Data Intensive Computing
Adaptive Composite Analysis
Principal Investigator: Amanda White
Challenge
Researchers and scientists must be able to fuse massive data sets from multiple sources to identify trends and anomalies. For example, to validate some climate models, researchers must combine data from multiple satellites (each satellite generates ~60GB/day), identify cloud systems, and assess relationships between those systems and cirrus clouds.
Approach
- This project fuses data from three out of six "A Train" satellites to identify objects and relationships.
- The MeDICi architecture enables distributed data analysis to reduce data size before fusing.
Impact
- Cloud radiative feedbacks are considered the largest uncertainty in climate models. Understanding relationships between cloud properties will lead to improved cloud models to accurately predict climate changes.
- Models could be used to obtain atmospheric conditions not measured by satellites as an input to statistical analysis (example: vertical velocity)
Collaborations
- PNNL Climate Physics Group
Accomplishments
- Developed taxonomy for data fusion approaches
- Presented Human-Centered Fusion Framework to 2007 IEEE Conference on Technologies for Homeland Security
- Implementing MeDICi for real-time streaming data analysis
- New start NASA-funded activity
The A-Train
Cloud radiative feedbacks