PREDICTING THE FUTURE CLIMATE OF THE U.S.

August 11, 2000

by Paul Tooby, SDSC

La Jolla, CA — One of the most significant unsolved scientific problems of our time does not lie billions of light-years away, nor is it hidden in the microscopic world of the atom or the gene. It is something that all of humanity is immersed in every day: Earth’s climate system.

For years, however, U.S. climate-modeling efforts have been falling behind European models and computing resources. The National Research Council has recently cautioned that “the United States lags behind other countries in its ability to model long-term climate change” and that it is inappropriate to rely so heavily on foreign climate modeling. Because climate science has major implications for energy policy, without a climate modeling capability the U.S. is seriously hampered when considering energy issues or negotiating future international energy policy.

To address the need for improved climate modeling, the Accelerated Climate Prediction Initiative (ACPI), a multi-institution effort including both government laboratories and universities, is working to create a state-of-the-art climate modeling capability for the U.S. Major participants include the Department of Energy (DoE), National Center for Atmospheric Research (NCAR), Los Alamos National Laboratory (LANL), Scripps Institution of Oceanography (SIO), Naval Postgraduate School (NPS), Pacific Northwest National Laboratory (PNNL), the University of Washington, U.S. Geological Survey (USGS), and the Desert Research Institute (DRI).

To “jump start” this major long-term program, SIO researcher Tim Barnett is coordinating a collaboration of researchers in an ACPI pilot demonstration. Co-PIs in the pilot program include Robert Malone of LANL, Bill Pennell of PNNL, Bert Semtner of NPS, Detlef Stammer of SIO, and Warren Washington of NCAR. “The pilot program we’re doing as part of the larger ACPI is intended to demonstrate all the key parts of the larger climate prediction initiative in an ‘end-to-end’ process. The different model components-ocean, atmosphere, land, sea ice, and other parts-have been integrated into a global climate model, the Parallel Climate Model (PCM), which we’ll start by using today’s observed ocean as initial conditions, then run climate simulations under various scenarios of greenhouse gas changes through 2050 or 2100, and finally downscale from global to regional predictions,” Barnett says.

There are three basic parts of this pilot program. The first step is to start the Parallel Climate Model with the initial conditions of the world’s oceans as they are now, which is important because the oceans hold some 99 percent of the heat in the combined ocean-atmosphere system. Including the actual ocean state data is new science that hasn’t been done before because these data have only recently become available. Previously, to model the impact of greenhouse gas emissions, researchers would have to start their climate models in 1860, before greenhouse gases had changed, and integrate forward. So, in addition to saving more than 100 years of computing simulation time, being able to use current ocean data ties the model more realistically to the real world. “Our simulations both ways, integrating forward from 1860 and starting with today’s ocean data, have turned out surprisingly close, giving us confidence in this climate model, the PCM,” Barnett says. The ocean data comes from SIO researcher Detlef Stammer’s Ocean General Circulation Model (OGCM), which assimilates or combines all existing data about the oceans into an overall description of the present ocean state.

The second step of the pilot program is to run the full global climate model, the PCM, using different scenarios of greenhouse gas emissions that come from fossil fuel burning and other industrial and agricultural activities, and vary depending on such factors as economic and population growth and conservation measures. The PCM contains models for major components of the climate system-ocean, atmosphere, land, and sea ice-and predicts global climate decades in the future at a spatial resolution of about 280 km. Typical scenario lengths are between 50 and 100 years, taking about 320 processor hours per model year. To average out model variability, ensembles of five to 10 simulations need to be run for each scenario, placing additional demands on computing resources.

In the third step, the global-scale climate prediction data from the PCM is passed to another set of models called regional climate models developed to “downscale” the global climate predictions to the regional scale. This is important because regional predictions are what governments, policymakers, and industry need in order to make decisions. “In the pilot program we’re focusing on the hydrology of the western U.S., and after the initialization and global PCM steps, eventually we predict the amount of water that will be running down, say, the Merced River 50 years from now. So we’ll be producing predictions that have practical use right from the start,” Barnett says. But downscaling uses much higher resolution, requiring more computing time. SIO researcher David Pierce points out that “doubling the resolution of the grid in 2-D space means you also have to cut the time steps in half, so it’s a factor of eight more calculations.” This fine resolution is required to provide meaningful data to the hydrology models that do the final step of predicting streamflow in Western rivers in the decades to come.

Although U.S. climate models have lagged behind European models, researchers at NCAR working with other groups have now assembled model components for the oceans, atmosphere, land, and sea ice into the fully coupled Parallel Climate Model (PCM). The atmospheric component of the PCM is the Community Climate Model 3 (CCM3) atmospheric general circulation model (AGCM) developed at NCAR, which is used at T42 resolution (a grid in which each side of a cell measures about 2.8 degrees or 280 kilometers). The CCM3 includes a land surface model that accounts for soil moisture and vegetation types as well as a river transport model developed at the University of Texas at Austin. The ocean component is the Parallel Ocean Program (POP) model developed jointly by LANL, NPS, and NCAR. The final model component is a sea ice model developed at NPS. The full PCM has been configured to run with a serial flux coupler designed to calculate the climate system components as efficiently as possible on a variety of parallel high capacity supercomputers.

While climate models are still maturing and have significant limitations, several factors, including progress in climate science, the availability of more and better climate data, advances in modeling and computing technologies, as well as the development of terascale computing resources, are all converging to put climate modeling on the verge of major progress in predictive capability.

“I think we’ve already solved the hardest scientific problem of our pilot program, which is taking the ocean data and initializing the PCM with that,” Barnett says. If researchers put the ocean data directly into the PCM, there is a mismatch that produces model drift. Fortunately, SIO researchers have found a solution, and SIO’s David Pierce is now running tests. Once this method is working it will be transferred to NCAR partners Warren Washington and his group, who operate the PCM. “They’ll go into production and run a series of climate scenarios from 2000 to at least 2050 and maybe 2100 if the computer time’s available. And we can now do runs like this here at SDSC because we have the PCM model up and running on Blue Horizon,” Barnett says.

Climate modeling involves many large interlocking models maintained by different groups that must be ported to various supercomputing platforms. SDSC scientist John Helly has worked with SIO’s Tim Barnett and others to ensure that the Parallel Climate Model will run on NPACI resources. In a Strategic Applications Collaboration (SAC), Giri Chukkapalli and Dong Ju Choi, members of SDSC’s Scientific Computing Department, worked with NCAR researchers Tom Bettge, Tony Craig, and Vincent Wayland to port the PCM to the IBM SP and NPACI IBM Blue Horizon teraflops platform. “To enable larger simulations, the POP ocean component of the PCM maintained at LANL is being modified to scale past 64 processors, and we’ve worked closely with researchers there to help them make these changes,” Chukkapalli says.

Barnett and his colleagues in the ACPI pilot program will use the largest NPACI resources including Blue Horizon for three purposes. While all the pieces of this project have been used separately, they’ve never been put together in a seamless, programmatic form. “So, one step is to do the studies on how we initialize this model, that’s the real science in this pilot program. That involves a number of test runs on NPACI resources,” Barnett says. Second, researchers may also do global scenario runs with the PCM on NPACI resources, where they change the greenhouse gases and integrate forward to the next century. “And the third and possibly most important use of the NPACI resources is to run the downscaling model to produce the regional impact statements, which involves the high-resolution nested models that take a great deal of computing time,” Barnett says.

Climate models are inherently among the most computationally demanding scientific models in both the length of simulations required and the volume of model data produced. The reason for this lies in climate dynamics: Researchers need to discern the relatively small global climate change “signal” or “signature” above the larger “noise” of the significant natural variability of the climate system. While the “memory” of the atmosphere is less than one month, the “memory” of the oceans, due to their high heat capacity, spans many hundreds of years, requiring long simulation times. Moreover, researchers need to run ensembles of five to 10 identical simulations to average out model variability, essentially noise generated by the nonlinear properties of the model itself.

Thus, climate simulations require very large amounts of computer time, and part of the ACPI pilot program involves not only porting and optimizing codes to run efficiently on various supercomputing platforms, but also learning how to compute these large models from different groups, exchange information, and make everything work seamlessly “end-to-end.” And because the models are too large and numerous to run at one place, part of the research is learning to work efficiently with different computing centers. “NPACI has a good allocation policy where you have a year to use the time, which makes it much more feasible for us to manage this complicated project,” Barnett says.

As new technologies have enabled larger and more complex research problems such as multi-component climate modeling to be undertaken, researchers are learning to work in larger collaborations across many disciplines and institutions. “We’ve deliberately organized this pilot program in a low profile way, with distributed management. So, we’re really testing the concept that a group of people can come together, make a decision on what needs to be done, each having a responsibility, and take that away,” Barnett says. “And so far it’s working very well.”

In the two year pilot program, ACPI researchers will demonstrate an end-to-end climate modeling capability, starting with real initial ocean conditions and proceeding through global simulations with various greenhouse scenarios to regional-scale predictions for the Western U.S. “The answers we get will be useful, but certainly not the final answers, and really what we’re doing is breaking ground to identify what the problems are in the full program,” Barnett says. “We’ve already got a page or two of things that should be handled differently in later stages. We’ll wind up with a blueprint for a much larger national program to bring a state-of-the-art climate modeling capability to the U.S.”

============================================================

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire