- pysteps.blending.skill_scores.clim_regr_values(n_cascade_levels, outdir_path, n_model=0, skill_kwargs=None)#
Obtains the climatological correlation values and regression parameters from a file called NWP_weights_window.bin in the outdir_path. If this file is not present yet, the values from [BPS04] are used.
n_cascade_levels (int) – The number of cascade levels to use.
outdir_path (string) – Path to folder where the historical skill are stored. Defaults to path_workdir from rcparams.
n_model (int, optional) – The index number of the (NWP) model when the climatological skill of multiple (NWP) models is stored. For calculations with one model, or when n_model is not provided, n_model = 0.
skill_kwargs (dict, optional) – Dictionary containing e.g. the outdir_path, nmodels and window_length parameters.
clim_cor_values (array-like) – Array of shape [n_cascade_levels] containing the climatological values of the lag 1 and lag 2 auto-correlation coefficients, obtained by calling a method implemented in pysteps.blending.skill_scores.get_clim_skill_scores.
regr_pars (array-like) – Array of shape [2, n_cascade_levels] containing the regression parameters. These are fixed values that should be hard-coded in this function. Unless changed by the user, the standard values from BPS2004 are used.
The literature climatological values assume 8 cascade levels. In case less or more cascade levels are used, the script will handle this by taking the first n values or extending the array with a small value. This is not ideal, but will be fixed once the clim_regr_file is made. Hence, this requires a warm-up period of the model. In addition, the regression parameter values (eq. 24 in BPS2004) are hard- coded and can only be optimized by the user after (re)fitting of the equation.