AI summary by TwinMind
Gavin McCormick: Tracking the whole world's carbon emissions -- with satellites & AI | TED Countdown
TED
11m 15s
Current Climate Monitoring Problems
- Scientists have surprisingly little information about specific sources of greenhouse gas emissions
- Current process relies on asking polluters to self-report pollution levels
- Manual compilation of reports takes years to complete
- Some countries haven't had emissions inventory in 20 years
- Cannot distinguish between real emissions reductions vs. companies transferring polluting assets
- Only country-level data available, not specific source identification
- Expensive systems only accessible to wealthy nations
Climate TRACE Solution Overview
- Coalition of scientists, activists, and tech companies addressing emissions monitoring gaps
- Goal: real-time, detailed, open, transparent, and free emissions tracking system
- Uses AI and satellite imagery instead of self-reporting
- Covers every major emissions source globally
Technical Approach - Power Plants
- Uses computer vision AI algorithms trained on hundreds of thousands of satellite photos
- Identifies pollution levels by visual analysis of power plant emissions
- Example: power plant emitting 2,930 tons CO2/hour vs. zero when offline
- Multiple satellite sources: NASA Landsat 8, China's Gaofen 6
- Photos available every few days for every major power plant worldwide
- Multiple validation methods:
- Thermal infrared imaging to detect heat signatures
- Atmospheric emissions monitoring downwind
- Cooling water intake valve analysis (ripples in rivers indicate heavy water usage)
- Commercial imagery from Planet satellites
Expanded Monitoring Capabilities
- Al Gore and Generation partners challenged expansion beyond power plants
- Now covers all human emissions from all major sources globally
- Specialized organizations monitoring different sectors:
- Transition Zero: steel mills (UK-based)
- Synthetic: factory farms inventory (US EPA lacks complete inventory)
- RMI: oil and gas production/refining
- Blue Sky Analytics: crop fires and forest fires (India-based)
- Johns Hopkins: ground transportation and road networks
- Ocean Mine: shipping emissions (every ship on planet)
Technology Infrastructure
- Thousands of satellites now available at reduced costs
- Many satellite feeds are free and publicly accessible
- Big data and AI costs have decreased more than satellite costs
- Society spends more resources monitoring internet memes than climate crisis
- Climate TRACE reallocates fraction of these technical capabilities to emissions monitoring
Collaborative Structure
- Shared database model similar to Wikipedia
- Software engineers volunteering nights/weekends
- Academics validating algorithms
- NGOs running different models
- Sensor/satellite companies donating code
- Cross-validation required for all data
- Real-time sensors integrated (unlike Wikipedia)
Climate TRACE Version 1 Results
- Released September 21 before COP26
- Free access at climatetrace.org
- Covers every country, every sector, every year
- Examples: Malaysia rice production 2020, Australia electricity emissions 2020
- Comparison with country self-reporting shows most nations negotiating in good faith
- Some missing emissions found in specific industries
- AI accuracy improves over time
Version 2 Development
- Goal: make every single emitting asset worldwide visible
- Provides tools for:
- Governments identifying emission sources within economies
- Companies greening supply chains by comparing factory cleanliness
- Asset managers investing $43 trillion in net-zero goals
- Eliminates ability to hide emissions globally
Impact and Transparency Goals
- Supports Paris Agreement transparency requirements
- Enables countries to see others' actions and build trust
- Addresses resource limitations of expensive traditional monitoring
- Accelerates climate action through better measurement tools
