UW group’s software adopted by IBM to make weather forecasts faster

0
137
The NCAR-Wyoming Supercomputing Center, located in Cheyenne, is a critical resource for data validation, including the development of the IBM GRAF software led by UW Professor Suresh Muknahallipatna. UW engineering and applied science researchers and students have privileged access to the facility. (UW Photo)

LARAMIE, WYOMING (Dec. 2, 2019) — Suresh Muknahallipatna, a professor in the Department of Electrical and Computer Engineering at the University of Wyoming, led the successful completion of a software program that will allow for accurate weather forecasts up to 12 times faster than traditionally possible.

 

 

The Weather Co., a subsidiary of IBM, announced the global rollout of IBM GRAF (Global High-Resolution Atmospheric Forecasting System) Nov. 14 through a media release with Business Insider.

Muknahallipatna led the high-performance computing group, a research group within UW’s Department of Electrical and Computer Engineering, to develop the software.

Since 2014, the group has been receiving research grants from the National Center for Atmospheric Research (NCAR) to refactor climate codes to execute on supercomputers using central processing units (CPUs) and accelerator hardware architectures.

The successful outcomes of these grants demonstrated to NCAR scientists the expertise and capability of the UW research group in refactoring and optimizing codes for speed. This led to the group being selected in 2017 to work on this project. The Weather Co. then collaborated with NCAR on IBM GRAF to develop the new modeling system.

NVIDIA Corp. — which is globally recognized as the manufacturer of the graphical processing units (GPU) that provide parallel processing capabilities to efficiently run high-performance computing (HPC) applications — donated GPU hardware to set up a HPC development cluster within Muknahallipatna’s department at UW.

Until recently, weather forecasts were updated every six to 12 hours, covering 10-15 square kilometers, using software executed on supercomputers using only CPUs. This created greater room for error when weather patterns shifted, making it difficult to accurately report forecasts for public consumption.

IBM GRAF is a refactored software that runs on supercomputers using both CPUs and GPUs, and can give hourly updates covering three square kilometers.

 

Muknahallipatna said the level of precision IBM GRAF will create will better serve the public.

“The power to forecast weather is a monumental achievement, and it is given importance only when the forecast is wrong,” Muknahallipatna said. “By updating at an hourly rate, the errors in forecasting can be corrected and reduced. And, with higher resolution of three square kilometers, the forecast can be localized.”

In the early stages of the HPC research group, the Department of Electrical and Computer Engineering and the UW Office of Research and Economic Development provided funding by returning the generated indirect cost from NCAR grants to support additional graduate students to work on NCAR research grants. Muknahallipatna was joined by several former and current UW students to help bring the project to life, including:

  • George Dickerson, a graduate student from Casper
  • Brett Gilman, an undergraduate student from Powell
  • Briley James, an undergraduate student from Laramie
  • Pranay Kommera, a Ph.D. candidate from Hyderabad, India
  • Sumathi Lakshmiranganatha, a Ph.D. candidate from Bengaluru, India
  • Henry O’Meara, a graduate student from Rapid City, South Dakota
  • Suzanne Piver, an undergraduate student from Cheyenne
  • Bradley Riotto, of Jackson, who graduated with a Master of Science degree earlier this year.

The software and the mathematical models of weather forecasting are developed and updated regularly by the scientists and programmers at NCAR. The expertise of their scientists on weather models is crucial in refactoring and validation of the code. Several individuals from NCAR’s Computational and Information Systems Laboratory (CISL) and the Mesoscale and Microscale Meteorology Laboratory (MMML) were involved in the production of IBM GRAF, along with teams from NVIDIA Corp.

Muknahallipatna thanks the following individuals for their contribution to the project:

  • Richard Loft, director, Technology Development Division, CISL, NCAR
  • Michael Duda, software engineer IV, MMML, NCAR
  • Supreeth Suresh, software engineer II, CISL, NCAR
  • Cena Miller, software engineer, CISL, NCAR
  • Raghu Kumar, senior DevTech, NVIDIA Corp
  • Carl Ponder, senior DevTech, NVIDIA Corp
  • Brent Leback, compiler manager, NVIDIA Corp

Muknahallipatna earned his bachelor’s and master’s degrees in electrical engineering from the University of Bangalore in India. He went on to earn his doctorate in electrical engineering in 1995 from UW, where he has since been a faculty member with the Department of Electrical and Computer Engineering.

 

His research interests include deep learning, machine learning, high-performance computing and augmented reality in medical applications.

Muknahallipatna and the high-performance computing group plan to continue research on refactoring codes of other weather forecasting models and the integration of the models with the Community Earth System model through their current NCAR research grant.