Flood Vulnerability Mapping Project

Output(Deployment Link:)

Flood Analysis of Kurla region (Ctrl + Click to open in a new tab)

Overview

This project builds a building-level flood vulnerability analysis for Kurla (Mumbai) using:

• OpenStreetMap (OSM) vector features (buildings, waterways)
• DEM raster elevation data (kurla.tif)
• Feature engineering (elevation, distance metrics, terrain, hydrology)
• Unsupervised clustering (K-Means)
• Static and interactive map outputs
• GeoPackage/CSV exports for GIS tools (QGIS)

The main workflow is in the notebook:

• flood.ipynb

Project Files

• flood.ipynb - Main notebook containing the full pipeline
• kurla.tif - DEM used for elevation and terrain features
• thane_dem.tif - Additional DEM available in folder
• interactive_flood_map.html - Base interactive map
• interactive_flood_map_enhanced.html - Enhanced interactive map with more features
• flood_vulnerability_map.png - Static vulnerability map
• flood_vulnerability_data.gpkg - Base geospatial export
• flood_vulnerability_data_enhanced.gpkg - Enhanced geospatial export
• flood_vulnerability_feature_report.csv - Tabular feature export
• cache/ - OSM/network cache artifacts

Features Used

Core features

• elevation - Mean building elevation from DEM zonal stats
• dist_to_river - Distance from each building to nearest waterway

Added terrain features

• slope_deg - DEM slope at representative building location
• local_relief_m - Local elevation range near each building

Added hydrology features

• dist_to_major_waterway - Distance to major waterways
• waterway_density_300m - Waterway length density within 300 m neighborhood
• dist_to_drainage_m - Distance to nearest drainage-like feature (drain, ditch, canal)

Modeling

Current approach

• Model: K-Means clustering (n_clusters=3)
• Input features (base model): elevation, dist_to_river
• Cluster labels are mapped to risk classes:
  • High Risk
  • Medium Risk
  • Low Risk

Evaluation

Because K-Means is unsupervised, use clustering metrics (not supervised accuracy unless ground truth labels exist):

• Silhouette score
• Davies-Bouldin index
• Calinski-Harabasz score

Output Screenshots:

Snapshot of output of the interactive Map:

Snapshot of Don Bsoco Institute of Technology:

Environment Setup

1) Create/activate Python environment

Example with conda:

conda create -n geo_env python=3.10 -y
conda activate geo_env

2) Install dependencies

pip install geopandas osmnx rasterio rasterstats pandas scikit-learn matplotlib folium shapely pyproj

Optional:

pip install geopy

How To Run

Open flood.ipynb and run cells in sequence.

Recommended execution order:

1. Install/import cells
2. OSM fetch + DEM load
3. Feature engineering (elevation, dist_to_river)
4. K-Means clustering + risk mapping
5. Base static and interactive map export
6. Step 8A (terrain features)
7. Step 8B (hydrology features)
8. Step 10 (nearest drainage feature)
9. Step 9 (enhanced output export)

Note:

• Run Step 10 before Step 9 if you want dist_to_drainage_m included in enhanced exports.

Outputs

After successful run, key outputs are:

• flood_vulnerability_map.png
• interactive_flood_map.html
• interactive_flood_map_enhanced.html
• flood_vulnerability_data.gpkg
• flood_vulnerability_data_enhanced.gpkg
• flood_vulnerability_feature_report.csv

QGIS Usage

1. Open QGIS
2. Drag and drop:
   • flood_vulnerability_data.gpkg (base)
   • or flood_vulnerability_data_enhanced.gpkg (recommended)
3. Style by risk_level
4. Use numeric columns for symbology/analysis:
   • elevation
   • dist_to_river
   • dist_to_major_waterway
   • waterway_density_300m
   • dist_to_drainage_m

Common Issues and Fixes

1) Missing variables error in Step 8

If you see messages like missing buildings_with_features, dem_array, or affine:

• Re-run earlier pipeline cells (through clustering/feature creation), or
• Ensure required files exist in folder (kurla.tif, flood_vulnerability_data.gpkg)

2) No waterway/drainage returned from OSM

• Try increasing search radius in point-based fetch
• Use fallback logic already included in notebook
• Confirm internet connectivity

3) CRS or distance looks wrong

• Ensure metric CRS conversion is done before distance calculations
• Re-run Step 8A and Step 8B in order

4) DEM overlap problems

• Use debug cell in notebook to check vector/DEM bounds and CRS
• Confirm kurla.tif covers target area

Next Improvements

• Include rainfall intensity grids and drainage network quality as features
• Move from unsupervised clustering to supervised classification when labeled flood events are available
• Add model comparison and automated report generation

License / Data Notes

• OSM data is sourced via OSMnx and subject to OSM/OpenStreetMap data terms.
• DEM source/license depends on where kurla.tif was obtained.