Skip to content

BUG: on transformation for climate and fir clim mix

Bug

Error description

When using the data handler DataHandlerMixedSamplingWithClimateAndFirFilter with no extend length (which corresponds to usage of data handler only using climFIR, but with separated chem/meteo vars), there are the expected plots for the climate FIR filter (chem and meteo) but also plots for the FIR filter for all meteo vars in addition. See paramterers:

data_handler=DataHandlerMixedSamplingWithClimateAndFirFilter,
window_history_size=3 * 24 - 1,
window_history_offset={"chem": 24, "meteo": 24},
window_history_end={"chem": 0, "meteo": 0},
extend_length_opts={"chem": 0, "meteo": 0},

Error message

First guess on error origin

This only effects the meteo variables. Unexpected plots are created much later than the expected ones. Maybe this is related to transformation?

Error origin

Reason: During transformation the following is called

# meteo transformation
if len(meteo_vars) > 0:
    kwargs_meteo = copy.deepcopy(kwargs)
    cls.prepare_build(kwargs_meteo, meteo_vars, cls.meteo_indicator)
    dh_transformation = (cls.data_handler_fir[cls.data_handler_fir_pos or 0], cls.data_handler_unfiltered)
    transformation_meteo = super().transformation(set_stations, tmp_path=tmp_path,
                                                  dh_transformation=dh_transformation, **kwargs_meteo)

But as cls.data_handler_fir_pos is initialized as None, this code will always use the index 0 for transformation. Make sure to check beforehand, which extend length is used and set cls.data_handler_fir_pos. Need to implement a method that sets the position before applying transformation. This should be similar to the following code block in cls.build:

if len(meteo_vars) > 0:
    if cls.data_handler_fir_pos is None:
        if "extend_length_opts" in kwargs:
            if isinstance(kwargs["extend_length_opts"], dict) and cls.meteo_indicator not in kwargs["extend_length_opts"].keys():
                cls.data_handler_fir_pos = 0  # use faster fir version without climate estimate
            else:
                cls.data_handler_fir_pos = 1  # use slower fir version with climate estimate
        else:
            cls.data_handler_fir_pos = 0  # use faster fir version without climate estimate

Solution

  • extract relevant code from the cls.build command to set cls.data_handler_fir_pos in new classmethod
  • replace extracted code parts in cls.build
  • add pos method call in cls.transformation as used in cls.build