tstoolbox.tstoolbox.gof

tstoolbox.tstoolbox.gof(obs_col=1, sim_col=2, stats='default', replace_nan=None, replace_inf=None, remove_neg=False, remove_zero=False, start_date=None, end_date=None, input_ts=None, round_index=None, clean=False, index_type='datetime', source_units=None, target_units=None, kge_sr=1.0, kge09_salpha=1.0, kge12_sgamma=1.0, kge_sbeta=1.0)

Will calculate goodness of fit statistics between two time-series.

The first time series must be the observed, the second the simulated series. You can only give two time-series.

Parameters:

• obs_col – If integer represents the column number of standard input. Can be If integer represents the column number of standard input. Can be a csv, wdm, hdf or xlsx file following format specified in ‘tstoolbox read …’.

• sim_col – If integer represents the column number of standard input. Can be a csv, wdm, hdf or xlsx file following format specified in ‘tstoolbox read …’.

• stats (Union[None, Literal['default', 'all', 'bias', 'pc_bias', 'apc_bias', 'rmsd', 'crmsd', 'corrcoef', 'coefdet', 'murphyss', 'nse', 'kge09', 'kge12', 'index_agreement', 'brierss', 'mae', 'mean', 'stdev', 'acc', 'd1', 'd1_p', 'd', 'dmod', 'drel', 'dr', 'ed', 'g_mean_diff', 'h10_mahe', 'h10_mhe', 'h10_rmshe', 'h1_mahe', 'h1_mhe', 'h1_rmshe', 'h2_mahe', 'h2_mhe', 'h2_rmshe', 'h3_mahe', 'h3_mhe', 'h3_rmshe', 'h4_mahe', 'h4_mhe', 'h4_rmshe', 'h5_mahe', 'h5_mhe', 'h5_rmshe', 'h6_mahe', 'h6_mhe', 'h6_rmshe', 'h7_mahe', 'h7_mhe', 'h7_rmshe', 'h8_mahe', 'h8_mhe', 'h8_rmshe', 'irmse', 'lm_index', 'maape', 'male', 'mapd', 'mape', 'mase', 'mb_r', 'mdae', 'mde', 'mdse', 'mean_var', 'me', 'mle', 'mse', 'msle', 'ned', 'nrmse_iqr', 'nrmse_mean', 'nrmse_range', 'nse_mod', 'nse_rel', 'rmse', 'rmsle', 'sa', 'sc', 'sga', 'sid', 'smape1', 'smape2', 'spearman_r', 've', 'watt_m'], List[Literal['default', 'all', 'bias', 'pc_bias', 'apc_bias', 'rmsd', 'crmsd', 'corrcoef', 'coefdet', 'murphyss', 'nse', 'kge09', 'kge12', 'index_agreement', 'brierss', 'mae', 'mean', 'stdev', 'acc', 'd1', 'd1_p', 'd', 'dmod', 'drel', 'dr', 'ed', 'g_mean_diff', 'h10_mahe', 'h10_mhe', 'h10_rmshe', 'h1_mahe', 'h1_mhe', 'h1_rmshe', 'h2_mahe', 'h2_mhe', 'h2_rmshe', 'h3_mahe', 'h3_mhe', 'h3_rmshe', 'h4_mahe', 'h4_mhe', 'h4_rmshe', 'h5_mahe', 'h5_mhe', 'h5_rmshe', 'h6_mahe', 'h6_mhe', 'h6_rmshe', 'h7_mahe', 'h7_mhe', 'h7_rmshe', 'h8_mahe', 'h8_mhe', 'h8_rmshe', 'irmse', 'lm_index', 'maape', 'male', 'mapd', 'mape', 'mase', 'mb_r', 'mdae', 'mde', 'mdse', 'mean_var', 'me', 'mle', 'mse', 'msle', 'ned', 'nrmse_iqr', 'nrmse_mean', 'nrmse_range', 'nse_mod', 'nse_rel', 'rmse', 'rmsle', 'sa', 'sc', 'sga', 'sid', 'smape1', 'smape2', 'spearman_r', 've', 'watt_m']]]) –

  [optional, Python: list, Command line: comma separated string, default is ‘default’]

  Comma separated list of statistical measures.

  You can select two groups of statistical measures.

  stats	Description
  default	A subset of common statistic measures
  all	All available statistic measures

  The ‘default’ set of statistics are:

  stats	Description
  me	Mean error or bias -inf < ME < inf, close to 0 is better
  pc_bias	Percent Bias -inf < PC_BIAS < inf, close to 0 is better
  apc_bias	Absolute Percent Bias 0 <= APC_BIAS < inf, close to 0 is better
  rmsd	Root Mean Square Deviation/Error 0 <= RMSD < inf, smaller is better
  crmsd	Centered Root Mean Square Deviation/Error
  corrcoef	Pearson Correlation coefficient (r) -1 <= r <= 1 1 perfect positive correlation 0 complete randomness -1 perfect negative correlation
  coefdet	Coefficient of determination (r^2) 0 <= r^2 <= 1 1 perfect correlation 0 complete randomness
  murphyss	Murphy Skill Score
  nse	Nash-Sutcliffe Efficiency -inf < NSE < 1, larger is better
  kge09	Kling-Gupta Efficiency, 2009 -inf < KGE09 < 1, larger is better
  kge12	Kling-Gupta Efficiency, 2012 -inf < KGE12 < 1, larger is better
  index_agreement	Index of agreement (d) 0 <= d < 1, larger is better
  brierss	Brier Skill Score
  mae	Mean Absolute Error 0 <= MAE < 1, larger is better
  mean	observed mean, simulated mean
  stdev	observed stdev, simulated stdev

  Additional statistics:

  stats	Description
  acc	Anomaly correlation coefficient (ACC) -1 <= r <= 1 1 positive correlation of variation in anomalies 0 complete randomness of variation in anomalies -1 negative correlation of variation in anomalies
  d1	Index of agreement (d1) 0 <= d1 < 1, larger is better
  d1_p	Legate-McCabe Index of Agreement 0 <= d1_p < 1, larger is better
  d	Index of agreement (d) 0 <= d < 1, larger is better
  dmod	Modified index of agreement (dmod) 0 <= dmod < 1, larger is better
  drel	Relative index of agreement (drel) 0 <= drel < 1, larger is better
  dr	Refined index of agreement (dr) -1 <= dr < 1, larger is better
  ed	Euclidean distance in vector space 0 <= ed < inf, smaller is better
  g_mean_diff	Geometric mean difference
  h1_mahe	H1 absolute error
  h1_mhe	H1 mean error
  h1_rmshe	H1 root mean square error
  h2_mahe	H2 mean absolute error
  h2_mhe	H2 mean error
  h2_rmshe	H2 root mean square error
  h3_mahe	H3 mean absolute error
  h3_mhe	H3 mean error
  h3_rmshe	H3 root mean square error
  h4_mahe	H4 mean absolute error
  h4_mhe	H4 mean error
  h4_rmshe	H4 root mean square error
  h5_mahe	H5 mean absolute error
  h5_mhe	H5 mean error
  h5_rmshe	H5 root mean square error
  h6_mahe	H6 mean absolute error
  h6_mhe	H6 mean error
  h6_rmshe	H6 root mean square error
  h7_mahe	H7 mean absolute error
  h7_mhe	H7 mean error
  h7_rmshe	H7 root mean square error
  h8_mahe	H8 mean absolute error
  h8_mhe	H8 mean error
  h8_rmshe	H8 root mean square error
  h10_mahe	H10 mean absolute error
  h10_mhe	H10 mean error
  h10_rmshe	H10 root mean square error
  irmse	Inertial root mean square error (IRMSE) 0 <= irmse < inf, smaller is better
  lm_index	Legate-McCabe Efficiency Index 0 <= lm_index < 1, larger is better
  maape	Mean Arctangent Absolute Percentage Error (MAAPE) 0 <= maape < pi/2, smaller is better
  male	Mean absolute log error 0 <= male < inf, smaller is better
  mapd	Mean absolute percentage deviation (MAPD)
  mape	Mean absolute percentage error (MAPE) 0 <= mape < inf, 0 indicates perfect correlation
  mase	Mean absolute scaled error
  mb_r	Mielke-Berry R value (MB R) 0 <= mb_r < 1, larger is better
  mdae	Median absolute error (MdAE) 0 <= mdae < inf, smaller is better
  mde	Median error (MdE) -inf < mde < inf, closer to zero is better
  mdse	Median squared error (MdSE) 0 < mde < inf, closer to zero is better
  mean_var	Mean variance
  me	Mean error -inf < me < inf, closer to zero is better
  mle	Mean log error -inf < mle < inf, closer to zero is better
  mse	Mean squared error 0 <= mse < inf, smaller is better
  msle	Mean squared log error 0 <= msle < inf, smaller is better
  ned	Normalized Euclidian distance in vector space 0 <= ned < inf, smaller is better
  nrmse_iqr	IQR normalized root mean square error 0 <= nrmse_iqr < inf, smaller is better
  nrmse_mean	Mean normalized root mean square error 0 <= nrmse_mean < inf, smaller is better
  nrmse_range	Range normalized root mean square error 0 <= nrmse_range < inf, smaller is better
  nse_mod	Modified Nash-Sutcliffe efficiency (NSE mod) -inf < nse_mod < 1, larger is better
  nse_rel	Relative Nash-Sutcliffe efficiency (NSE rel) -inf < nse_mod < 1, larger is better
  rmse	Root mean square error 0 <= rmse < inf, smaller is better
  rmsle	Root mean square log error 0 <= rmsle < inf, smaller is better
  sa	Spectral Angle (SA) -pi/2 <= sa < pi/2, closer to 0 is better
  sc	Spectral Correlation (SC) -pi/2 <= sc < pi/2, closer to 0 is better
  sga	Spectral Gradient Angle (SGA) -pi/2 <= sga < pi/2, closer to 0 is better
  sid	Spectral Information Divergence (SID) -pi/2 <= sid < pi/2, closer to 0 is better
  smape1	Symmetric Mean Absolute Percentage Error (1) (SMAPE1) 0 <= smape1 < 100, smaller is better
  smape2	Symmetric Mean Absolute Percentage Error (2) (SMAPE2) 0 <= smape2 < 100, smaller is better
  spearman_r	Spearman rank correlation coefficient -1 <= spearman_r <= 1 1 perfect positive correlation 0 complete randomness -1 perfect negative correlation
  ve	Volumetric Efficiency (VE) 0 <= ve < 1, smaller is better
  watt_m	Watterson’s M (M) -1 <= watt_m < 1, larger is better

• replace_nan (float) – If given, indicates which value to replace NaN values with in the two arrays. If None, when a NaN value is found at the i-th position in the observed OR simulated array, the i-th value of the observed and simulated array are removed before the computation.

• replace_inf (float) – If given, indicates which value to replace Inf values with in the two arrays. If None, when an inf value is found at the i-th position in the observed OR simulated array, the i-th value of the observed and simulated array are removed before the computation.

• remove_neg (boolean) – If True, when a negative value is found at the i-th position in the observed OR simulated array, the i-th value of the observed AND simulated array are removed before the computation.

• remove_zero (boolean) – If true, when a zero value is found at the i-th position in the observed OR simulated array, the i-th value of the observed AND simulated array are removed before the computation.

• start_date (str) –

  [optional, defaults to first date in time-series, input filter]

  The start_date of the series in ISOdatetime format, or ‘None’ for beginning.

• end_date (str) –

  [optional, defaults to last date in time-series, input filter]

  The end_date of the series in ISOdatetime format, or ‘None’ for end.

• round_index –

  [optional, default is None which will do nothing to the index, output format]

  Round the index to the nearest time point. Can significantly improve the performance since can cut down on memory and processing requirements, however be cautious about rounding to a very course interval from a small one. This could lead to duplicate values in the index.

• clean –

  [optional, default is False, input filter]

  The ‘clean’ command will repair a input index, removing duplicate index values and sorting.

• index_type (str) –

  [optional, default is ‘datetime’, output format]

  Can be either ‘number’ or ‘datetime’. Use ‘number’ with index values that are Julian dates, or other epoch reference.

• source_units (str) –

  [optional, default is None, transformation]

  If unit is specified for the column as the second field of a ‘:’ delimited column name, then the specified units and the ‘source_units’ must match exactly.

  Any unit string compatible with the ‘pint’ library can be used.

• target_units (str) –

  [optional, default is None, transformation]

  The purpose of this option is to specify target units for unit conversion. The source units are specified in the header line of the input or using the ‘source_units’ keyword.

  The units of the input time-series or values are specified as the second field of a ‘:’ delimited name in the header line of the input or in the ‘source_units’ keyword.

  Any unit string compatible with the ‘pint’ library can be used.

  This option will also add the ‘target_units’ string to the column names.

• tablefmt (str) –

  [optional, default is ‘csv’, output format]

  The table format. Can be one of ‘csv’, ‘tsv’, ‘plain’, ‘simple’, ‘grid’, ‘pipe’, ‘orgtbl’, ‘rst’, ‘mediawiki’, ‘latex’, ‘latex_raw’ and ‘latex_booktabs’.

• float_format –

  [optional, output format]

  Format for float numbers.

• kge_sr (float) –

  [optional, defaults to 1.0]

  Scaling factor for kge09 and kge12 correlation.

• kge09_salpha (float) –

  [optional, defaults to 1.0]

  Scaling factor for kge09 alpha.

• kge12_sgamma (float) –

  [optional, defaults to 1.0]

  Scaling factor for kge12 beta.

• kge_sbeta (float) –

  [optional, defaults to 1.0]

  Scaling factor for kge09 and kge12 beta.

• input_ts (str) –

  [DEPRECATED] [optional, defaults to None]

  The older approach was implicit by using the input_ts and columns to create a two-column dataset where the first column was the observed values and the second column was the simulated values.

  The new approach is to specify the input time series file using obs_col and sim_col parameters.

  OLD WAY:

  ... --input_ts=blah.csv ...
  # Where blah.csv contains two columns, the first
  # is the observed values and the second
  # is the simulated values.
  

  NEW WAY:

  ... --obs_col=blah.csv,2 --sim_col=blah.csv,5 ...
  # Where blah.csv contains at least 5 columns,
  # the second column is the observed values
  # and the fifth column is the simulated values.