RESCUEAI: Pakistan's AI Flood Warning System Protects 9 Million People Using Satellites and WhatsApp

RESCUEAI (branded as BACHAO, Urdu for "save") is an AI-driven flood warning platform in Pakistan. It integrates NASA satellite imagery, real-time river level monitoring, machine learning algorithms, and WhatsApp to deliver timely, personalized evacuation routes and evacuation suggestions to 9 million people in flood-prone areas.

Its core goal is to address gaps in traditional warning systems, such as delayed information transmission, insufficient coverage, and lack of tailored guidance.

Pakistan, located in the Indus River basin, faces annual flood threats during monsoon seasons. The 2022 super flood caused over 1700 deaths, affected 33 million people, and led to $30 billion in economic losses.

Traditional disaster warning systems in the region have limitations like lagging information, limited reach, and no personalized guidance. RESCUEAI was developed to solve these pain points by combining satellite technology, real-time data, and instant messaging tools for flood-prone areas like Punjab province.

RESCUEAI's core architecture consists of four layers:

1. Multi-source Data Ingestion: NASA FIRMS/Sentinel-1 SAR satellite data (3-hour updates), real-time river level data from Pakistan Meteorological Department (30-minute updates), NDMA official notices (30-minute updates), and crowd-sourced reports via WhatsApp.

2. AI Risk Scoring Engine: Uses a weighted model (satellite data:40%, water level rise rate:45%, ground reports:15%) to calculate risk scores (0-100). Alert levels: Red (>80, emergency evacuation), Orange (60-79, risk prompt), Green (<60, normal).

3. Road Cut Prediction & Path Planning: Predicts road submergence time using road elevation, current water level, rise rate, and distance to river. Plans highest-altitude safe routes to nearest shelters.

4. WhatsApp Smart Bot: Supports user registration, road status queries, shelter location, and automatic red alert pushes (no extra app needed).

Technical highlights: Multi-language support (Urdu, Punjabi, Sindhi), node-cron for scheduled data updates, MongoDB with GeoJSON for geospatial data storage.

Currently, RESCUEAI covers 9 high-risk districts in Punjab province: Rajanpur, DG Khan, Muzaffargarh, Layyah, Multan, Bahawalpur, Rahim Yar Khan, Mianwali, Bhakkar (all along Indus River and its tributaries).

The system is designed for scalability—it can add new data sources and language packs to expand to other provinces like Sindh and Khyber Pakhtunkhwa.

Current Limitations: Cannot integrate Twilio SMS due to telecom policies; deep integration with local carriers (e.g., Jazz) requires additional certifications; no support for Facebook Messenger, Telegram, or email.

Future Plans: Access higher-resolution commercial satellite data; integrate weather forecast models for 72-hour advance warnings; develop offline mode for network outages; deepen integration with local government emergency systems for real-time rescue resource dispatch.

RESCUEAI's value lies in its design philosophy—using widely available tools (WhatsApp), open data (NASA satellites), and simple interactions to make AI accessible to ordinary people.

It provides a replicable model for developing countries to enhance disaster resilience amid climate change. The system proves that AI's true value is in delivering timely, actionable info to those most in need, not just complex algorithms.

Project link: https://github.com/manahils567-ux/RESCUEAI