SI_Flooding

Flood Risk Analysis and Visualization Toolkit

This project processes model-predicted flood areas, compares them to FEMA's Special Flood Hazard Areas (SFHA), and visualizes the results on a satellite basemap for any user-defined area of interest (AOI) in the United States.
Now supports loading a pre-trained flood prediction model via a .joblib file.

---

Features

• Loads flood prediction data exported from Google Earth Engine (GEE) as CSV with GeoJSON geometries.
• Optionally loads a pre-trained flood prediction model (.joblib) for local prediction.
• Converts CSV to a GeoDataFrame with robust geometry parsing.
• Fetches FEMA NFHL flood zone polygons for your AOI via the ArcGIS REST API.
• Identifies model-predicted flooded areas that fall outside FEMA's mapped SFHA zones.
• Visualizes all results on a satellite basemap, highlighting areas of interest.
• Saves outputs (GeoJSON, PNG, PDF, and JSON summary report) to an outputs/ directory.

---

Model Training and Results

Model Overview

The flood prediction model used in this toolkit is trained using satellite imagery and historical flood data. The workflow is designed to be flexible and can be adapted to any area of interest (AOI) in the United States.

• Input Data:
  The model leverages multi-temporal satellite imagery (e.g., Sentinel-1 SAR, Sentinel-2 optical) and, where available, historical flood event records.
• Features:
  Features may include spectral indices (NDWI, MNDWI), backscatter values, topography, land cover, and temporal change metrics.
• Labels:
  Training labels are derived from known flood extents, such as those mapped by government agencies or from high-confidence remote sensing flood maps.
• Model Type:
  A supervised machine learning classifier (e.g., Random Forest, Gradient Boosting, or similar) is trained to distinguish flooded from non-flooded areas based on the extracted features.

Using a Pre-trained Model (.joblib)

• You can use a pre-trained model saved as a .joblib file (e.g., flood_model.joblib) for local flood prediction.
• The script will load this model and apply it to your input features if specified in the configuration.
• Example:

  from joblib import load
  model = load('flood_model.joblib')
  predictions = model.predict(X)  # X is your feature matrix

Prediction and Export

• The trained model is applied to new satellite imagery or feature data to generate flood predictions for the AOI.
• Results are exported from Google Earth Engine as a CSV file, or generated locally using the .joblib model, with each row representing a spatial feature (polygon or point) and a flood_predicted value (1 for flooded, 0 for not flooded).

Results Interpretation

• The toolkit compares model-predicted flooded areas to FEMA's mapped Special Flood Hazard Areas (SFHA).
• Areas predicted as flooded by the model but not mapped as SFHA by FEMA are highlighted, helping to identify potential gaps in official flood risk mapping.
• All results are visualized on a satellite basemap for easy interpretation.

Note:
Model performance (accuracy, recall, etc.) will vary depending on the quality and quantity of training data, the AOI, and the features used. Users are encouraged to validate predictions with local knowledge or additional data sources where possible.

---

Requirements

• Python 3.8+
• pandas
• geopandas
• shapely
• matplotlib
• contextily
• requests
• joblib
• scikit-learn
• dataclasses (Python 3.7+)
• urllib3

Install dependencies with:

pip install -r requirements.txt

---

Quickstart

Get up and running in 3 simple steps:

from flood_analysis import FloodAnalyzer, FloodAnalysisConfig

# 1. Configure your analysis (add model_path if using a .joblib model)
config = FloodAnalysisConfig(
    csv_file_path='your_flood_predictions.csv',  # Path to your GEE export or feature CSV
    model_path='flood_model.joblib',             # Path to your pre-trained model (optional)
    output_dir='results'                         # Where to save outputs
)

# 2. Initialize and run analysis
analyzer = FloodAnalyzer(config)
report = analyzer.run_analysis()

# 3. View results
print(f"Analysis complete! Found {report['flood_predictions']} flood predictions")
print(f"Areas outside FEMA zones: {report['areas_outside_fema']}")

That's it! Your analysis will run automatically, create visualizations, and save all outputs to the specified directory.

---

Usage

1. Prepare your CSV
   Export your flood prediction data from GEE, or generate it locally using your .joblib model. The CSV must include:

   • .geo column (GeoJSON geometry as string)
   • flood_predicted column (1 for flooded, 0 for not flooded)
     or
   • Feature columns required by your model, if you want to generate predictions locally.

2. Place your CSV and model file
   Put your CSV and .joblib model file in the project root or specify their paths in the config.

3. Run the analysis

   python flood_analysis.py

   (Replace flood_analysis.py with your actual script name.)

4. Outputs

   • outputs/flood_predictions.geojson: All model predictions as GeoJSON
   • outputs/outside_fema_predictions.geojson: Model-predicted flooded areas outside FEMA SFHA
   • outputs/flood_analysis_map.png and .pdf: Visualization
   • outputs/analysis_report.json: Summary statistics

---

Configuration

Edit the configuration at the top of the script or in the FloodAnalysisConfig dataclass:

• csv_file_path: Path to your CSV file
• model_path: Path to your .joblib model file (optional)
• initial_crs: CRS of your data (default: EPSG:4326)
• fema_sfha_layer_id: FEMA MapServer layer ID for SFHA (default: 27)
• sfha_zones: List of FEMA flood zone codes considered as SFHA

Advanced Configuration Example

config = FloodAnalysisConfig(
    csv_file_path='data/miami_flood_predictions.csv',
    model_path='models/miami_flood_model.joblib',
    initial_crs='EPSG:4326',
    web_mercator_crs='EPSG:3857',
    fema_sfha_layer_id=27,
    sfha_zones=['A', 'AE', 'AH', 'AO', 'AR', 'A99', 'V', 'VE', 'VO'],
    output_dir='miami_analysis_results'
)

---

Troubleshooting

• No FEMA features found in AOI
  Your area of interest may not overlap with any FEMA-mapped flood zones. Check your bounding box and ensure your AOI is covered by FEMA data.

• Geometry parsing errors
  Ensure your .geo column contains valid GeoJSON strings.

• Basemap not displaying
  Requires internet connection and the contextily package.

• Model loading errors
  Ensure your .joblib file is present and compatible with your feature columns.

---

Example

---

License

MIT License

---

Acknowledgments

• FEMA National Flood Hazard Layer (NFHL)
• Google Earth Engine
• Open source Python geospatial community

---

For questions or contributions, please open an issue or pull request!