Nova 2.0

The aim of the project is to tackle the suddenness of GLOFs, and to create an alert system which helps in the prevention of mass destruction due to GLOFs.

Comment 1

Inspiration

Climate Change was something which was something evident but still, people refused to take step. Slowly, people started seeing the destruction Climate Change was bringing. Among them, one of the toughest and hardest to detect and prevent was Glacier Lake Outburst Floods. Glacial Lake Outburst Floods (GLOFs) are a growing but understudied natural disaster, accelerated by climate change. Existing monitoring systems are slow, manual, and reactive, providing warnings only when it's too late. 🌍 Key Factors That Inspired NOVA 2.0: ✔ The 2013 Kedarnath disaster, where 5,700+ people died due to a sudden GLOF. ✔ The 2021 Chamoli disaster, where an unexpected glacier break led to flash floods, killing 61 people. ✔ The 2023 Sikkim floods, highlighted the lack of real-time monitoring and prediction. 🚀 Our Goal: To create an AI-powered early warning system that predicts GLOFs before they happen, helping authorities take preventive action instead of reacting to devastation.

What it does

NOVA 2.0 is an AI-driven early warning system that predicts Glacial Lake Outburst Floods (GLOFs) by using satellite data, AI-powered risk scoring, and predictive flood modeling. ✅ Key Features: ✔ Glacier Irregular Breaking Tracking (GIBT) – Identifies stress fractures in glaciers before they collapse. ✔ AI-Based Risk Scoring – Continuously assesses glacial lake expansion, ice dam stability, and seismic triggers to assign risk levels. ✔ Real-Time Satellite Monitoring – Uses Sentinel-1, Sentinel-2, ICEYE SAR, and TerraSAR-X for high-precision tracking. ✔ Flood Impact Estimation – Predicts how far floodwaters will spread, allowing for early evacuation planning. Unlike traditional methods that react after a flood occurs, NOVA 2.0 predicts disasters before they strike.

How we built it

Building NOVA 2.0 required a combination of satellite data processing, AI-based risk modeling, and geospatial flood prediction. 📌 Phase 1: Data Collection & Processing Gathered real-time satellite imagery from Sentinel-1, Sentinel-2, ICEYE SAR, and TerraSAR-X. Used Google Earth Engine for geospatial data analysis. Applied Radiometric Terrain-Corrected SAR (RTC SAR) processing to remove distortions. 📌 Phase 2: AI Model Development Trained machine learning models on historical GLOF data. Implemented anomaly detection algorithms for glacier shifts and ice fractures. 📌 Phase 3: Glacier Irregular Breaking Tracking (GIBT) Designed an AI-based tracking model to predict glacier stress fractures and weak points. Integrated past case studies to improve prediction accuracy. 📌 Phase 4: Risk Scoring & Flood Prediction Developed a dynamic risk classification system for glacial lakes. Used GIS-based flood simulation to predict flood velocity, depth, and impact zones.

Challenges we ran into

Every innovation comes with challenges, and NOVA 2.0 was no exception. 🔹 Data Limitations: Glacial lakes are remote, making real-time monitoring difficult. Limited historical datasets for certain regions made AI training challenging. 🔹 AI Model Accuracy & Optimization: Initial models had false positives, predicting risk in stable lakes. Fine-tuning the AI model required multiple iterations. 🔹 Computational Constraints:

  1. Processing large-scale geospatial data required high-performance computing.
  2. Optimizing models for real-time prediction was difficult due to processing delays.
  3. Despite these challenges, we successfully built an AI model capable of detecting early warning signs weeks in advance.

Accomplishments that we're proud of

We are proud of the research and development done till now. While working on this project, it made us read a lot of research papers, and study many geographical phenomenons. We faced many issues like unavailability of satellite data, incorrect data, etc., and Google Cloud restrictions. To solve those we had to come up with innovative new ways like combining multiple data together, filling of the data in between, and using predictive analysis to identify incorrect or inaccurate spikes caused by other issues from the satellite.

What we learned

Throughout this journey, we gained valuable insights into AI-driven climate monitoring and disaster prevention. 📌 The Power of AI in Climate Science: AI can detect patterns that humans cannot, making early prediction possible. Machine learning models improve with more data, so real-time updates are essential. 📌 The Importance of Multi-Source Data Fusion: Combining multiple satellite sources (SAR, optical, RTC) improved accuracy. Historical case studies helped fine-tune risk prediction. 📌 The Need for Collaboration: Working with geospatial experts, climate scientists, and disaster agencies can improve deployment.

What's next for Nova 2.0

NOVA 2.0 is just the beginning. Our future roadmap includes: 📌 Expanding to High-Risk Regions: Deploy in other regions where glaciers are also retreating rapidly. 📌 Refining AI Models for Better Accuracy: Train models with more diverse datasets to reduce false positives. 📌 Enhancing Flood Impact Simulations: Improve hydrodynamic modelling to predict real-time water flow patterns. 📌 Collaborating with Disaster Response Agencies: Work with NDMA (India), UNDP, and global disaster relief organizations for large-scale deployment. 📌 Developing User-Friendly Risk Dashboards: Make risk analysis data accessible to government agencies and researchers. Our ultimate goal: To make NOVA 2.0 the global standard for GLOF early warning systems.

Built With

Share this project:

Updates