| """Generate synthetic indoor air pollution & clean cooking dataset.""" |
|
|
| from __future__ import annotations |
|
|
| from pathlib import Path |
|
|
| import numpy as np |
| import pandas as pd |
|
|
| SEED = 42 |
| N_PER_SCENARIO = 10_000 |
|
|
| YEAR_RANGE = np.arange(2010, 2025) |
| YEAR_WEIGHTS = np.linspace(0.85, 1.3, len(YEAR_RANGE)) |
| YEAR_WEIGHTS = YEAR_WEIGHTS / YEAR_WEIGHTS.sum() |
|
|
| CLEAN_FUELS = {"lpg", "electricity", "biogas", "ethanol"} |
| BIOMASS_FUELS = {"wood", "charcoal", "dung", "crop_residue"} |
|
|
| FUEL_PM25_FACTOR = { |
| "wood": 1.0, |
| "charcoal": 0.85, |
| "dung": 1.15, |
| "crop_residue": 1.05, |
| "kerosene": 0.65, |
| "lpg": 0.2, |
| "electricity": 0.05, |
| "biogas": 0.2, |
| "ethanol": 0.2, |
| } |
|
|
| FUEL_CO_FACTOR = { |
| "wood": 1.0, |
| "charcoal": 0.8, |
| "dung": 1.1, |
| "crop_residue": 1.0, |
| "kerosene": 0.6, |
| "lpg": 0.25, |
| "electricity": 0.05, |
| "biogas": 0.25, |
| "ethanol": 0.25, |
| } |
|
|
| STOVE_PM25_FACTOR = {"traditional": 1.0, "improved": 0.75, "clean": 0.3} |
|
|
| FUEL_BC_FACTOR = { |
| "wood": 0.08, |
| "charcoal": 0.05, |
| "dung": 0.1, |
| "crop_residue": 0.07, |
| "kerosene": 0.03, |
| "lpg": 0.01, |
| "electricity": 0.0, |
| "biogas": 0.01, |
| "ethanol": 0.01, |
| } |
|
|
| FUEL_CONSUMPTION_FACTOR = { |
| "wood": 1.2, |
| "charcoal": 0.8, |
| "dung": 1.0, |
| "crop_residue": 0.9, |
| "kerosene": 0.4, |
| "lpg": 0.2, |
| "electricity": 0.0, |
| "biogas": 0.1, |
| "ethanol": 0.1, |
| } |
|
|
| KITCHEN_VENTILATION = {"enclosed": 0.2, "semi_open": 0.4, "open": 0.65} |
|
|
| SCENARIOS = { |
| "traditional_biomass": { |
| "setting_probs": {"rural": 0.7, "peri_urban": 0.2, "urban": 0.1}, |
| "fuel_probs": { |
| "wood": 0.45, |
| "charcoal": 0.2, |
| "dung": 0.2, |
| "crop_residue": 0.1, |
| "kerosene": 0.05, |
| }, |
| "cookstove_probs": {"traditional": 0.85, "improved": 0.1, "clean": 0.05}, |
| "kitchen_probs": {"enclosed": 0.6, "semi_open": 0.3, "open": 0.1}, |
| "pm25_mean": 420, |
| "pm25_sd": 180, |
| "co_mean": 18, |
| "co_sd": 7, |
| "ventilation_mean": 0.25, |
| "cooking_hours_mean": 3.0, |
| "intervention_coverage": 0.1, |
| "adoption_rate": 0.4, |
| "compliance_mean": 0.6, |
| "baseline_pneumonia": 16.5, |
| "household_size_mean": 5.4, |
| "household_size_sd": 1.6, |
| "children_u5_mean": 1.2, |
| }, |
| "improved_stove_rollout": { |
| "setting_probs": {"rural": 0.6, "peri_urban": 0.25, "urban": 0.15}, |
| "fuel_probs": { |
| "wood": 0.35, |
| "charcoal": 0.2, |
| "dung": 0.15, |
| "crop_residue": 0.1, |
| "kerosene": 0.05, |
| "lpg": 0.1, |
| "electricity": 0.05, |
| }, |
| "cookstove_probs": {"traditional": 0.45, "improved": 0.45, "clean": 0.1}, |
| "kitchen_probs": {"enclosed": 0.5, "semi_open": 0.35, "open": 0.15}, |
| "pm25_mean": 260, |
| "pm25_sd": 120, |
| "co_mean": 12, |
| "co_sd": 5, |
| "ventilation_mean": 0.35, |
| "cooking_hours_mean": 2.7, |
| "intervention_coverage": 0.6, |
| "adoption_rate": 0.55, |
| "compliance_mean": 0.65, |
| "baseline_pneumonia": 15.5, |
| "household_size_mean": 5.2, |
| "household_size_sd": 1.5, |
| "children_u5_mean": 1.1, |
| }, |
| "clean_fuel_transition": { |
| "setting_probs": {"urban": 0.45, "peri_urban": 0.35, "rural": 0.2}, |
| "fuel_probs": { |
| "lpg": 0.35, |
| "electricity": 0.25, |
| "biogas": 0.1, |
| "charcoal": 0.15, |
| "wood": 0.1, |
| "ethanol": 0.05, |
| }, |
| "cookstove_probs": {"clean": 0.65, "improved": 0.2, "traditional": 0.15}, |
| "kitchen_probs": {"enclosed": 0.4, "semi_open": 0.35, "open": 0.25}, |
| "pm25_mean": 70, |
| "pm25_sd": 35, |
| "co_mean": 4, |
| "co_sd": 2, |
| "ventilation_mean": 0.5, |
| "cooking_hours_mean": 2.3, |
| "intervention_coverage": 0.8, |
| "adoption_rate": 0.75, |
| "compliance_mean": 0.8, |
| "baseline_pneumonia": 13.0, |
| "household_size_mean": 4.9, |
| "household_size_sd": 1.4, |
| "children_u5_mean": 0.9, |
| }, |
| } |
|
|
| SCENARIO_FILES = { |
| "traditional_biomass": "indoor_air_traditional_biomass.csv", |
| "improved_stove_rollout": "indoor_air_improved_stove_rollout.csv", |
| "clean_fuel_transition": "indoor_air_clean_fuel_transition.csv", |
| } |
|
|
|
|
| def _choice(rng: np.random.Generator, prob_map: dict[str, float]) -> str: |
| keys = list(prob_map.keys()) |
| weights = np.array(list(prob_map.values()), dtype=float) |
| weights = weights / weights.sum() |
| return rng.choice(keys, p=weights) |
|
|
|
|
| def _sample_year(rng: np.random.Generator) -> int: |
| return int(rng.choice(YEAR_RANGE, p=YEAR_WEIGHTS)) |
|
|
|
|
| def _simulate_scenario(name: str, params: dict[str, float], seed: int) -> pd.DataFrame: |
| rng = np.random.default_rng(seed) |
| records = [] |
|
|
| for idx in range(N_PER_SCENARIO): |
| setting = _choice(rng, params["setting_probs"]) |
| kitchen_type = _choice(rng, params["kitchen_probs"]) |
| fuel_type = _choice(rng, params["fuel_probs"]) |
| cookstove_type = _choice(rng, params["cookstove_probs"]) |
| household_size = int( |
| np.clip( |
| rng.normal(params["household_size_mean"], params["household_size_sd"]), 2, 11 |
| ) |
| ) |
| children_u5 = int(np.clip(rng.poisson(params["children_u5_mean"]), 0, 4)) |
| cooking_hours = float(np.clip(rng.normal(params["cooking_hours_mean"], 0.6), 1.2, 5.5)) |
| ventilation_index = float( |
| np.clip(rng.normal(KITCHEN_VENTILATION[kitchen_type], 0.07), 0.1, 0.9) |
| ) |
|
|
| intervention_received = rng.random() < params["intervention_coverage"] |
| adoption = intervention_received and rng.random() < params["adoption_rate"] |
| compliance = float(np.clip(rng.normal(params["compliance_mean"], 0.15), 0.2, 1.0)) |
| if not adoption: |
| compliance = 0.0 |
|
|
| clean_access = fuel_type in CLEAN_FUELS or cookstove_type == "clean" |
|
|
| base_pm25 = max(20.0, rng.normal(params["pm25_mean"], params["pm25_sd"])) |
| vent_factor = 1.1 - ventilation_index * 0.6 |
| pm25 = ( |
| base_pm25 |
| * FUEL_PM25_FACTOR[fuel_type] |
| * STOVE_PM25_FACTOR[cookstove_type] |
| * vent_factor |
| ) |
| pm25 = float(np.clip(pm25, 15, 1200)) |
|
|
| co_ppm = ( |
| rng.normal(params["co_mean"], params["co_sd"]) |
| * FUEL_CO_FACTOR[fuel_type] |
| * STOVE_PM25_FACTOR[cookstove_type] |
| * (1.05 - ventilation_index * 0.5) |
| ) |
| co_ppm = float(np.clip(co_ppm, 0.5, 80)) |
|
|
| exposure_hours = float(cooking_hours * rng.uniform(0.5, 0.9)) |
| exposure_index = float(pm25 * exposure_hours / 24) |
|
|
| lri_rr = 1 + np.clip((pm25 - 50) / 450, 0, 1) * 0.9 |
| stove_health_factor = 1.0 |
| if cookstove_type == "clean": |
| stove_health_factor = 0.85 |
| elif cookstove_type == "improved": |
| stove_health_factor = 0.98 |
|
|
| pneumonia_incidence = params["baseline_pneumonia"] * lri_rr * stove_health_factor |
| if clean_access and compliance > 0: |
| pneumonia_incidence *= 1 - 0.05 * compliance |
| pneumonia_incidence = float(np.clip(pneumonia_incidence, 5, 40)) |
|
|
| copd_risk_index = float(np.clip(0.2 + (pm25 / 600) * 0.8, 0.2, 1.0)) |
| low_birthweight_risk = float(np.clip(0.08 + (pm25 / 600) * 0.12, 0.05, 0.35)) |
|
|
| health_burden_score = float( |
| np.clip( |
| (pneumonia_incidence / 30) * 0.4 |
| + copd_risk_index * 0.35 |
| + low_birthweight_risk * 0.25, |
| 0, |
| 1, |
| ) |
| ) |
|
|
| black_carbon_kg_per_day = float( |
| FUEL_BC_FACTOR[fuel_type] * cooking_hours * (0.3 if cookstove_type == "clean" else 1) |
| ) |
| fuel_kg_per_day = float(FUEL_CONSUMPTION_FACTOR[fuel_type] * cooking_hours) |
|
|
| if fuel_type in BIOMASS_FUELS: |
| base_collection = 1.3 if setting == "rural" else 0.8 |
| fuel_collection_hours = float(np.clip(rng.normal(base_collection, 0.4), 0.1, 3.5)) |
| else: |
| fuel_collection_hours = float(np.clip(rng.normal(0.2, 0.1), 0.0, 0.6)) |
|
|
| if pm25 >= 500: |
| pm25_category = "extreme" |
| elif pm25 >= 250: |
| pm25_category = "high" |
| elif pm25 >= 100: |
| pm25_category = "moderate" |
| else: |
| pm25_category = "low" |
|
|
| record = { |
| "record_id": f"{name[:3].upper()}-{idx:05d}", |
| "scenario": name, |
| "year": _sample_year(rng), |
| "setting": setting, |
| "household_size": household_size, |
| "children_u5": children_u5, |
| "primary_cook_female": int(rng.random() < 0.82), |
| "cooking_hours_per_day": cooking_hours, |
| "kitchen_type": kitchen_type, |
| "ventilation_index": ventilation_index, |
| "fuel_type": fuel_type, |
| "cookstove_type": cookstove_type, |
| "clean_cooking_access": int(clean_access), |
| "intervention_received": int(intervention_received), |
| "adoption_compliance": compliance, |
| "pm25_kitchen_ugm3": pm25, |
| "co_ppm": co_ppm, |
| "exposure_hours": exposure_hours, |
| "exposure_index": exposure_index, |
| "pm25_category": pm25_category, |
| "pneumonia_incidence_per100": pneumonia_incidence, |
| "child_lri_risk_ratio": lri_rr, |
| "copd_risk_index": copd_risk_index, |
| "low_birthweight_risk": low_birthweight_risk, |
| "health_burden_score": health_burden_score, |
| "black_carbon_kg_per_day": black_carbon_kg_per_day, |
| "fuel_kg_per_day": fuel_kg_per_day, |
| "fuel_collection_hours": fuel_collection_hours, |
| } |
| records.append(record) |
|
|
| return pd.DataFrame(records) |
|
|
|
|
| def main() -> None: |
| output_dir = Path("data") |
| output_dir.mkdir(parents=True, exist_ok=True) |
|
|
| for idx, (name, params) in enumerate(SCENARIOS.items()): |
| df = _simulate_scenario(name, params, SEED + idx * 211) |
| df.to_csv(output_dir / SCENARIO_FILES[name], index=False) |
| print(f"Saved {name} -> {SCENARIO_FILES[name]}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
