environmental_insights.models

environmental_insights.models#

ensure_download#

def ensure_download(dataset: str,
                    data_type: str,
                    *,
                    timestamp: Optional[str] = None,
                    month: Optional[Union[int, str]] = None,
                    day: Optional[str] = None,
                    hour: Optional[Union[int, str]] = None,
                    model_level: Optional[str] = None,
                    pollutant: Optional[str] = None,
                    station: Optional[str] = None,
                    token: Optional[str] = None,
                    output_dir: Optional[Union[str, Path]] = None) -> None

Ensure the requested file(s) exist locally by invoking the unified downloader.

load_model_united_kingdom#

def load_model_united_kingdom(
        model_level: str,
        pollutant: str,
        model_category: str,
        token: Optional[str] = None) -> lgb.LGBMRegressor

Load a pretrained UK air pollution LGBM model for a specific category, downloading only the required booster + params.

Model directory structure (identical for both namespaces): {namespace}/Models/{level}/{pollutant}/All_Stations/{pollutant}_{category} but when model_category == “All”, use just {pollutant} with no suffix.

Namespaces:

  • ML-HAPPE (for existing categories)

  • SynthHAPPE_v2 (for Climate_Projections_Models and Transport_Infrastructure_Policy_Models)

load_model_global#

def load_model_global(model_level: str,
                      pollutant: str,
                      model_category: str,
                      token: Optional[str] = None) -> lgb.LGBMRegressor

Load a pretrained GLOBAL ML-HAPPG air pollution LGBM model for a specific category, downloading only the required booster + params.

For ‘temporal’ category: Models/{model_level}/temporal/{pollutant} For all other categories: Models/{model_level}/{pollutant}/All_Stations/{suffix} where suffix = pollutant if model_category == ‘All’ else f”{pollutant}_{model_category}”

make_concentration_predictions_united_kingdom#

def make_concentration_predictions_united_kingdom(
        estimating_model: lgb.LGBMRegressor, observation_data: pd.DataFrame,
        feature_names: List[str]) -> pd.DataFrame

Predict concentrations using a trained LGBM model.

make_concentration_predictions_global#

def make_concentration_predictions_global(
        estimating_model: lgb.LGBMRegressor, observation_data: pd.DataFrame,
        feature_names: List[str]) -> pd.DataFrame

Predict concentrations using a trained GLOBAL LGBM model.

rename_global_input_columns#

def rename_global_input_columns(df: pd.DataFrame) -> pd.DataFrame

Rename columns in a DataFrame to match the GLOBAL model’s expected names.

Contents