Predict. Prepare. Protect. A high-precision, AI-driven emergency management platform that replaces reactive disaster response with proactive, ML-powered intelligence — built specifically for Bihar's annual flood crisis.
A high-fidelity GIS interface providing district-level forensic risk coloring and real-time situation room telemetry.
An interactive 7-day projection engine where researchers can adjust precipitation and river parameters to simulate hydraulic outcomes.
Live monitoring of 18 gauge stations across 5 major basins with active XGBoost-driven breach probability monitoring.
Bihar is India's most flood-prone state, with 76% of the North Bihar population living under recurring threats. Every monsoon, rivers like the Kosi and Gandak breach aging embankments, causing catastrophic damage.
- Information Latency: Manual gauge reporting takes 6–12 hours to reach decision-makers. By then, villages are often already submerged.
- Reactive Response: Current systems deploy after flooding begins. Evacuation windows are missed, and resource staging is delayed by critical hours.
- Infrastructure Decay: Embankments built in the 1950s have minimal maintenance budgets. No predictive model exists to prioritize repairs based on breach probability.
- Resource Inefficiency: Relief is distributed evenly rather than by statistical vulnerability, leaving the most at-risk zones under-served.
| System | Primary Function | The Forensic Gap & Failure Point |
|---|---|---|
| CWC Forecasting | Text-based bulletins | Manual collection with 12hr lag; no spatial visualization or ML logic. |
| FMIS Bihar | Web portal data tables | Raw data only—lacks actionable intelligence or "What-If" scenario logic. |
| NDMA Decision Support | Macro-level coordination | Static risk maps based on history, not dynamic real-time ML triggers. |
| Academic Research | Published ML papers | High theoretical value but never deployed as a functional, real-time UI. |
Core Thesis: If we can predict where and when an embankment will breach — before it happens — we can transform flood response from reactive chaos into proactive precision.
AERIS uses dual XGBoost models trained on Bihar-specific river telemetry, rainfall intensity, and infrastructure age to provide:
- Live Risk Map: Interactive GIS with 24-hour risk trend sparklines and population-at-risk counters.
- Forecast Simulator: Multi-variate 7-day simulation engine with granular model execution logs.
- River Pulse: Real-time telemetry monitoring (18 stations) with automated alert color-coding.
- Resilience Scorecard: Multi-dimensional ranking of district preparedness (Infrastructure, Medical, Shelter).
- Post-Mortem: Forensic hydraulic stress analysis and timeline reconstruction (Learning from History).
- ML Predictor: Manual parameter override for "What-If" scenario execution.
- Research Review: Scholarly technical overview of the system's architecture and problem-solution fit.
- ML Engine: XGBoost (Dual models: Breach Prediction + Flood Classification).
- Backend API: FastAPI (Python) with real-time inference latency optimizations.
- Frontend GUI: React 19 + Vite + Tailwind CSS 4 + Framer Motion.
- GIS Core: React-Leaflet + GeoJSON with responsive zoom accessibility (7.5 - 8.5).
- Design System: Editorial Design (Playfair Display, Charcoal/Teal palette, Glassmorphism).
# Install dependencies
pip install -r requirements.txt
# Start the API server
python -m uvicorn api.main:app --host 0.0.0.0 --port 8000 --reloadcd frontend
npm install
npm run dev- Evacuation Lead Time: 7-day predictive warnings enable pre-positioned resources.
- Localized Precision: Trained on Bihar-specific river basins, not generic national rainfall models.
- Forensic Accountability: Transparency in district rankings drives infrastructure investment.
AERIS — Built for Bihar. Powered by AI. Designed for Impact.