Access to Carter cluster yields sunny forecast for atmospheric science students

March 12, 2014

Melissa Valentino, headed toward being a weather officer for the Air Force after graduation, knows the importance of computer modeling to her field so she already had incorporated a computer science minor into her educational plan.

But an ITaP and faculty partnership that gives students in Mike Baldwin’s classes training in high-performance computing — and the opportunity to run weather models on one of the world’s most powerful supercomputers — let the Purdue senior in atmospheric science take her plan to another level.

“This class and the experience with high-performance computing has definitely opened my eyes to the complexity involved in weather models and it gave me hands-on experience not many have at my age,” says Valentino, a captain and flight commander in Purdue’s Air Force ROTC wing.

With help from ITaP’s Stephen Harrell, Baldwin, an associate professor in the Department of Earth, Atmospheric and Planetary Sciences, employed one of Purdue’s community cluster supercomputers in his graduate level numerical weather prediction class during the fall semester of 2013. For spring 2014, he and Harrell are working with undergraduates in a senior level weather analysis and forecasting class.

The students get to use the Carter community cluster, which made the TOP500 list of the world’s most powerful supercomputers two years running. ITaP Research Computing (RCAC), which operates five community clusters for Purdue researchers, has set aside a portion of Carter, dubbed the Scholar Cluster, for any instructor with a need to incorporate high-performance computing in classes.

With weather modeling on high-performance computing systems now integral to the field, the skill is important for his students, Baldwin says. This is the case whether they’re continuing in research careers, forecasting for the Air Force, working as TV weather forecasters or consulting for businesses and industries, on which weather and climate can have major effects.

“Because atmospheric science is increasingly becoming a computing- and data-intensive discipline, there is a corresponding demand for those in our field to have programming, data management, analysis, and visualization skills,” says Kimberly Hoogewind, a second year doctoral student who also was in Baldwin’s 2013 class. “I feel that the exposure within the course will be extremely beneficial going forward, particularly when preparing to enter the job market.”

Some things Baldwin has his students do — such as ensemble forecasting using multiple models to gauge not only a forecast’s accuracy but its uncertainty, and running the state-of-the-art Weather Research and Forecasting (WRF) model — would be difficult, if not impossible, to do without using a supercomputer.

“If we didn't have high-performance computing, being able to run multiple models would take far too long to be useful,” Valentino says. “With Carter, we gained the ability to actually run WRF with real data and analyze it in a timely manner for the class.”

Baldwin and Harrell started working together when the professor became a faculty adviser to Purdue’s national student supercomputing competition team, sponsored by ITaP, which Harrell helps coach. They soon started talking about ways to incorporate high-performance computing in classes.

“We want to do more and more of this, weave the computing throughout our curriculum,” Baldwin says. “Who better to talk about high-performance computing than people at ITaP Research Computing who actually do it?”

Baldwin, in essence, teaches the students how numerical weather models work, while Harrell teaches them how to get the models working on a supercomputer, including skills to automate tasks and coax the software to take fuller advantage of the machine.

“It was a nice balance between technical information and practical application,” says Hoogewind, who will be using the community clusters heavily in her doctoral work. “For me, one of the main advantages of having Stephen available in the classroom was simply having him available as a resource. There were occasions where I had a computing question that wasn't completely related to what we were doing in class, yet Stephen was happy to lend his expertise.”

Harrell says ITaP is looking for more instructors who want to make use of the Scholar Cluster. For information, contact ITaP Research Computing at rcac-help@purdue.edu.

Originally posted: March 12, 2014