Simulating the Breathing of the Biosphere
The biosphere and atmosphere interact. Climatic variables, including temperature, humidity, wind, and precipitation, affect individual plants and plant communities, while the type and diversity of plant cover influence the atmosphere. Understanding the breathing of the biosphere—the fluxes of carbon dioxide, water vapor, and trace gases between plants and the atmosphere—is a challenging task, involving numerous coupled, non-linear, biophysical processes.
At the regional to global scale, the computational demands require at least daily merging of several remote-sensing datasets. Scaling such a computation means scaling the science as well; for example, the science algorithm must encompass rainforests as well as croplands.
The Breathing Earth System Simulator (BESS) computation on the MODISAzure cloud addresses both the science and computational challenges. BESS scientists Youngryel Ryu and Dennis Baldocchi developed the science computation, which synthesizes data from satellite imagery, global climate models, and ground-based sensors. MODISAzure computer scientists, Jie Li, You-Wei Cheah, and Catharine van Ingen addressed the computational challenges by building a four-stage data processing pipeline on Windows Azure.
Cloud computing addresses three barriers to such a computation:
- Access to sufficient resources. The BESS computation used more than 500,000 CPU hours, ingesting 14 TB of data from NASA and generating 1.5 TB of results. MODISAzure scales from 5 to 240 deployed virtual machines in the cloud.
- Tedium. The global-scale science computation breaks into more than 70,000 tasks and consumes 500,000 input files. MODISAzure marshals the right input files for each task and handles synchronization and error recovery.
- Complexity. The bulk of the input files were the result of preprocessing approximately 800,000 initial NASA satellite files into “sinusoidal” files that contain the daytime aggregate value at fixed, equal-sized pixels.
Computations like BESS enable us to examine complex climate changes globally and locally. Global-scale fluctuations can be simulated with climate models, extending our understanding of complex climate change phenomena.
The results can also be combined with local observations to address specific questions, such the implications of rice farming in the Sacramento Delta region.
Learn more about this research:
Primary Researchers
Marty Humphrey is an associate professor in the Department of Computer Science at the University of Virginia. He received B.S. and M.S. degrees in Electrical Engineering from Clarkson University in 1986 and 1989, respectively. He received his Ph.D. degree in Computer Science from the University of Massachusetts in 1996. He has been on the faculty of the University of Virginia since 1988. Dr. Humphrey’s research interests include the use of cloud computing for cancer research and for environmental research.
Youngryel Ryu received his Ph.D. at the University of California, Berkeley, and participated in a summer internship at Microsoft Research in San Francisco after the Ph.D. course. He was a post-doctoral researcher at Harvard University, and is currently an assistant professor at Seoul National University. His main research topics include land-atmosphere interactions and application of computing resources for data-intensive science.
