Disaster Instability Early Warning Engine

Hazards become disasters when buffering capacity collapses under accumulated pressure. This project models disaster escalation as a transition, not as a one-off event.

Instead of predicting “damage” directly, the engine answers a more actionable question:

Is this event’s system equilibrium becoming unstable and which forces are driving it?

What This Project Is (and is not)

What it is

A diagnostic + early-warning engine that:

• computes an interpretable Instability Index (leading signal)

• decomposes each event into pressure forces vs buffer forces

• trains an ML model to estimate Major Disaster Escalation Risk (P(major))

• provides a Streamlit UI for:

  • single-event diagnosis
  • counterfactual scenario testing
  • cohort / map pressure-field visualization

What it is not

• Not a “Kaggle damage prediction model”
• Not a black-box catastrophe forecaster
• Not an automated decision maker

This is a human-centered decision-support system.

Dataset

Kaggle: Disaster Events 2025 by emirhanakku…