Whether trust

We Don't just share climate data—prove it. WeatherTrust uses AI and ZK-proofs to verify satellite readings, anchoring absolute climate truth on-chain.

Comment

Inspiration

Climate decisions—from carbon credit verification to weather forecasting—depend heavily on satellite data. However, these readings are frequently corrupted by sensor drift, cloud contamination, and cyberattacks. Currently, users have no way to verify if the data they are using is authentic. If a malicious actor alters a temperature reading by even 1 degree, it can silently corrupt a multi-million dollar carbon credit or a climate model. We built WeatherTrust to secure the foundational layer of climate data. Our inspiration was simple: Don't just share climate data. Prove how far to trust it.

What it does

WeatherTrust is a 3D, cinematic intelligence platform that acts as a verifiable trust layer.

  • Global Target Acquisition: Users can search 40+ global regions (e.g., New Delhi, Tokyo, London). The platform instantly fetches live, real-time weather data from the Open-Meteo API and scatters simulated satellite data points across that specific geographic bounding box.
  • AI Cross-Verification: The AI compares the satellite reading against the live ground truth. If the delta (error) is too large due to a cloud edge or sensor drift, it calculates a low Trust Score (0-1) and flags the data red.
  • Trust Gate Filtering: Users can drag a real-time slider to instantly filter out low-trust anomalies on the 12K WebGL Earth.
  • Tamper-Proof Ledger: Every observation batch is hashed into a Merkle Tree. Users can click "Verify Carbon" to generate a cryptographic certificate, or click "Tamper" to simulate a cyberattack and watch the Merkle proof instantly break.

How we built it

We built a microservices architecture that balances heavy cryptography with high-fidelity 3D rendering:

  • Frontend (Live Demo): Built with Vanilla JS, Three.js (PBR rendering, real SGP4 orbital mechanics, 1,500+ instanced debris objects, custom GLSL shaders for Thermal Mode), and Chart.js.
  • Backend (In Repo): Built with FastAPI (Python) and Scikit-Learn (Isolation Forest ML model) to train and evaluate anomaly detection.
  • Web3/Crypto (In Repo): Built with Circom (Groth16 ZK-SNARKs), Solidity, and OpenZeppelin MerkleProof for the ClimateSentinel smart contract.
  • Data APIs: Integrated Open-Meteo for live ground truth weather data and the KeepTrack API for real satellite TLEs.

Challenges we ran into

  • Balancing 3D and Data: Rendering 1,500 debris objects and calculating real-time SGP4 orbital paths while maintaining 60fps was a massive WebGL challenge. We solved this using InstancedMesh and optimized post-processing (UnrealBloom).
  • Bulletproofing the Demo: Hackathon networks are notoriously unreliable. We built a bulletproof fallback system where the frontend will automatically generate mathematically accurate mock data if the Open-Meteo API goes offline, ensuring the 3D Earth never breaks during a live presentation.
  • Scientific Accuracy: We had to implement a geographic temperature gradient (where latitudes further from the equator are mathematically colder) so the simulated satellite data accurately reflected real-world physics.

Accomplishments that we're proud of

  • The Tamper Demo: Clicking a button and watching the Merkle hashes mismatch in the terminal log is the perfect way to prove why blockchain is necessary for climate data.
  • The Cinematic UI: Building a genuine 12K-style mission control dashboard with functional Thermal and God's Eye camera modes.
  • Mathematical Proof: We didn't just fake the AI. We trained an Isolation Forest model and evaluated its precision and recall, and we cross-verify our live frontend data against an independent Open-Meteo API.

What we learned

We learned how to bridge the gap between heavy cryptographic backend logic and high-fidelity frontend rendering. We also learned that the best way to present AI/Blockchain isn't through a boring dashboard, but through an immersive, interactive experience that lets the user see the data being corrupted and verified. Finally, we learned how to architect a system that remains visually stunning even when external APIs fail.

What's next for WeatherTrust

  • Integrating real GOES-16 ABI data streams via the FastAPI backend (moving beyond simulated data).
  • Deploying the ClimateSentinel.sol contract to the Ethereum Sepolia testnet with a live Etherscan oracle.
  • Expanding the ZKP circuits to prove carbon credit compliance without revealing the underlying proprietary supply chain data.
Share this project:

Updates