Yellowstone’s Greatest Hits Highlighted at AGU Fall Meeting

By Marijke Unger
03/19/2017 - 4:45pm

NCAR’s Yellowstone supercomputer enjoyed a moment in the spotlight at the AGU Fall Meeting, held in San Francisco during the week of 11 December 2016. In a session that spanned both oral and poster presentations, scientists described successes and advancements in their research that were made possible through Yellowstone and similar resources.

The session, entitled “Success Stories from Yellowstone and Other Petascale Systems,” was co-chaired by NCAR Director Jim Hurrell and CISL’s Communications and Outreach specialist, Marijke Unger. Additional co-conveners for the session were Jim Kinter, Director of the Center for Ocean-Land-Atmosphere Studies (COLA) and Professor of Climate Dynamics at George Mason University, and Doug Nychka, Director of IMAGe in CISL.

Jennifer Kay, a University of Colorado professor and one of two invited speakers for the session, emphasized the importance of the Yellowstone resource beyond scientific research. “Yellowstone has been a transformative tool, not only for science but also for education,” Kay said. Kay, who was recently awarded the American Meteorological Society’s Henry G. Houghton Award for innovative use of observations and global climate models, took advantage of educational allocations to introduce her class to supercomputing. “The students loved logging onto a supercomputer, many of them had no prior experience at all,” Kay said, adding that “the CESM community and their commitment to pursue ‘community’ projects made the large ensemble project possible.”

Craig Schwartz, a researcher at NCAR’s Mesoscale and Microscale Meteorology (MMM) Laboratory, described the importance of supercomputing resources for forecasters and modelers, but like Kay, also noted that Yellowstone has served as “a surprisingly strong outreach tool.” According to Schwartz, “forecasters of all types, public, private, and amateur, have really embraced the forecasting system.” It has been used for power outage modeling, urban flooding modeling, and to develop critical severe weather guidance.

The second invited talk was given by Dimitri Mavriplis, a professor at the University of Wyoming, who has been working on high fidelity simulations of wind and air flow, using high performance computing to model wind farms and their impact. “We are aerodynamicists starting at the blades, and scaling up, rather than starting at the wind farm level and scaling down,” Mavriplis said, as he explained the importance of understanding air flow around the individual turbine blades, as well as the larger-scale turbulent flows across entire wind farms.

Airflow through wind turbine
From a simulation run on the Yellowstone supercomputer, these contour lines and isosurfaces provide valuable information about turbulence and aerodynamic drag in this visualization of air flow through the blades of a wind turbine. (Image courtesy of Dimitri Mavriplis, University of Wyoming)

Other NCAR presenters included:

  • Hanli Liu, of NCAR’s High Altitude Observatory, who mesmerized the audience with high resolution simulations of atmospheric gravity waves, which are an important mechanism for middle and upper atmosphere circulation and transport.
  • George Bryan, NCAR MMM scientist, who juxtaposed tornado and hurricane simulations, explaining that simulations really lead the way in understanding, given the challenges of direct observations. He also noted the difference in scale: a hurricane simulation takes about a million core hours on Yellowstone, while a tornado simulation takes about a 100 hours.
  • Jean-François Lamarque, Community Earth System Model (CESM) Chief Scientist and jointly appointed Senior Scientist for the Atmospheric Chemistry Observations and Modeling Laboratory (ACOM) and Climate and Global Dynamics (CGD), discussed using Yellowstone in support of the Coupled Model Intercomparison Project Phase 6 (CMIP6). According to Lamarque, “one of the strengths of Yellowstone over the years has been its stability. It has allowed us to run a beautiful, extremely successful experiment.”

University of Colorado professor Peter Hamlington discussed the role of Yellowstone and the importance of high performance systems dedicated to scientific research, citing the need to answer grand-challenge questions, such as what climate will look like 10, 20, 100, or 500 years from now. “A number of important insights have been obtained from these large simulations,” said Hamlington. “Sometimes they are surprising, as well.”

Nicolas Rodriguez Jeangros
Nicolas Rodriguez Jeangros, of the Colorado School of Mines, discusses his poster on “Spatiotemporal Categorical Map Fusion (SCaMF) – a methodology to fuse multiple categorical maps in space and time using parallel computing” during the AGU Fall Meeting poster session. (Photo by Marijke Unger)

Ben Cash, a scientist at COLA and GMU, echoed the importance of dedicated computing resources to advance scientific understanding. He presented results of a project that received early resource allocations on Yellowstone as part of the Advanced Scientific Discovery (ASD) initiative. ASD projects get early access to new supercomputers in an effort to both test the new system under a realistic load, and advance large projects that are ready to run and require extensive resources. Cash’s ASD simulation, called Project Minerva, focused on the South Asian Monsoon, which remains a particularly complex simulation challenge.

In addition to the talks, eight posters were also accepted to session. The poster abstracts appear on the 2016 AGU website.

“These talks and posters represent a wonderful cross-section of the scientific advances that are possible with dedicated, high performance computing systems,” said Marijke Unger, who co-chaired the session. “They are fascinating snapshots of projects that span the ‘small’ scale of turbine blades all the way to planetary system scales.”