def load_observations_for_station(station_name, element_name):
    """Load observations for a specific stations."""
    element_id = data_io.get_element_id(element_name, "fc")

    # TODO Add support
    if element_id != 999:
        raise NotImplementedError()

    observations = data_io.read_observations(element_id, station_name)
    return observations
def load_and_interpolate_forecast(model, element_name,
                                  station_name, issue):
    """Load a given dataset of forecast data with interpolated values.

    parameters
    ----------
    model: str, name of model
    element_name: str, string representation of the provided parameter
    station_name: str, canonical name of the station
    issue: str, string representation of the model issue hour
    """
    # Forecast columns: 1 to 4
    forecast_cols = ['value' + str(x) for x in range(1, 5)]

    # Station location
    station_lat, station_lon = data_io.get_station_location(station_name)

    # Read in forecast
    element_id = data_io.get_element_id(element_name, model)
    forecast_data = data_io.read_forecast_data(
        model, element_id, station_name, issue)

    # Check whether to do any interpolation
    empty_columns = \
        forecast_data.ix[:, forecast_cols].isnull().values.any(axis=0)
    if empty_columns.sum() == 3:
        # Just a single column provided. Don't do interpolation.
        logging.debug(
              "Not interpolating for model '%s', element '%s', station '%s'" %
              (model, str(element_id), station_name))
        # Select non-empty column
        non_empty_col = forecast_cols[(~empty_columns).nonzero()[0][0]]
        interpolated_forecast = forecast_data[non_empty_col]
    else:
        # Do interpolation using meta-data
        meta_data = \
            data_io.read_meta_data(model, element_id, station_name, issue)
        interpolated_forecast = interpolate(
            station_lat, station_lon,
            forecast_data.ix[:, forecast_cols],
            meta_data['latitude'], meta_data['longitude'],
            meta_data['distance']
        )
    forecast_data['interpolated_forecast'] = interpolated_forecast

    # Drop grid point data.
    forecast_data.drop(forecast_cols, axis=1, inplace=True)

    # Transform EPS data from row format to column format
    forecast_data = \
        transform_forecast_group_data(forecast_data, model, element_name)
    return forecast_data