponding p value.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
spearman_r()
- Details:
minimum
-1.0
maximum
1.0
perfect
1.0
orientation
positive
Example
>>> HindcastEnsemble.verify(metric='spearman_r', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.9336 0.9311 0.9293 0.9474 ... 0.9465 0.9346 0.9328
Spearman’s Rank Correlation Coefficient p value¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [6]: print(f"\n\nKeywords: {metric_aliases['spearman_r_p_value']}")
Keywords: ['spearman_r_p_value', 's_pval', 'spvalue', 'spval']
- climpred.metrics._spearman_r_p_value(forecast, verif, dim=None, **metric_kwargs)[source]¶
Probability that forecast and verification data are monotonically uncorrelated.
Two-tailed p value associated with the Spearman’s rank correlation coefficient (
spearman_r
), assuming that all samples are independent. Usespearman_r_eff_p_value
to account for autocorrelation in the forecast and verification data.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
spearman_r_p_value()
- Details:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
negative
Example
>>> HindcastEnsemble.verify(metric='spearman_r_p_value', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 6.248e-24 1.515e-23 ... 4.288e-24 8.254e-24
Spearman’s Rank Correlation Effective p value¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [7]: print(f"\n\nKeywords: {metric_aliases['spearman_r_eff_p_value']}")
Keywords: ['spearman_r_eff_p_value', 's_pval_eff', 'spvalue_eff', 'spval_eff']
- climpred.metrics._spearman_r_eff_p_value(forecast, verif, dim=None, **metric_kwargs)[source]¶
Probability that forecast and verification data are monotonically uncorrelated, accounting for autocorrelation.
Note
Weights are not included here due to the dependence on temporal autocorrelation.
Note
This metric can only be used for hindcast-type simulations.
The effective p value is computed by replacing the sample size in the t-statistic with the effective sample size, . The same Spearman’s rank correlation coefficient is used as when computing the standard p value.
where is computed via the autocorrelation in the forecast and verification data.
where and are the lag-1 autocorrelation coefficients for the forecast and verification data.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
spearman_r_eff_p_value()
- Details:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
negative
- Reference:
Bretherton, Christopher S., et al. “The effective number of spatial degrees of freedom of a time-varying field.” Journal of climate 12.7 (1999): 1990-2009.
Example
>>> HindcastEnsemble.verify(metric='spearman_r_eff_p_value', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.02034 0.0689 0.2408 ... 0.2092 0.2315 0.2347
Distance Metrics¶
This class of metrics simply measures the distance (or difference) between forecasted values and observed values.
Mean Squared Error (MSE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [8]: print(f"\n\nKeywords: {metric_aliases['mse']}")
Keywords: ['mse']
- climpred.metrics._mse(forecast, verif, dim=None, **metric_kwargs)[source]¶
Mean Sqaure Error (MSE).
The average of the squared difference between forecasts and verification data. This incorporates both the variance and bias of the estimator. Because the error is squared, it is more sensitive to large forecast errors than
mae
, and thus a more conservative metric. For example, a single error of 2°C counts the same as two 1°C errors when usingmae
. On the other hand, the 2°C error counts double formse
. See Jolliffe and Stephenson, 2011.- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
- Reference:
Ian T. Jolliffe and David B. Stephenson. Forecast Verification: A Practitioner’s Guide in Atmospheric Science. John Wiley & Sons, Ltd, Chichester, UK, December 2011. ISBN 978-1-119-96000-3 978-0-470-66071-3. URL: http://doi.wiley.com/10.1002/9781119960003.
Example
>>> HindcastEnsemble.verify(metric='mse', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.006202 0.006536 0.007771 ... 0.02417 0.02769
Root Mean Square Error (RMSE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [9]: print(f"\n\nKeywords: {metric_aliases['rmse']}")
Keywords: ['rmse']
- climpred.metrics._rmse(forecast, verif, dim=None, **metric_kwargs)[source]¶
Root Mean Sqaure Error (RMSE).
The square root of the average of the squared differences between forecasts and verification data.
- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
Example
>>> HindcastEnsemble.verify(metric='rmse', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.07875 0.08085 0.08815 ... 0.1371 0.1555 0.1664
Mean Absolute Error (MAE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [10]: print(f"\n\nKeywords: {metric_aliases['mae']}")
Keywords: ['mae']
- climpred.metrics._mae(forecast, verif, dim=None, **metric_kwargs)[source]¶
Mean Absolute Error (MAE).
The average of the absolute differences between forecasts and verification data. A more robust measure of forecast accuracy than
mse
which is sensitive to large outlier forecast errors.- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
- Reference:
Ian T. Jolliffe and David B. Stephenson. Forecast Verification: A Practitioner’s Guide in Atmospheric Science. John Wiley & Sons, Ltd, Chichester, UK, December 2011. ISBN 978-1-119-96000-3 978-0-470-66071-3. URL: http://doi.wiley.com/10.1002/9781119960003.
Example
>>> HindcastEnsemble.verify(metric='mae', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.06484 0.06684 0.07407 ... 0.1193 0.1361 0.1462
Median Absolute Error¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [11]: print(f"\n\nKeywords: {metric_aliases['median_absolute_error']}")
Keywords: ['median_absolute_error']
- climpred.metrics._median_absolute_error(forecast, verif, dim=None, **metric_kwargs)[source]¶
Median Absolute Error.
The median of the absolute differences between forecasts and verification data. Applying the median function to absolute error makes it more robust to outliers.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
median_absolute_error()
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
Example
>>> HindcastEnsemble.verify(metric='median_absolute_error', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.06077 0.06556 0.06368 ... 0.1131 0.142 0.1466
Spread¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [12]: print(f"\n\nKeywords: {metric_aliases['spread']}")
Keywords: ['spread']
- climpred.metrics._spread(forecast, verif, dim=None, **metric_kwargs)[source]¶
Ensemble spread taking the standard deviation over the member dimension.
- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
obs.std()
orientation
negative
Example
>>> HindcastEnsemble.verify(metric='spread', comparison='m2o', alignment='same_verifs', ... dim=['member','init']) <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.1468 0.1738 0.1922 0.2096 ... 0.2142 0.2178 0.2098
Multiplicative bias¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [13]: print(f"\n\nKeywords: {metric_aliases['mul_bias']}")
Keywords: ['mul_bias', 'm_b', 'multiplicative_bias']
- climpred.metrics._mul_bias(forecast, verif, dim=None, **metric_kwargs)[source]¶
Multiplicative bias.
- Parameters
- Details:
minimum
-∞
maximum
∞
perfect
1.0
orientation
None
Example
>>> HindcastEnsemble.verify(metric='multiplicative_bias', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.719 0.9991 1.072 1.434 ... 1.854 2.128 2.325 2.467
Normalized Distance Metrics¶
Distance metrics like mse
can be normalized to 1. The normalization factor
depends on the comparison type choosen. For example, the distance between an ensemble
member and the ensemble mean is half the distance of an ensemble member with other
ensemble members. See _get_norm_factor()
.
Normalized Mean Square Error (NMSE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [14]: print(f"\n\nKeywords: {metric_aliases['nmse']}")
Keywords: ['nmse', 'nev']
- climpred.metrics._nmse(forecast, verif, dim=None, **metric_kwargs)[source]¶
Normalized MSE (NMSE), also known as Normalized Ensemble Variance (NEV).
Mean Square Error (
mse
) normalized by the variance of the verification data.where is 1 when using comparisons involving the ensemble mean (
m2e
,e2c
,e2o
) and 2 when using comparisons involving individual ensemble members (m2c
,m2m
,m2o
). See_get_norm_factor()
.Note
climpred
uses a single-valued internal reference forecast for the NMSE, in the terminology of Murphy 1988. I.e., we use a single climatological variance of the verification data within the experimental window for normalizing MSE.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
comparison (str) – Name comparison needed for normalization factor fac, see
_get_norm_factor()
(Handled internally by the compute functions)
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
better than climatology
0.0 - 1.0
worse than climatology
> 1.0
- Reference:
Griffies, S. M., and K. Bryan. “A Predictability Study of Simulated North Atlantic Multidecadal Variability.” Climate Dynamics 13, no. 7–8 (August 1, 1997): 459–87. https://doi.org/10/ch4kc4.
Murphy, Allan H. “Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient.” Monthly Weather Review 116, no. 12 (December 1, 1988): 2417–24. https://doi.org/10/fc7mxd.
Example
>>> HindcastEnsemble.verify(metric='nmse', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.1732 0.1825 0.217 0.2309 ... 0.5247 0.6749 0.7732
Normalized Mean Absolute Error (NMAE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [15]: print(f"\n\nKeywords: {metric_aliases['nmae']}")
Keywords: ['nmae']
- climpred.metrics._nmae(forecast, verif, dim=None, **metric_kwargs)[source]¶
Normalized Mean Absolute Error (NMAE).
Mean Absolute Error (
mae
) normalized by the standard deviation of the verification data.where is 1 when using comparisons involving the ensemble mean (
m2e
,e2c
,e2o
) and 2 when using comparisons involving individual ensemble members (m2c
,m2m
,m2o
). See_get_norm_factor()
.Note
climpred
uses a single-valued internal reference forecast for the NMAE, in the terminology of Murphy 1988. I.e., we use a single climatological standard deviation of the verification data within the experimental window for normalizing MAE.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
comparison (str) – Name comparison needed for normalization factor fac, see
_get_norm_factor()
(Handled internally by the compute functions)
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
better than climatology
0.0 - 1.0
worse than climatology
> 1.0
- Reference:
Griffies, S. M., and K. Bryan. “A Predictability Study of Simulated North Atlantic Multidecadal Variability.” Climate Dynamics 13, no. 7–8 (August 1, 1997): 459–87. https://doi.org/10/ch4kc4.
Murphy, Allan H. “Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient.” Monthly Weather Review 116, no. 12 (December 1, 1988): 2417–24. https://doi.org/10/fc7mxd.
Example
>>> HindcastEnsemble.verify(metric='nmae', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.3426 0.3532 0.3914 0.3898 ... 0.6303 0.7194 0.7726
Normalized Root Mean Square Error (NRMSE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [16]: print(f"\n\nKeywords: {metric_aliases['nrmse']}")
Keywords: ['nrmse']
- climpred.metrics._nrmse(forecast, verif, dim=None, **metric_kwargs)[source]¶
Normalized Root Mean Square Error (NRMSE).
Root Mean Square Error (
rmse
) normalized by the standard deviation of the verification data.where is 1 when using comparisons involving the ensemble mean (
m2e
,e2c
,e2o
) and 2 when using comparisons involving individual ensemble members (m2c
,m2m
,m2o
). See_get_norm_factor()
.Note
climpred
uses a single-valued internal reference forecast for the NRMSE, in the terminology of Murphy 1988. I.e., we use a single climatological variance of the verification data within the experimental window for normalizing RMSE.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
comparison (str) – Name comparison needed for normalization factor fac, see
_get_norm_factor()
(Handled internally by the compute functions)
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
better than climatology
0.0 - 1.0
worse than climatology
> 1.0
- Reference:
Bushuk, Mitchell, Rym Msadek, Michael Winton, Gabriel Vecchi, Xiaosong Yang, Anthony Rosati, and Rich Gudgel. “Regional Arctic Sea–Ice Prediction: Potential versus Operational Seasonal Forecast Skill.” Climate Dynamics, June 9, 2018. https://doi.org/10/gd7hfq.
Hawkins, Ed, Steffen Tietsche, Jonathan J. Day, Nathanael Melia, Keith Haines, and Sarah Keeley. “Aspects of Designing and Evaluating Seasonal-to-Interannual Arctic Sea-Ice Prediction Systems.” Quarterly Journal of the Royal Meteorological Society 142, no. 695 (January 1, 2016): 672–83. https://doi.org/10/gfb3pn.
Murphy, Allan H. “Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient.” Monthly Weather Review 116, no. 12 (December 1, 1988): 2417–24. https://doi.org/10/fc7mxd.
Example
>>> HindcastEnsemble.verify(metric='nrmse', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.4161 0.4272 0.4658 0.4806 ... 0.7244 0.8215 0.8793
Mean Square Error Skill Score (MSESS)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [17]: print(f"\n\nKeywords: {metric_aliases['msess']}")
Keywords: ['msess', 'ppp', 'msss']
- climpred.metrics._msess(forecast, verif, dim=None, **metric_kwargs)[source]¶
Mean Squared Error Skill Score (MSESS).
where is 1 when using comparisons involving the ensemble mean (
m2e
,e2c
,e2o
) and 2 when using comparisons involving individual ensemble members (m2c
,m2m
,m2o
). See_get_norm_factor()
.This skill score can be intepreted as a percentage improvement in accuracy. I.e., it can be multiplied by 100%.
Note
climpred
uses a single-valued internal reference forecast for the MSSS, in the terminology of Murphy 1988. I.e., we use a single climatological variance of the verification data within the experimental window for normalizing MSE.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
comparison (str) – Name comparison needed for normalization factor fac, see
_get_norm_factor()
(Handled internally by the compute functions)
- Details:
minimum
-∞
maximum
1.0
perfect
1.0
orientation
positive
better than climatology
> 0.0
equal to climatology
0.0
worse than climatology
< 0.0
- Reference:
Griffies, S. M., and K. Bryan. “A Predictability Study of Simulated North Atlantic Multidecadal Variability.” Climate Dynamics 13, no. 7–8 (August 1, 1997): 459–87. https://doi.org/10/ch4kc4.
Murphy, Allan H. “Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient.” Monthly Weather Review 116, no. 12 (December 1, 1988): 2417–24. https://doi.org/10/fc7mxd.
Pohlmann, Holger, Michael Botzet, Mojib Latif, Andreas Roesch, Martin Wild, and Peter Tschuck. “Estimating the Decadal Predictability of a Coupled AOGCM.” Journal of Climate 17, no. 22 (November 1, 2004): 4463–72. https://doi.org/10/d2qf62.
Bushuk, Mitchell, Rym Msadek, Michael Winton, Gabriel Vecchi, Xiaosong Yang, Anthony Rosati, and Rich Gudgel. “Regional Arctic Sea–Ice Prediction: Potential versus Operational Seasonal Forecast Skill. Climate Dynamics, June 9, 2018. https://doi.org/10/gd7hfq.
Example
>>> HindcastEnsemble.verify(metric='msess', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.8268 0.8175 0.783 0.7691 ... 0.4753 0.3251 0.2268
Mean Absolute Percentage Error (MAPE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [18]: print(f"\n\nKeywords: {metric_aliases['mape']}")
Keywords: ['mape']
- climpred.metrics._mape(forecast, verif, dim=None, **metric_kwargs)[source]¶
Mean Absolute Percentage Error (MAPE).
Mean absolute error (
mae
) expressed as the fractional error relative to the verification data.- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
Example
>>> HindcastEnsemble.verify(metric='mape', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 1.536 1.21 1.421 1.149 ... 1.078 1.369 1.833 1.245
Symmetric Mean Absolute Percentage Error (sMAPE)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [19]: print(f"\n\nKeywords: {metric_aliases['smape']}")
Keywords: ['smape']
- climpred.metrics._smape(forecast, verif, dim=None, **metric_kwargs)[source]¶
Symmetric Mean Absolute Percentage Error (sMAPE).
Similar to the Mean Absolute Percentage Error (
mape
), but sums the forecast and observation mean in the denominator.- Parameters
- Details:
minimum
0.0
maximum
1.0
perfect
0.0
orientation
negative
See also
Example
>>> HindcastEnsemble.verify(metric='smape', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.3801 0.3906 0.4044 0.3819 ... 0.4822 0.5054 0.5295
Unbiased Anomaly Correlation Coefficient (uACC)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [20]: print(f"\n\nKeywords: {metric_aliases['uacc']}")
Keywords: ['uacc']
- climpred.metrics._uacc(forecast, verif, dim=None, **metric_kwargs)[source]¶
Bushuk’s unbiased Anomaly Correlation Coefficient (uACC).
This is typically used in perfect model studies. Because the perfect model Anomaly Correlation Coefficient (ACC) is strongly state dependent, a standard ACC (e.g. one computed using
pearson_r
) will be highly sensitive to the set of start dates chosen for the perfect model study. The Mean Square Skill Score (MESSS
) can be related directly to the ACC asMESSS = ACC^(2)
(see Murphy 1988 and Bushuk et al. 2019), so the unbiased ACC can be derived asuACC = sqrt(MESSS)
.where is 1 when using comparisons involving the ensemble mean (
m2e
,e2c
,e2o
) and 2 when using comparisons involving individual ensemble members (m2c
,m2m
,m2o
). See_get_norm_factor()
.Note
Because of the square root involved, any negative
MSESS
values are automatically converted to NaNs.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
comparison (str) – Name comparison needed for normalization factor
fac
, see_get_norm_factor()
(Handled internally by the compute functions)
- Details:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
positive
better than climatology
> 0.0
equal to climatology
0.0
- Reference:
Bushuk, Mitchell, Rym Msadek, Michael Winton, Gabriel Vecchi, Xiaosong Yang, Anthony Rosati, and Rich Gudgel. “Regional Arctic Sea–Ice Prediction: Potential versus Operational Seasonal Forecast Skill.” Climate Dynamics, June 9, 2018. https://doi.org/10/gd7hfq.
Allan H. Murphy. Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient. Monthly Weather Review, 116(12):2417–2424, December 1988. https://doi.org/10/fc7mxd.
Example
>>> HindcastEnsemble.verify(metric='uacc', comparison='e2o', alignment='same_verifs', ... dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.9093 0.9041 0.8849 0.877 ... 0.6894 0.5702 0.4763
Murphy Decomposition Metrics¶
Metrics derived in [Murphy1988] which decompose the MSESS
into a correlation term,
a conditional bias term, and an unconditional bias term. See
https://www-miklip.dkrz.de/about/murcss/ for a walk through of the decomposition.
Standard Ratio¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [21]: print(f"\n\nKeywords: {metric_aliases['std_ratio']}")
Keywords: ['std_ratio']
- climpred.metrics._std_ratio(forecast, verif, dim=None, **metric_kwargs)[source]¶
Ratio of standard deviations of the forecast over the verification data.
where and are the standard deviations of the forecast and the verification data over the experimental period, respectively.
- Parameters
- Details:
minimum
0.0
maximum
∞
perfect
1.0
orientation
N/A
- Reference:
Example
>>> HindcastEnsemble.verify(metric='std_ratio', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.7567 0.8801 0.9726 1.055 ... 1.075 1.094 1.055
Conditional Bias¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [22]: print(f"\n\nKeywords: {metric_aliases['conditional_bias']}")
Keywords: ['conditional_bias', 'c_b', 'cond_bias']
- climpred.metrics._conditional_bias(forecast, verif, dim=None, **metric_kwargs)[source]¶
Conditional bias between forecast and verification data.
where and are the standard deviations of the forecast and verification data over the experimental period, respectively.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
pearson_r()
:param and
Datasetstd()
:- Details:
minimum
-∞
maximum
1.0
perfect
0.0
orientation
negative
- Reference:
Example
>>> HindcastEnsemble.verify(metric='conditional_bias', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.1705 0.03435 -0.05988 ... -0.1436 -0.175 -0.1434
Unconditional Bias¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [23]: print(f"\n\nKeywords: {metric_aliases['unconditional_bias']}")
Keywords: ['unconditional_bias', 'u_b', 'a_b', 'bias', 'additive_bias']
Simple bias of the forecast minus the observations.
- climpred.metrics._unconditional_bias(forecast, verif, dim=None, **metric_kwargs)[source]¶
Unconditional additive bias.
- Parameters
- Details:
minimum
-∞
maximum
∞
perfect
0.0
orientation
negative
- Reference:
Example
>>> HindcastEnsemble.verify(metric='unconditional_bias', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 -0.01158 -0.02512 -0.0408 ... -0.1322 -0.1445
Conditional bias is removed by
remove_bias()
.>>> HindcastEnsemble = HindcastEnsemble.remove_bias(alignment='same_verifs') >>> HindcastEnsemble.verify(metric='unconditional_bias', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 3.203e-18 -1.068e-18 ... 2.882e-17 -2.776e-17
Bias Slope¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [24]: print(f"\n\nKeywords: {metric_aliases['bias_slope']}")
Keywords: ['bias_slope']
- climpred.metrics._bias_slope(forecast, verif, dim=None, **metric_kwargs)[source]¶
Bias slope between verification data and forecast standard deviations.
where is the Pearson product-moment correlation between the forecast and the verification data and and are the standard deviations of the verification data and forecast over the experimental period, respectively.
- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
pearson_r()
and
:param
std()
:- Details:
minimum
0.0
maximum
∞
perfect
1.0
orientation
negative
- Reference:
Example
>>> HindcastEnsemble.verify(metric='bias_slope', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.7016 0.8049 0.8877 0.9836 ... 1.002 1.004 0.961
Murphy’s Mean Square Error Skill Score¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [25]: print(f"\n\nKeywords: {metric_aliases['msess_murphy']}")
Keywords: ['msess_murphy', 'msss_murphy']
- climpred.metrics._msess_murphy(forecast, verif, dim=None, **metric_kwargs)[source]¶
Murphy’s Mean Square Error Skill Score (MSESS).
where represents the Pearson product-moment correlation coefficient between the forecast and verification data and represents the standard deviation of the verification data over the experimental period. See
conditional_bias
andunconditional_bias
for their respective formulations.- Parameters
forecast (xarray object) – Forecast.
verif (xarray object) – Verification data.
dim (str) – Dimension(s) to perform metric over.
metric_kwargs (dict) – see
pearson_r()
,
- Details:
minimum
-∞
maximum
1.0
perfect
1.0
orientation
positive
- Reference:
Murphy, Allan H. “Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient.” Monthly Weather Review 116, no. 12 (December 1, 1988): 2417–24. https://doi.org/10/fc7mxd.
Example
>>> HindcastEnsemble = HindcastEnsemble.remove_bias(alignment='same_verifs') >>> HindcastEnsemble.verify(metric='msess_murphy', comparison='e2o', ... alignment='same_verifs', dim='init') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.8306 0.8351 0.8295 0.8532 ... 0.8471 0.813 0.8097
Probabilistic¶
Probabilistic metrics include the spread of the ensemble simulations in their calculations and assign a probability value between 0 and 1 to their forecasts [Jolliffe2011].
Continuous Ranked Probability Score (CRPS)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [26]: print(f"\n\nKeywords: {metric_aliases['crps']}")
Keywords: ['crps']
- climpred.metrics._crps(forecast, verif, dim=None, **metric_kwargs)[source]¶
Continuous Ranked Probability Score (CRPS).
The CRPS can also be considered as the probabilistic Mean Absolute Error (
mae
). It compares the empirical distribution of an ensemble forecast to a scalar observation. Smaller scores indicate better skill.where is the cumulative distribution function (CDF) of the forecast (since the verification data are not assigned a probability), and H() is the Heaviside step function where the value is 1 if the argument is positive (i.e., the forecast overestimates verification data) or zero (i.e., the forecast equals verification data) and is 0 otherwise (i.e., the forecast is less than verification data).
Note
The CRPS is expressed in the same unit as the observed variable. It generalizes the Mean Absolute Error (MAE), and reduces to the MAE if the forecast is determinstic.
- Parameters
forecast (xr.object) – Forecast with member dim.
verif (xr.object) – Verification data without member dim.
dim (list of str) – Dimension to apply metric over. Expects at least member. Other dimensions are passed to xskillscore and averaged.
metric_kwargs (dict) – optional, see
crps_ensemble()
- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
- Reference:
Matheson, James E., and Robert L. Winkler. “Scoring Rules for Continuous Probability Distributions.” Management Science 22, no. 10 (June 1, 1976): 1087–96. https://doi.org/10/cwwt4g.
See also
crps_ensemble()
Example
>>> HindcastEnsemble.verify(metric='crps', comparison='m2o', dim='member', ... alignment='same_inits') <xarray.Dataset> Dimensions: (lead: 10, init: 52) Coordinates: * init (init) object 1954-01-01 00:00:00 ... 2005-01-01 00:00:00 * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead, init) float64 0.1722 0.1202 0.01764 ... 0.05428 0.1638
Continuous Ranked Probability Skill Score (CRPSS)¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [27]: print(f"\n\nKeywords: {metric_aliases['crpss']}")
Keywords: ['crpss']
- climpred.metrics._crpss(forecast, verif, dim=None, **metric_kwargs)[source]¶
Continuous Ranked Probability Skill Score.
This can be used to assess whether the ensemble spread is a useful measure for the forecast uncertainty by comparing the CRPS of the ensemble forecast to that of a reference forecast with the desired spread.
Note
When assuming a Gaussian distribution of forecasts, use default
gaussian=True
. If not gaussian, you may specify the distribution type, xmin/xmax/tolerance for integration (seecrps_quadrature()
).- Parameters
forecast (xr.object) – Forecast with
member
dim.verif (xr.object) – Verification data without
member
dim.dim (list of str) – Dimension to apply metric over. Expects at least member. Other dimensions are passed to xskillscore and averaged.
metric_kwargs (dict) –
optional gaussian (bool, optional): If
True
, assume Gaussian distribution forbaseline skill. Defaults to
True
.
- Details:
minimum
-∞
maximum
1.0
perfect
1.0
orientation
positive
better than climatology
> 0.0
worse than climatology
< 0.0
- Reference:
Matheson, James E., and Robert L. Winkler. “Scoring Rules for Continuous Probability Distributions.” Management Science 22, no. 10 (June 1, 1976): 1087–96. https://doi.org/10/cwwt4g.
Gneiting, Tilmann, and Adrian E Raftery. “Strictly Proper Scoring Rules, Prediction, and Estimation.” Journal of the American Statistical Association 102, no. 477 (March 1, 2007): 359–78. https://doi.org/10/c6758w.
Example
>>> HindcastEnsemble.verify(metric='crpss', comparison='m2o', ... alignment='same_inits', dim='member') <xarray.Dataset> Dimensions: (init: 52, lead: 10) Coordinates: * init (init) object 1954-01-01 00:00:00 ... 2005-01-01 00:00:00 * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead, init) float64 0.2644 0.3636 0.7376 ... 0.7526 0.7702 0.5126
>>> import scipy >>> PerfectModelEnsemble..isel(lead=[0, 1]).verify(metric='crpss', comparison='m2m', ... dim='member', gaussian=False, cdf_or_dist=scipy.stats.norm, xmin=-10, ... xmax=10, tol=1e-6) <xarray.Dataset> Dimensions: (init: 12, lead: 2, member: 9) Coordinates: * init (init) object 3014-01-01 00:00:00 ... 3257-01-01 00:00:00 * lead (lead) int64 1 2 * member (member) int64 1 2 3 4 5 6 7 8 9 Data variables: tos (lead, init, member) float64 0.9931 0.9932 0.9932 ... 0.9947 0.9947
See also
crps_ensemble()
Continuous Ranked Probability Skill Score Ensemble Spread¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [28]: print(f"\n\nKeywords: {metric_aliases['crpss_es']}")
Keywords: ['crpss_es']
- climpred.metrics._crpss_es(forecast, verif, dim=None, **metric_kwargs)[source]¶
Continuous Ranked Probability Skill Score Ensemble Spread.
If the ensemble variance is smaller than the observed
mse
, the ensemble is said to be under-dispersive (or overconfident). An ensemble with variance larger than the verification data indicates one that is over-dispersive (underconfident).- Parameters
forecast (xr.object) – Forecast with
member
dim.verif (xr.object) – Verification data without
member
dim.dim (list of str) – Dimension to apply metric over. Expects at least member. Other dimensions are passed to xskillscore and averaged.
metric_kwargs (dict) – see
crps_ensemble()
:param and
mse()
:- Details:
minimum
-∞
maximum
0.0
perfect
0.0
orientation
positive
under-dispersive
> 0.0
over-dispersive
< 0.0
- Reference:
Kadow, Christopher, Sebastian Illing, Oliver Kunst, Henning W. Rust, Holger Pohlmann, Wolfgang A. Müller, and Ulrich Cubasch. “Evaluation of Forecasts by Accuracy and Spread in the MiKlip Decadal Climate Prediction System.” Meteorologische Zeitschrift, December 21, 2016, 631–43. https://doi.org/10/f9jrhw.
Example
>>> HindcastEnsemble.verify(metric='crpss_es', comparison='m2o', ... alignment='same_verifs', dim='member') <xarray.Dataset> Dimensions: (init: 52, lead: 10) Coordinates: * init (init) object 1964-01-01 00:00:00 ... 2015-01-01 00:00:00 * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead, init) float64 -0.01121 -0.05575 ... -0.1263 -0.007483
Brier Score¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [29]: print(f"\n\nKeywords: {metric_aliases['brier_score']}")
Keywords: ['brier_score', 'brier', 'bs']
- climpred.metrics._brier_score(forecast, verif, dim=None, **metric_kwargs)[source]¶
Brier Score for binary events.
The Mean Square Error (
mse
) of probabilistic two-category forecasts where the verification data are either 0 (no occurrence) or 1 (occurrence) and forecast probability may be arbitrarily distributed between occurrence and non-occurrence. The Brier Score equals zero for perfect (single-valued) forecasts and one for forecasts that are always incorrect.where is the forecast probability of .
Note
The Brier Score requires that the observation is binary, i.e., can be described as one (a “hit”) or zero (a “miss”). So either provide a function with with binary outcomes logical in metric_kwargs or create binary verifs and probability forecasts by hindcast.map(logical).mean(‘member’). This Brier Score is not the original formula given in Brier’s 1950 paper.
- Parameters
forecast (xr.object) – Raw forecasts with
member
dimension if logical provided in metric_kwargs. Probability forecasts in [0,1] if logical is not provided.verif (xr.object) – Verification data without
member
dim. Raw verification if logical provided, else binary verification.dim (list or str) – Dimensions to aggregate. Requires member if logical provided in metric_kwargs to create probability forecasts. If logical not provided in metric_kwargs, should not include member.
metric_kwargs (dict) –
optional logical (callable): Function with bool result to be applied to verification
data and forecasts and then
mean('member')
to get forecasts and verification data in interval [0,1].see
brier_score()
- Details:
minimum
0.0
maximum
1.0
perfect
0.0
orientation
negative
- Reference:
https://www.nws.noaa.gov/oh/rfcdev/docs/ Glossary_Forecast_Verification_Metrics.pdf
See also
brier_score()
Example
Define a boolean/logical function for binary scoring:
>>> def pos(x): return x > 0 # checking binary outcomes
Option 1. Pass with keyword
logical
: (specifically designed forPerfectModelEnsemble
, where binary verification can only be created after comparison)>>> HindcastEnsemble.verify(metric='brier_score', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', logical=pos) <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.115 0.1121 0.1363 0.125 ... 0.1654 0.1675 0.1873
Option 2. Pre-process to generate a binary multi-member forecast and binary verification product:
>>> HindcastEnsemble.map(pos).verify(metric='brier_score', ... comparison='m2o', dim=['member', 'init'], alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.115 0.1121 0.1363 0.125 ... 0.1654 0.1675 0.1873
Option 3. Pre-process to generate a probability forecast and binary verification product. because
member
not present inhindcast
anymore, usecomparison='e2o'
anddim='init'
:>>> HindcastEnsemble.map(pos).mean('member').verify(metric='brier_score', ... comparison='e2o', dim='init', alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized' Data variables: SST (lead) float64 0.115 0.1121 0.1363 0.125 ... 0.1654 0.1675 0.1873
Threshold Brier Score¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [30]: print(f"\n\nKeywords: {metric_aliases['threshold_brier_score']}")
Keywords: ['threshold_brier_score', 'tbs']
- climpred.metrics._threshold_brier_score(forecast, verif, dim=None, **metric_kwargs)[source]¶
Brier score of an ensemble for exceeding given thresholds.
where is the cumulative distribution function (CDF) of the forecast distribution , is a point estimate of the true observation (observational error is neglected), denotes the Brier score and denotes the Heaviside step function, which we define here as equal to 1 for and 0 otherwise.
- Parameters
forecast (xr.object) – Forecast with
member
dim.verif (xr.object) – Verification data without
member
dim.dim (list of str) – Dimension to apply metric over. Expects at least member. Other dimensions are passed to xskillscore and averaged.
threshold (int, float, xr.object) – Threshold to check exceedance, see properscoring.threshold_brier_score.
metric_kwargs (dict) – optional, see
threshold_brier_score()
- Details:
minimum
0.0
maximum
1.0
perfect
0.0
orientation
negative
- Reference:
Brier, Glenn W. Verification of forecasts expressed in terms of probability.” Monthly Weather Review 78, no. 1 (1950). https://doi.org/10.1175/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2.
See also
threshold_brier_score()
Example
>>> # get threshold brier score for each init >>> HindcastEnsemble.verify(metric='threshold_brier_score', comparison='m2o', ... dim='member', threshold=.2, alignment='same_inits') <xarray.Dataset> Dimensions: (lead: 10, init: 52) Coordinates: * init (init) object 1954-01-01 00:00:00 ... 2005-01-01 00:00:00 * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 threshold float64 0.2 skill <U11 'initialized' Data variables: SST (lead, init) float64 0.0 0.0 0.0 0.0 0.0 ... 0.25 0.36 0.09 0.01
>>> # multiple thresholds averaging over init dimension >>> HindcastEnsemble.verify(metric='threshold_brier_score', comparison='m2o', ... dim=['member', 'init'], threshold=[.2, .3], alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, threshold: 2) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * threshold (threshold) float64 0.2 0.3 skill <U11 'initialized' Data variables: SST (lead, threshold) float64 0.08712 0.005769 ... 0.1312 0.01923
Ranked Probability Score¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [31]: print(f"\n\nKeywords: {metric_aliases['rps']}")
Keywords: ['rps']
- climpred.metrics._rps(forecast, verif, dim=None, **metric_kwargs)[source]¶
Ranked Probability Score.
- Parameters
Note
If
category_edges
is xr.Dataset or tuple of xr.Datasets, climpred will broadcast the grouped dimensionsseason
,month
,weekofyear
,dayfofyear
onto the dimensionsinit
for forecast andtime
for observations. seeclimpred.utils.broadcast_time_grouped_to_time
.- Details:
minimum
0.0
maximum
∞
perfect
0.0
orientation
negative
See also
Example
>>> category_edges = np.array([-.5, 0., .5, 1.]) >>> HindcastEnsemble.verify(metric='rps', comparison='m2o', dim=['member', 'init'], ... alignment='same_verifs', category_edges=category_edges) <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 observations_category_edge <U67 '[-np.inf, -0.5), [-0.5, 0.0), [0.0, 0.5... forecasts_category_edge <U67 '[-np.inf, -0.5), [-0.5, 0.0), [0.0, 0.5... skill <U11 'initialized' Data variables: SST (lead) float64 0.115 0.1123 ... 0.1687 0.1875
Provide category_edges as xr.Dataset for category_edges varying along dimensions.
>>> category_edges = xr.DataArray([9.5, 10., 10.5, 11.], dims='category_edge').assign_coords(category_edge=[9.5, 10., 10.5, 11.]).to_dataset(name='tos') >>> # category_edges = np.array([9.5, 10., 10.5, 11.]) # identical >>> PerfectModelEnsemble.verify(metric='rps', comparison='m2c', ... dim=['member','init'], category_edges=category_edges) <xarray.Dataset> Dimensions: (lead: 20) Coordinates: * lead (lead) int64 1 2 3 4 5 6 7 ... 15 16 17 18 19 20 observations_category_edge <U71 '[-np.inf, 9.5), [9.5, 10.0), [10.0, 10.... forecasts_category_edge <U71 '[-np.inf, 9.5), [9.5, 10.0), [10.0, 10.... Data variables: tos (lead) float64 0.08951 0.1615 ... 0.1399 0.2274
Provide category_edges as tuple for different category_edges to categorize forecasts and observations.
>>> q = [1 / 3, 2 / 3] >>> forecast_edges = HindcastEnsemble.get_initialized().groupby('init.month').quantile(q=q, dim=['init','member']).rename({'quantile':'category_edge'}) >>> obs_edges = HindcastEnsemble.get_observations().groupby('time.month').quantile(q=q, dim='time').rename({'quantile':'category_edge'}) >>> category_edges = (obs_edges, forecast_edges) >>> HindcastEnsemble.verify(metric='rps', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', ... category_edges=category_edges) <xarray.Dataset> Dimensions: (lead: 10) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 observations_category_edge <U101 '[-np.inf, 0.3333333333333333), [0.3333... forecasts_category_edge <U101 '[-np.inf, 0.3333333333333333), [0.3333... skill <U11 'initialized' Data variables: SST (lead) float64 0.1248 0.1756 ... 0.3081 0.3413
Reliability¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [32]: print(f"\n\nKeywords: {metric_aliases['reliability']}")
Keywords: ['reliability']
- climpred.metrics._reliability(forecast, verif, dim=None, **metric_kwargs)[source]¶
Returns the data required to construct the reliability diagram for an event. The the relative frequencies of occurrence of an event for a range of forecast probability bins.
- Parameters
forecast (xr.object) – Raw forecasts with
member
dimension if logical provided in metric_kwargs. Probability forecasts in [0,1] if logical is not provided.verif (xr.object) – Verification data without
member
dim. Raw verification if logical provided, else binary verification.dim (list or str) – Dimensions to aggregate. Requires member if logical provided in metric_kwargs to create probability forecasts. If logical not provided in metric_kwargs, should not include member.
logical (callable, optional) – Function with bool result to be applied to verification data and forecasts and then
mean('member')
to get forecasts and verification data in interval [0,1]. Passed via metric_kwargs.probability_bin_edges (array_like, optional) – Probability bin edges used to compute the reliability. Bins include the left most edge, but not the right. Passed via metric_kwargs. Defaults to 6 equally spaced edges between 0 and 1+1e-8.
- Returns
- The relative frequency of occurrence for each
probability bin
- Return type
reliability (xr.object)
- Details:
perfect
flat distribution
See also
Example
Define a boolean/logical function for binary scoring:
>>> def pos(x): return x > 0 # checking binary outcomes
Option 1. Pass with keyword
logical
: (especially designed forPerfectModelEnsemble
, where binary verification can only be created after comparison))>>> HindcastEnsemble.verify(metric='reliability', comparison='m2o', ... dim=['member','init'], alignment='same_verifs', logical=pos) <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 SST_samples (forecast_probability) float64 22.0 5.0 1.0 3.0 21.0 skill <U11 'initialized' Data variables: SST (lead, forecast_probability) float64 0.09091 ... 1.0
Option 2. Pre-process to generate a binary forecast and verification product:
>>> HindcastEnsemble.map(pos).verify(metric='reliability', ... comparison='m2o', dim=['init', 'member'], alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 SST_samples (forecast_probability) float64 22.0 5.0 1.0 3.0 21.0 skill <U11 'initialized' Data variables: SST (lead, forecast_probability) float64 0.09091 ... 1.0
Option 3. Pre-process to generate a probability forecast and binary verification product. because
member
not present inhindcast
, usecomparison='e2o'
anddim='init'
:>>> HindcastEnsemble.map(pos).mean('member').verify(metric='reliability', ... comparison='e2o', dim='init', alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 SST_samples (forecast_probability) float64 22.0 5.0 1.0 3.0 21.0 skill <U11 'initialized' Data variables: SST (lead, forecast_probability) float64 0.09091 ... 1.0
Discrimination¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [33]: print(f"\n\nKeywords: {metric_aliases['discrimination']}")
Keywords: ['discrimination']
- climpred.metrics._discrimination(forecast, verif, dim=None, **metric_kwargs)[source]¶
Returns the data required to construct the discrimination diagram for an event. The histogram of forecasts likelihood when observations indicate an event has occurred and has not occurred.
- Parameters
forecast (xr.object) – Raw forecasts with
member
dimension if logical provided in metric_kwargs. Probability forecasts in [0,1] if logical is not provided.verif (xr.object) – Verification data without
member
dim. Raw verification if logical provided, else binary verification.dim (list or str) – Dimensions to aggregate. Requires member if logical provided in metric_kwargs to create probability forecasts. If logical not provided in metric_kwargs, should not include member. At least one dimension other than member is required.
logical (callable, optional) – Function with bool result to be applied to verification data and forecasts and then
mean('member')
to get forecasts and verification data in interval [0,1]. Passed via metric_kwargs.probability_bin_edges (array_like, optional) – Probability bin edges used to compute the histograms. Bins include the left most edge, but not the right. Passed via metric_kwargs. Defaults to 6 equally spaced edges between 0 and 1+1e-8.
- Returns
Discrimination (xr.object) with added dimension “event” containing the histograms of forecast probabilities when the event was observed and not observed
- Details:
perfect
distinct distributions
See also
Example
Define a boolean/logical function for binary scoring:
>>> def pos(x): return x > 0 # checking binary outcomes
Option 1. Pass with keyword
logical
: (especially designed forPerfectModelEnsemble
, where binary verification can only be created after comparison)>>> HindcastEnsemble.verify(metric='discrimination', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', logical=pos) <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5, event: 2) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 * event (event) bool True False skill <U11 'initialized' Data variables: SST (lead, event, forecast_probability) float64 0.07407...
Option 2. Pre-process to generate a binary forecast and verification product:
>>> HindcastEnsemble.map(pos).verify(metric='discrimination', ... comparison='m2o', dim=['member','init'], alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5, event: 2) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 * event (event) bool True False skill <U11 'initialized' Data variables: SST (lead, event, forecast_probability) float64 0.07407...
Option 3. Pre-process to generate a probability forecast and binary verification product. because
member
not present inhindcast
, usecomparison='e2o'
anddim='init'
:>>> HindcastEnsemble.map(pos).mean('member').verify(metric='discrimination', ... comparison='e2o', dim='init', alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, forecast_probability: 5, event: 2) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * forecast_probability (forecast_probability) float64 0.1 0.3 0.5 0.7 0.9 * event (event) bool True False skill <U11 'initialized' Data variables: SST (lead, event, forecast_probability) float64 0.07407...
Rank Histogram¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [34]: print(f"\n\nKeywords: {metric_aliases['rank_histogram']}")
Keywords: ['rank_histogram']
- climpred.metrics._rank_histogram(forecast, verif, dim=None, **metric_kwargs)[source]¶
Rank histogram or Talagrand diagram.
- Parameters
- Details:
perfect
flat distribution
See also
Example
>>> HindcastEnsemble.verify(metric='rank_histogram', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs') <xarray.Dataset> Dimensions: (lead: 10, rank: 11) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 * rank (rank) float64 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 skill <U11 'initialized' Data variables: SST (lead, rank) int64 12 3 2 1 1 3 1 2 6 5 16 ... 0 1 0 0 3 0 2 6 6 34
>>> PerfectModelEnsemble.verify(metric='rank_histogram', comparison='m2c', ... dim=['member', 'init']) <xarray.Dataset> Dimensions: (lead: 20, rank: 10) Coordinates: * lead (lead) int64 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 * rank (rank) float64 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 Data variables: tos (lead, rank) int64 1 4 2 1 2 1 0 0 0 1 2 ... 0 2 0 1 2 1 0 3 1 2 0
Contingency-based metrics¶
A number of metrics can be derived from a contingency table. To use this in climpred
, run .verify(metric='contingency', score=...)
where score can be chosen from xskillscore.
- climpred.metrics._contingency(forecast, verif, score='table', dim=None, **metric_kwargs)[source]¶
Contingency table.
- Parameters
forecast (xr.object) – Raw forecasts.
verif (xr.object) – Verification data.
score (str) – Score derived from contingency table. Attribute from
Contingency
. Usescore=table
to return a contingency table or any other contingency score, e.g.score=hit_rate
.observation_category_edges (array_like) – Category bin edges used to compute the observations CDFs. Bins include the left most edge, but not the right. Passed via metric_kwargs.
forecast_category_edges (array_like) – Category bin edges used to compute the forecast CDFs. Bins include the left most edge, but not the right. Passed via metric_kwargs
See also
Example
>>> category_edges = np.array([-0.5, 0.0, 0.5, 1.0]) >>> HindcastEnsemble.verify(metric='contingency', score='table', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', ... observation_category_edges=category_edges, ... forecast_category_edges=category_edges).isel(lead=[0, 1]).SST <xarray.DataArray 'SST' (lead: 2, observations_category: 3, forecasts_category: 3)> array([[[221, 29, 0], [ 53, 217, 0], [ 0, 0, 0]], [[234, 16, 0], [ 75, 194, 1], [ 0, 0, 0]]]) Coordinates: * lead (lead) int32 1 2 observations_category_bounds (observations_category) <U11 '[-0.5, 0.0)' ... forecasts_category_bounds (forecasts_category) <U11 '[-0.5, 0.0)' ...... * observations_category (observations_category) int64 1 2 3 * forecasts_category (forecasts_category) int64 1 2 3 skill <U11 'initialized'
>>> # contingency-based dichotomous accuracy score >>> category_edges = np.array([9.5, 10.0, 10.5]) >>> PerfectModelEnsemble.verify(metric='contingency', score='hit_rate', ... comparison='m2c', dim=['member','init'], ... observation_category_edges=category_edges, ... forecast_category_edges=category_edges) <xarray.Dataset> Dimensions: (lead: 20) Coordinates: * lead (lead) int64 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Data variables: tos (lead) float64 1.0 1.0 1.0 1.0 0.9091 ... 1.0 1.0 1.0 nan 1.0
Receiver Operating Characteristic¶
# Enter any of the below keywords in ``metric=...`` for the compute functions.
In [35]: print(f"\n\nKeywords: {metric_aliases['roc']}")
Keywords: ['roc']
- climpred.metrics._roc(forecast, verif, dim=None, **metric_kwargs)[source]¶
Receiver Operating Characteristic.
- Parameters
observations (xarray.object) – Labeled array(s) over which to apply the function. If
bin_edges=='continuous'
, observations are binary.forecasts (xarray.object) – Labeled array(s) over which to apply the function. If
bin_edges=='continuous'
, forecasts are probabilities.dim (str, list of str) – The dimension(s) over which to aggregate. Defaults to None, meaning aggregation over all dims other than
lead
.logical (callable, optional) – Function with bool result to be applied to verification data and forecasts and then
mean('member')
to get forecasts and verification data in interval [0,1]. Passed via metric_kwargs.bin_edges (array_like, str) – Bin edges for categorising observations and forecasts. Similar to np.histogram, all but the last (righthand-most) bin include the left edge and exclude the right edge. The last bin includes both edges.
bin_edges
will be sorted in ascending order. Ifbin_edges=='continuous'
, calculatebin_edges
from forecasts, equal tosklearn.metrics.roc_curve(f_boolean, o_prob)
. Passed via metric_kwargs. Defaults to ‘continuous’.drop_intermediate (bool) – Whether to drop some suboptimal thresholds which would not appear on a plotted ROC curve. This is useful in order to create lighter ROC curves. Defaults to False. Defaults to
True
insklearn.metrics.roc_curve
. Passed via metric_kwargs.return_results (str) –
Passed via metric_kwargs. Defaults to ‘area’. Specify how return is structed:
’area’: return only the
area under curve
of ROC’all_as_tuple’: return
true positive rate
andfalse positive rate
at each bin and area under the curve of ROC as tuple’all_as_metric_dim’: return
true positive rate
andfalse positive rate
at each bin andarea under curve
of ROC concatinated into newmetric
dimension
- Returns
- reduced by dimensions
dim
, seereturn_results
parameter.
true positive rate
andfalse positive rate
containprobability_bin
dimension with ascendingbin_edges
as coordinates.
- reduced by dimensions
- Return type
roc (xr.object)
- Details for area under curve:
minimum
0.0
maximum
1.0
perfect
1.0
orientation
positive
See also
Example
>>> bin_edges = np.array([-0.5, 0.0, 0.5, 1.0]) >>> HindcastEnsemble.verify(metric='roc', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', ... bin_edges=bin_edges, ... ).SST <xarray.DataArray 'SST' (lead: 10)> array([0.84385185, 0.82841667, 0.81358547, 0.8393463 , 0.82551752, 0.81987778, 0.80719573, 0.80081909, 0.79046553, 0.78037564]) Coordinates: * lead (lead) int32 1 2 3 4 5 6 7 8 9 10 skill <U11 'initialized'
Get area under the curve, false positive rate and true positive rate as
metric
dimension by specifyingreturn_results='all_as_metric_dim'
:>>> def f(ds): return ds > 0 >>> HindcastEnsemble.map(f).verify(metric='roc', comparison='m2o', ... dim=['member', 'init'], alignment='same_verifs', ... bin_edges='continuous', return_results='all_as_metric_dim' ... ).SST.isel(lead=[0, 1]) <xarray.DataArray 'SST' (lead: 2, metric: 3, probability_bin: 3)> array([[[0. , 0.116 , 1. ], [0. , 0.8037037 , 1. ], [0.84385185, 0.84385185, 0.84385185]], [[0. , 0.064 , 1. ], [0. , 0.72222222, 1. ], [0.82911111, 0.82911111, 0.82911111]]]) Coordinates: * probability_bin (probability_bin) float64 2.0 1.0 0.0 * lead (lead) int32 1 2 * metric (metric) <U19 'false positive rate' ... 'area under curve' skill <U11 'initialized'
User-defined metrics¶
You can also construct your own metrics via the climpred.metrics.Metric
class.
|
Master class for all metrics. |
First, write your own metric function, similar to the existing ones with required
arguments forecast
, observations
, dim=None
, and **metric_kwargs
:
from climpred.metrics import Metric
def _my_msle(forecast, observations, dim=None, **metric_kwargs):
"""Mean squared logarithmic error (MSLE).
https://peltarion.com/knowledge-center/documentation/modeling-view/build-an-ai-model/loss-functions/mean-squared-logarithmic-error."""
# function
return ( (np.log(forecast + 1) + np.log(observations + 1) ) ** 2).mean(dim)
Then initialize this metric function with climpred.metrics.Metric
:
_my_msle = Metric(
name='my_msle',
function=_my_msle,
probabilistic=False,
positive=False,
unit_power=0,
)
Finally, compute skill based on your own metric:
skill = hindcast.verify(metric=_my_msle, comparison='e2o', alignment='same_verif', dim='init')
Once you come up with an useful metric for your problem, consider contributing this metric to climpred, so all users can benefit from your metric, see Contributing to xarray.
References¶
- Jolliffe2011(1,2)
Ian T. Jolliffe and David B. Stephenson. Forecast Verification: A Practitioner’s Guide in Atmospheric Science. John Wiley & Sons, Ltd, Chichester, UK, December 2011. ISBN 978-1-119-96000-3 978-0-470-66071-3. URL: http://doi.wiley.com/10.1002/9781119960003.
- Murphy1988
Allan H. Murphy. Skill Scores Based on the Mean Square Error and Their Relationships to the Correlation Coefficient. Monthly Weather Review, 116(12):2417–2424, December 1988. https://doi.org/10/fc7mxd.