Former SIParCS interns evaluating next-generation climate models

By Brian Bevirt
09/17/2012 - 12:00am

Throughout its first six years, CISL's SIParCS program has been offering graduate and undergraduate students the chance to discover what careers would be like in computational science, computer science, applied mathematics, or the computational geosciences. This year, five former interns were at NCAR for a forward-looking workshop designed to find the most effective dynamical cores for next-generation global atmosphere models that will run on the most powerful supercomputers. (Dynamical cores calculate the fluid flow component of general circulation models.)

One of these five was Paul Ullrich, an organizer of the DCMIP workshop and a mentor for a working group that examined one of the dynamical cores. Also among these five, Matthew Norman lectured about using General Purpose Graphical Processing Units (GPGPUs) in atmospheric modeling; he was a member of SIParCS' first class in 2007. Another, Lucas Harris, both contributed to the workshop's lecture series and served as a modeling mentor for one of the dynamical cores.

In July–August 2012, the Dynamical Core Model Intercomparison Project (DCMIP) and Summer School on Future Generation Non-Hydrostatic Weather and Climate Models was hosted by the National Science Foundation, the U.S. Department of Energy (DOE) Office of Science, the National Oceanographic and Atmospheric Administration (NOAA), NCAR, and the University of Michigan. The purpose of this student-run international model intercomparison project was to test and evaluate 18 dynamical cores that represent a broad spectrum of the modeling approaches in the international weather and climate modeling community.

These five former interns credit SIParCS with advancing their careers, and some of them credit SIParCS with providing a valuable new direction for their careers. While they were all at NCAR in summer 2012, CISL asked for their perspectives on the program.

Matthew Norman, Oak Ridge National Laboratory
Matthew Norman, Oak Ridge National Laboratory. —Photo by Brian Bevirt, CISL

Matthew Norman, SIParCS class of 2007

Now working as a computational climate scientist in the Scientific Computing Group at the National Center for Computational Sciences at DOE's Oak Ridge National Laboratory, Matthew participated in the 2012 DCMIP workshop and presented an intensive lecture titled, "Emerging computational aspects and challenges for GCMs III: Some basics on General Purpose Graphical Processing Units (GPGPUs), pros and cons (and personal perspectives) of GPGPUs for atmospheric models, experiences and recommendations from a practitioner’s viewpoint."

In 2007, Matthew came from North Carolina State University to participate in CISL's first SIParCS class. As part of his summer project, he produced the paper titled "Non-Polynomial Based Conservative Remapping Schemes." Recalling that project, he said, "My main work was investigating non-oscillatory polynomial and non-polynomial reconstructions for use in atmospheric transport methods on the sphere, and it led to some interesting results. I thought my mentor [Ram Nair] guided me well, gearing things toward my interests while keeping the work applicable to science. It was an intense and fast-paced learning experience."

His SIParCS experience changed the course of his career. "This internship solidified for me a new direction I was wanting to take in my research as a Master's student, and it gave me the beginning contacts I needed to establish myself in the field of geophysical computational science. It gave me my first hard introduction to supercomputing as well. I changed my Master's thesis topic to the topic I worked on during this internship, and it was dense enough that I actually wrote my entire thesis while I was there that summer. I consider this the first big step that set into action the rest of my career. I believe it had a large influence on receiving DOE's Computational Science Graduate Fellowship, which in turn had a large influence on receiving my position as a staff computational scientist at Oak Ridge National Lab directly out of graduate school."

Looking back on his 2007 summer in the SIParCS program, Matthew said, "I think the strongest influences of the program are the contacts developed while working at NCAR and the introduction to the importance and challenges of supercomputing in science. So I believe [the SIParCS] leaders ... should seek to maintain and further improve these aspects."

Paul Ullrich, University of California at Davis
Paul Ullrich, University of California at Davis. —Photo by Brian Bevirt, CISL

Paul Ullrich, SIParCS class of 2008

One of the five organizers of the two-week DCMIP workshop and its associated two-week summer school, Paul returned to NCAR from the University of Michigan to present three lectures: "Numerical Methods I: Review of spatial (horizontal) discretizations," "Numerical Methods II: Review of temporal discretizations, numerical stability," and "Design philosophies of the MCORE dynamical core, discussion of the scientific reasoning and motivation behind the design." He also served in the workshop's breakout group sessions as modeling mentor for the dynamical core named MCORE.

When he came to NCAR in 2008, Paul had just completed his undergraduate studies at the University of Michigan. "SIParCS was my first experience with in-depth research at a national lab. I had just started my Ph.D. program and it was the first time I didn't have any course work and could focus full-time on research. So I worked for Peter Lauritzen and he had an idea about how to design a remapping scheme from a cubed-sphere grid to a lat-lon grid and vice versa to maintain conservative properties of tracers. This is very important for interfacing between the model components like a land model that runs on a lat-lon grid and an atmosphere model that runs on a cubed-sphere grid. My job was to come up with a scheme that interfaced between those two grids. We got things done so rapidly that we started a paper we finished a month after I returned to Michigan. [Ullrich, P.A., P.H. Lauritzen, and C. Jablonowski, 2009: Geometrically Exact Conservative Remapping (GECoRe): Regular latitude-longitude and cubed-sphere grids, Mon. Wea. Rev., 137 (6), 1721–1741.] It was a nice low-hanging research topic that was an excellent opportunity for somebody just starting off, and it got my feet wet in the topic." Peter Lauritzen was then the lead author for another paper based on an offshoot of this work: Lauritzen, P.H., R.D. Nair, and P.A. Ullrich, 2010: A conservative semi-Lagrangian multi-tracer transport scheme (CSLAM) on the cubed-sphere grid, J. Comp. Phys., 229 (5), 1401–1424. The paper Paul wrote for his SIParCS project was titled, "Conservative and Non-Conservative Remapping Schemes on the Sphere."

About his SIParCS experience, Paul continued, "It was a stepping stone for a lot of the research I did going forward. My Ph.D. research was the design of a high-order finite-volume method for atmospheric modeling. All the reconstruction work I'd done with Peter was the basis for the design of that finite-volume model. It was the first chapter in my thesis and a major component of that went into MCORE, the standalone experimental atmospheric model that came out of my research."

Paul's career is continuing to move forward: "I am now a post-doctoral researcher working for Christiane Jablonowski at Michigan. On September first I start as an assistant professor at UC Davis."

As a professor, Paul plans to continue supporting CISL's internship program. "SIParCS is an incredible opportunity for students. When I get my own students I'm going to strongly encourage them to participate in the SIParCS program. It can be a career-changing opportunity: going to NCAR, meeting with leading researchers, learning which topics are outstanding, and getting a feel for where things are heading in the field. NCAR is a place where everybody talks atmospheric science. If you're in an academic institution as a graduate student, you might be the only one working on your research topic. Even though there are others around you working in atmospheric science, it can be difficult to find somebody with whom you can have a close research relationship."

Pete Bosler, University of Michigan
Pete Bosler, University of Michigan. —Photo by Jennifer Williamson, CISL

Pete Bosler, SIParCS class of 2009

Now a Ph.D. candidate at the University of Michigan, Pete is planning to defend his thesis in March 2013. He participated in SIParCS 2009 as an undergraduate at Michigan. As part of his summer project, he produced the paper titled "Metadata for the Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS) using the Earth System Modeling Framework."

About that project, he said, "I was with the ESMF group at the Mesa Lab, which is a fantastic place to work. The group was led by Cecelia Deluca [who also participated in DCMIP], and my mentor was Sylvia Murphy [a speaker at DCMIP]. The main group was at the Mesa Lab with Bob Oehmke and Ryan O'Kuinghttons, who was another SIParCS student who later got a job with the ESMF group. We were collaborating with people from Princeton and Mississippi using very efficient conference calls. It was impressive to see a large software project that was so well coordinated. Everyone was working on code that potentially everybody else needed to use, and they were changing it all the time, but it was done in such a coordinated way that everything worked. It was a well-designed project that stressed documentation. Whenever you wrote code, it couldn't be submitted to the system until you wrote what it does, why it does it, and how you use it. My job was to put ESMF's metadata component into a Navy model for the Navy research lab at Stennis Space Center."

"The most immediate benefit of my SIParCS experience was how it improved my coding. The people I met and my exposure to the environment at NCAR continue to benefit me today. I was able to attend seminars at MMM [NCAR's Mesoscale and Microscale Meteorology Division] while I did my project, and I saw how science is done professionally instead of just as part of a thesis. When I came here I got a better idea of how research is really done. I'm now working on a Lagrangian technique for modeling fluids, and I'm doing a lot of coding, and my experience working with ESMF and seeing how code can be written and documented has really helped me."

Pete also offered these thoughts about his 2009 internship at CISL: "I was very grateful to be a part of SIParCS, because both my first- and second-choice projects were filled by other applicants. The SIParCS office then forwarded my appliction and CV to the ESMF group, who created that opportunity for me. So they matched my skills with a project where everybody benefited."

Lucas Harris, Geophysical Fluid Dynamics Laboratory
Lucas Harris, Geophysical Fluid Dynamics Laboratory. —Photo by Brian Bevirt, CISL

Lucas Harris, SIParCS class of 2009

Lucas recently completed a year and a half as a postdoc at Princeton University, and is now working as a physical scientist at NOAA's Geophysical Fluid Dynamics Laboratory (GFDL). Lucas participated in the 2012 DCMIP workshop, gave a presentation titled "FV3-GFDL: The GFDL Finite-Volume Cubed-sphere Dynamical Core" and served in the workshop's breakout group sessions as modeling mentor for the GFDL dynamical core.

In 2009, Lucas came from the University of Washington to participate in SIParCS. As part of his summer project, he worked with Peter Lauritzen and produced the paper titled "A flux-form version of the conservative semi-Lagrangian multi-tracer transport scheme (CSLAM)." Recalling that project, he said, "It was a good chance to do something different than what I was doing at Washington. My advisor, Dale Durran, does mostly limited-area mesoscale modeling. SIParCS was my first exposure to global modeling using a cubed-sphere grid. It was also my first experience developing numerical algorithms for tracer transport."

Regarding how his internship influenced his career, Lucas said, "SIParCS helped my career quite a bit. A couple months after my summer at NCAR, I got an email from a professor at National Taiwan University – I'd met him at a conference – who had just been visiting Shian-Jiann Lin at GFDL. He needed somebody who had experience doing both the sort of work I did at NCAR on the cubed-sphere grid and the sort of work I was doing at Washington on nested grids and grid refinement. After completing my Ph.D. the following year, I went to GFDL to work with S-J, first as a postdoc, and now as a physical scientist."

Lucas offered some insight into his time as an intern from the perspective of his work experience: "SIParCS is a very good program for people like me, grad students who need a different perspective than what I was getting at Washington. It really gave me something different to include in my research and broadened my options quite a bit. It was also beneficial to meet people here [at NCAR]. My advisor at Washington, Dale Durran, did his thesis here at NCAR working for Joe Klemp [both a speaker and MPAS modeling mentor at DCMIP], and SIParCS gave me a chance to network with some of these people who broadened my perspective."

Kiran Katta, University of Texas at El Paso
Kiran Katta, University of Texas at El Paso and NCAR Advanced Study Program (left) chats with his former SIParCS mentor Ram Nair. Ram was an organizer of the summer 2012 DCMIP workshop. —Photo by Brian Bevirt, CISL

Kiran Katta, SIParCS class of 2010

Kiran Kumar Katta's participation in the DCMIP working group included his work as a team member evaluating the FV3-GFDL dynamical core. In 2010, he was a graduate student at the University of Texas at El Paso when he came to NCAR for his SIParCS internship. As part of his summer project, he produced the paper titled "Third-order Non-oscillatory Transport Scheme." Now finishing his doctoral studies at the University of Texas at El Paso while working in NCAR's Advanced Study Program (ASP), Kiran said, "SIParCS is a very good opportunity to learn what research is being done at national labs, and how to contribute. I took advantage of my opportunities to meet people with immense knowledge in many areas."

Kiran's career was also influenced by his SIParCS experience. "It [SIParCS] was very helpful in deciding my career path, and I built a productive relationship with a scientist mentor [Ram Nair], who helped me find a long-term opportunity here as an ASP visitor. Now I am completing the project required for my doctoral thesis."

The SIParCS opportunity

SIParCS seeks highly qualified and motivated candidates for internships producing projects that advance CISL's research and development efforts in computational science, mathematics, software engineering, and facility engineering. All SIParCS projects meet real needs within the laboratory and advance its research priorites. They also provide work experience, networking contacts, and other career advancement for people who can serve as the computational Earth System sciences workforce of the future. Through its six years, the program continues to adapt and improve its mandate to integrate research and education to increase the number of trained scientists and engineers capable of using and maintaining 21st-century supercomputers.