hydromt.stats.extremes.eva#

hydromt.stats.extremes.eva(da: DataArray, ev_type: str = 'BM', min_dist: int = 0, qthresh: float = 0.9, period: str = '365.25D', min_sample_size: int = 0, distribution: str | None = None, rps: ndarray[Any, dtype[int64]] = array([2, 5, 10, 25, 50, 100, 250, 500]), criterium: str = 'AIC') → Dataset[source]#

Return Extreme Value Analysis.

Extreme value analysis based on block maxima (BM) or Peaks Over Threshold (POT). The method selects the peaks, fits a distribution and calculates return values for provided return periods.

Parameters:

  • da (xr.DataArray) – Timeseries data, must have a regular spaced ‘time’ dimension.

  • ev_type ({"POT", "BM"}) – Peaks over threshold (POT) or block maxima (BM) peaks, by default “BM”

  • period (str, optional) – Period string, by default “365.25D”. See pandas.Timedelta for options.

  • min_dist (int, optional) – Minimum distance between peaks measured in time steps, by default 0

  • qthresh (float, optional) – Quantile threshold used with peaks over threshold method, by default 0.9

  • min_sample_size (int, optional) – Minimum number of finite values in a valid block, by default 0. Peaks of invalid blocks are set to NaN.

  • distribution (str, optional) – Short name of distribution. If None (default) the optimal block maxima distribution (“gumb” or “gev” for BM and “exp” or “gpd” for POT) is selected based on criterium.

  • rps (np.ndarray, optional) – Array of return periods, by default [2, 5, 10, 25, 50, 100, 250, 500]

  • criterium ({'AIC', 'AICc', 'BIC'}) – Selection criterium, by default “AIC”

Returns:

Dataset with peaks timeseries, distribution name and parameters and return values.

Return type:

xr.Dataset