environmentaltools.temporal.fit_gpd_bootstrap

environmentaltools.temporal.fit_gpd_bootstrap(df_eventos, alpha, umb, param, bca, nyears)[source]

Compute GPD parameters with bootstrap resampling and confidence intervals.

Automatically fits Generalized Pareto Distribution using L-moments with bootstrap resampling for specified thresholds. Also tests if Exponential distribution is appropriate when shape parameter CI includes zero.

Parameters:

  • df_eventos (pd.DataFrame or np.ndarray) – Event time series data

  • alpha (float) – Significance level for confidence intervals

  • umb (array-like) – Threshold values to test

  • param ({'parametric', 'non-parametric'}) – Bootstrap resampling method

  • bca ({'standard', 'bca'}) – Confidence interval computation method

  • nyears (int) – Number of years in the dataset

Returns:

  • boot (list of lists) – Bootstrap results [GPD_results, Exponential_results (if applicable)] Each contains bootstrap parameter matrices for each threshold

  • orig (list of lists) – Original parameter estimates [GPD_estimates, Exponential_estimates]

  • ci (list of lists) – Confidence intervals [GPD_CI, Exponential_CI]

  • tr (np.ndarray) – Return periods evaluated

  • peaks (np.ndarray) – Peak values used (for last threshold processed)

  • npeaks (int) – Number of peaks (for last threshold)

  • eventanu (float) – Average events per year (for last threshold)

Notes

For each threshold:

  1. Extracts peaks exceeding threshold

  2. Fits GPD using L-moments with bootstrap

  3. Computes confidence intervals

  4. If CI for shape includes 0, also fits Exponential distribution

Exponential is a special case of GPD with shape=0.

Examples

>>> umb = np.array([2.0, 2.5])
>>> boot, orig, ci, tr, peaks, npeaks, eventanu = fit_gpd_bootstrap(
...     df['Hs'], alpha=0.05, umb=umb, param='parametric',
...     bca='standard', nyears=10
... )