Upload folder using huggingface_hub

README.md

@@ -0,0 +1,111 @@
+---
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+  - tabular-regression
+language:
+  - en
+tags:
+  - climate
+  - environment
+  - africa
+  - synthetic-data
+  - sub-saharan-africa
+  - drought
+  - flood
+  - disaster-impact
+  - climate-disasters
+size_categories:
+  - 10K<n<100K
+---
+
+# Drought and Flood Impact - Africa
+
+A comprehensive synthetic dataset documenting drought and flood events, their severity, impacts, and recovery metrics across Sub-Saharan African countries.
+
+## Dataset Description
+
+This dataset provides detailed records of drought and flood events across 15 Sub-Saharan African countries. It includes event characteristics, population impacts, economic damages, early warning indicators, and recovery progress. Data is stratified by malnutrition burden levels to analyze disaster-nutrition linkages.
+
+### Key Statistics
+
+| Metric | Value |
+|--------|-------|
+| Total Records | 14,998 |
+| Countries Covered | 15 |
+| Time Period | 2018-2025 |
+| Files | 3 (high_burden, moderate_burden, low_burden) |
+| Features | 31 |
+
+### Countries Included
+DRC, Ethiopia, Ghana, Kenya, Malawi, Mali, Mozambique, Niger, Nigeria, Rwanda, Senegal, South Africa, Tanzania, Uganda, Zambia
+
+## Column Descriptions
+
+| Column | Type | Description |
+|--------|------|-------------|
+| record_id | string | Unique identifier (DF_XXXXXX) |
+| country | string | African country name |
+| year | int | Year of event (2018-2025) |
+| event_type | string | Type of event (drought/flood) |
+| event_subtype | string | Specific event classification |
+| event_duration_days | int | Duration in days |
+| rainfall_deficit_pct | float | Rainfall deviation from normal (%) |
+| severity_index | float | Composite severity score (0-10) |
+| severity_category | string | Severity classification |
+| affected_area_km2 | float | Geographic extent (km²) |
+| population_affected | int | Number of people affected |
+| population_displaced | int | Number of people displaced |
+| deaths | int | Number of deaths |
+| injuries | int | Number of injuries |
+| housing_destroyed | int | Houses destroyed |
+| housing_damaged | int | Houses damaged |
+| agricultural_land_affected_ha | float | Agricultural land impacted (hectares) |
+| crop_loss_value_usd | float | Value of crop losses (USD) |
+| livestock_loss_heads | int | Livestock lost (heads) |
+| infrastructure_damage_usd | float | Infrastructure damage (USD) |
+| total_economic_damage_usd | float | Total economic damage (USD) |
+| damage_gdp_pct | float | Damage as % of GDP |
+| early_warning_issued | int | Whether early warning was issued (0/1) |
+| warning_lead_time_hours | int | Warning lead time (hours) |
+| emergency_response_activated | int | Emergency response status (0/1) |
+| response_time_hours | int | Response time (hours) |
+| international_aid_requested | int | Whether aid was requested (0/1) |
+| international_aid_received_usd | float | International aid received (USD) |
+| recovery_duration_months | int | Recovery duration (months) |
+| recovery_progress_pct | float | Recovery progress (%) |
+| long_term_displacement_risk | float | Risk of permanent displacement (0-1) |
+
+### Event Types
+- drought (meteorological, agricultural, hydrological)
+- flood (flash, riverine, coastal)
+
+### Severity Categories
+- Minor, Moderate, Severe, Extreme, Catastrophic
+
+## Usage Example
+
+```python
+import pandas as pd
+
+# Load high burden dataset
+df = pd.read_csv('drought_flood_impact_africa_high_burden.csv')
+
+# Analyze by event type
+event_impacts = df.groupby('event_type')['total_economic_damage_usd'].sum()
+print(event_impacts)
+
+# Filter catastrophic events
+catastrophic = df[df['severity_category'] == 'Catastrophic']
+
+# Calculate average recovery by severity
+recovery_analysis = df.groupby('severity_category')['recovery_duration_months'].mean()
+```
+
+## Research Sources
+
+This synthetic dataset is inspired by and aligned with data from:
+- [EM-DAT International Disaster Database](https://www.emdat.be/)
+- [FEWS NET](https://fews.net/)
+- [Global Flood Awareness System](https://www.globalfloods.eu/)
+- [UNDRR Disaster Data](https://www.undrr.org/)

generate_dataset.py

@@ -0,0 +1,222 @@
+"""
+Drought and Flood Impact Dataset Generator for Africa
+
+================================================================================
+PARAMETER EVIDENCE TABLE
+================================================================================
+| Parameter                    | Value/Range           | Source                   |
+|------------------------------|-----------------------|--------------------------|
+| Drought events (2015-2023)   | 171 recorded          | EM-DAT 2024              |
+| Flood events (2015-2023)     | 456 recorded          | EM-DAT 2024              |
+| Horn of Africa drought       | 5 consecutive seasons | WMO 2024                 |
+| Economic loss (drought)      | $30B (2018-2023)      | AUC 2024                 |
+| Population affected drought  | 150M (2020-2023)      | OCHA 2024                |
+| Flood displacement           | 6.3M (2022 alone)     | IDMC 2023                |
+| Cyclone Idai damage          | $2.2B (Mozambique)    | World Bank 2019          |
+| Recovery time average        | 3-7 years             | UNDRR 2024               |
+================================================================================
+
+DAG Structure:
+climate_conditions -> event_type -> severity -> damage_assessment
+vulnerability_factors -> impact_magnitude
+infrastructure_resilience -> recovery_trajectory
+response_capacity -> recovery_duration
+"""
+
+import numpy as np
+import pandas as pd
+from pathlib import Path
+
+COUNTRIES = [
+    "Kenya", "Uganda", "Nigeria", "Ghana", "Tanzania", "Ethiopia",
+    "Malawi", "Zambia", "Senegal", "Rwanda", "Niger", "Mali",
+    "DRC", "Mozambique", "South Africa"
+]
+
+YEARS = list(range(2018, 2026))
+
+EVENT_TYPES = ["drought", "flood", "flash_flood", "riverine_flood", "cyclone"]
+
+COUNTRY_EVENT_PROBABILITY = {
+    "Niger": {"drought": 0.625, "flood": 0.375},
+    "Mali": {"drought": 0.64, "flood": 0.36},
+    "Ethiopia": {"drought": 0.58, "flood": 0.42},
+    "Kenya": {"drought": 0.54, "flood": 0.46},
+    "Mozambique": {"drought": 0.22, "flood": 0.50, "cyclone": 0.28},
+    "Malawi": {"drought": 0.38, "flood": 0.62},
+    "DRC": {"drought": 0.25, "flood": 0.75},
+    "Tanzania": {"drought": 0.42, "flood": 0.58},
+    "Uganda": {"drought": 0.33, "flood": 0.67},
+    "Zambia": {"drought": 0.58, "flood": 0.42},
+    "Ghana": {"drought": 0.30, "flood": 0.70},
+    "Nigeria": {"drought": 0.33, "flood": 0.67},
+    "Senegal": {"drought": 0.58, "flood": 0.42},
+    "Rwanda": {"drought": 0.30, "flood": 0.70},
+    "South Africa": {"drought": 0.55, "flood": 0.45},
+}
+
+def generate_dataset(scenario: str, n_samples: int, seed: int) -> pd.DataFrame:
+    rng = np.random.default_rng(seed)
+    
+    burden_mod = {"low_burden": 0.7, "moderate_burden": 1.0, "high_burden": 1.5}[scenario]
+    
+    countries = rng.choice(COUNTRIES, size=n_samples)
+    years = rng.choice(YEARS, size=n_samples)
+    
+    data = {
+        "record_id": [f"DF_{i:06d}" for i in range(n_samples)],
+        "country": countries,
+        "year": years,
+    }
+    
+    event_types = []
+    for c in countries:
+        probs = COUNTRY_EVENT_PROBABILITY.get(c, {"drought": 0.40, "flood": 0.60})
+        event_types.append(rng.choice(list(probs.keys()), p=list(probs.values())))
+    data["event_type"] = event_types
+    data["event_subtype"] = np.select(
+        [np.array(event_types) == "flood", np.array(event_types) == "drought"],
+        [rng.choice(["riverine", "flash", "coastal"], n_samples, p=[0.5, 0.35, 0.15]),
+         rng.choice(["meteorological", "agricultural", "hydrological"], n_samples, p=[0.4, 0.35, 0.25])],
+        rng.choice(["tropical_storm", "cyclone", "severe_storm"], n_samples)
+    )
+    
+    data["event_duration_days"] = np.select(
+        [np.array(event_types) == "drought", 
+         np.array(event_types) == "flood", 
+         np.array(event_types) == "cyclone"],
+        [rng.integers(60, 365, n_samples),
+         rng.integers(3, 30, n_samples),
+         rng.integers(1, 7, n_samples)]
+    )
+    
+    data["rainfall_deficit_pct" if "drought" else "rainfall_excess_pct"] = np.select(
+        [np.array(event_types) == "drought", np.array(event_types) != "drought"],
+        [rng.integers(30, 80, n_samples) * burden_mod,
+         rng.integers(100, 400, n_samples) * burden_mod]
+    )
+    
+    data["severity_index"] = np.clip(
+        rng.beta(2, 5, n_samples) * 10 * burden_mod, 1, 10
+    ).round(2)
+    
+    severity = data["severity_index"]
+    data["severity_category"] = np.select(
+        [severity < 3, severity < 5, severity < 7, severity < 9, severity >= 9],
+        ["Minor", "Moderate", "Severe", "Extreme", "Catastrophic"]
+    )
+    
+    data["affected_area_km2"] = np.clip(
+        rng.exponential(5000, n_samples) * (severity / 5) * burden_mod, 100, 500000
+    ).astype(int)
+    
+    data["population_affected"] = np.clip(
+        data["affected_area_km2"] * rng.exponential(50, n_samples) * 
+        (severity / 5) * burden_mod, 1000, 25000000
+    ).astype(int)
+    
+    data["population_displaced"] = np.clip(
+        data["population_affected"] * rng.uniform(0.05, 0.35, n_samples) * 
+        (severity / 5), 100, 5000000
+    ).astype(int)
+    
+    data["deaths"] = np.clip(
+        rng.poisson(50, n_samples) * (severity / 5) * burden_mod, 0, 5000
+    ).astype(int)
+    
+    data["injuries"] = np.clip(
+        data["deaths"] * rng.uniform(2, 8, n_samples), 0, 30000
+    ).astype(int)
+    
+    data["housing_destroyed"] = np.clip(
+        data["population_displaced"] * rng.uniform(0.2, 0.6, n_samples), 0, 500000
+    ).astype(int)
+    
+    data["housing_damaged"] = np.clip(
+        data["housing_destroyed"] * rng.uniform(1.5, 4, n_samples), 0, 1500000
+    ).astype(int)
+    
+    data["agricultural_land_affected_ha"] = np.clip(
+        data["affected_area_km2"] * 100 * rng.uniform(0.1, 0.5, n_samples), 100, 5000000
+    ).astype(int)
+    
+    data["crop_loss_value_usd"] = np.clip(
+        data["agricultural_land_affected_ha"] * rng.uniform(200, 800, n_samples) * 
+        (severity / 5), 10000, 5000000000
+    ).astype(int)
+    
+    data["livestock_loss_heads"] = np.clip(
+        rng.poisson(1000, n_samples) * (severity / 3) * burden_mod, 0, 500000
+    ).astype(int)
+    
+    data["infrastructure_damage_usd"] = np.clip(
+        data["affected_area_km2"] * rng.uniform(5000, 50000, n_samples) * 
+        (severity / 5), 100000, 10000000000
+    ).astype(int)
+    
+    data["total_economic_damage_usd"] = (
+        data["crop_loss_value_usd"] + 
+        data["infrastructure_damage_usd"] + 
+        data["housing_destroyed"] * rng.uniform(5000, 20000, n_samples)
+    ).astype(int)
+    
+    data["damage_gdp_pct"] = np.clip(
+        data["total_economic_damage_usd"] / 1e9 * rng.uniform(0.5, 2, n_samples), 0.1, 25
+    ).round(2)
+    
+    data["early_warning_issued"] = rng.choice([0, 1], size=n_samples, p=[0.3, 0.7])
+    data["warning_lead_time_hours"] = np.where(
+        data["early_warning_issued"] == 1,
+        np.clip(rng.exponential(48, n_samples), 1, 168).astype(int),
+        0
+    )
+    
+    data["emergency_response_activated"] = rng.choice([0, 1], size=n_samples, p=[0.25, 0.75])
+    data["response_time_hours"] = np.where(
+        data["emergency_response_activated"] == 1,
+        np.clip(rng.exponential(36, n_samples), 2, 120).astype(int),
+        0
+    )
+    
+    data["international_aid_requested"] = rng.choice([0, 1], size=n_samples, p=[0.6, 0.4])
+    data["international_aid_received_usd"] = np.where(
+        data["international_aid_requested"] == 1,
+        np.clip(data["total_economic_damage_usd"] * rng.uniform(0.02, 0.15, n_samples), 
+                100000, 500000000).astype(int),
+        0
+    )
+    
+    data["recovery_duration_months"] = np.clip(
+        (severity * 3 + rng.integers(3, 12, n_samples)) * 
+        (1 - data["international_aid_received_usd"] / (data["total_economic_damage_usd"] + 1) * 0.3), 
+        1, 84
+    ).astype(int)
+    
+    data["recovery_progress_pct"] = np.clip(
+        100 - (severity * 5) + rng.integers(-10, 20, n_samples), 10, 100
+    ).astype(int)
+    
+    data["long_term_displacement_risk"] = np.clip(
+        data["population_displaced"] / (data["population_affected"] + 1) * 
+        (1 - data["recovery_progress_pct"] / 150), 0, 1
+    ).round(3)
+    
+    return pd.DataFrame(data)
+
+def main():
+    output_dir = Path(__file__).parent
+    
+    scenarios = [
+        ("low_burden", 4000, 42),
+        ("moderate_burden", 5000, 43),
+        ("high_burden", 6000, 44),
+    ]
+    
+    for scenario, n, seed in scenarios:
+        df = generate_dataset(scenario, n, seed)
+        output_file = output_dir / f"drought_flood_impact_africa_{scenario}.csv"
+        df.to_csv(output_file, index=False)
+        print(f"Generated {output_file}: {len(df)} records")
+
+if __name__ == "__main__":
+    main()