This image portrays ocean surface currents as simulated by MPAS-Ocean, with acknowledgment to Los Alamos National Laboratory, E3SM, and the U.S. Department of Energy.
A newly developed solver algorithm for the MPAS-Ocean model marks a significant advancement in climate research. It accomplishes this by diminishing the time required for computations and enhancing the model’s precision. This innovation, which melds Fortran and C++ programming, represents a significant leap in climate modeling, both in terms of efficiency and reliability.
Ocean waves, while offering a calming backdrop on beaches, play a crucial role in scientific endeavors like weather prediction and climate studies. The ocean, alongside the atmosphere, is often a primary and computationally intensive element in Earth system models such as the Department of Energy’s Energy Exascale Earth System Model (E3SM).
Advancement in Oceanic Modeling
Contemporary ocean models generally examine two wave types: the fast-moving barotropic and the slower-moving baroclinic systems. To better simulate these two types simultaneously, a collaborative team from DOE’s Oak Ridge, Los Alamos, and Sandia National Laboratories has innovated a new solver algorithm. This algorithm has cut down the operational duration of the MPAS-Ocean, the ocean circulation component of E3SM, by 45%.
The team conducted software testing on various supercomputers, including Summit at ORNL’s Oak Ridge Leadership Computing Facility, Compy at Pacific Northwest National Laboratory, and Cori and Perlmutter at Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center. Their findings have been published in the International Journal of High Performance Computing Applications.
Progress in Climate Modeling Computing
The challenge of merging Trilinos, a collection of open-source software in C++ for supercomputing, with Earth system models typically written in Fortran was addressed using ForTrilinos. This software library integrates Fortran interfaces with existing C++ packages, streamlining the creation of the new solver focusing on barotropic waves.
Hyun Kang, an ORNL computational Earth system scientist and the study’s lead author, highlights the convenience of this interface, allowing the use of all C++ package components in Fortran without the need for translation.
Enhancements to MPAS-Ocean
This development builds upon earlier research published in the Journal of Advances in Modeling Earth Systems. Researchers from ORNL and Los Alamos National Laboratory manually created a code to enhance MPAS-Ocean. The new ForTrilinos-enabled solver addresses previous limitations, particularly when running MPAS-Ocean with fewer compute cores for a given problem size.
The original MPAS-Ocean solver used explicit subcycling, requiring numerous small time intervals to synchronize barotropic and baroclinic wave calculations without destabilizing the model. In contrast, the newly developed barotropic solver employs a semi-implicit method, ensuring unconditional stability and allowing the use of fewer, larger time steps without compromising accuracy. This approach saves considerable computing power and time.
The Trilinos and Fortrilinos software libraries, optimized by a community of developers for various climate applications, now enhance the performance of the MPAS-Ocean solver. These optimizations enable other scientists to expedite their climate research.
Future Developments and Impact
Despite its current scalability limitations on high-performance computing systems, the solver performs exceptionally up to a certain processor count. The semi-implicit method’s requirement for frequent processor communication can hinder performance. To address this, the researchers are optimizing processor communications and adapting the solver for GPU usage.
Furthermore, the team has refined the time stepping method for the baroclinic system, boosting MPAS-Ocean’s efficiency. These advancements aim to hasten and improve the reliability and accuracy of climate predictions, which are vital for climate security and for informed, high-resolution forecasting.
“This new solver for the barotropic mode not only speeds up computations but also ensures more stable integration of models, particularly MPAS-Ocean,” says Kang. “By enhancing the speed of this model, we can reduce the energy consumption associated with extensive computational resource use, thereby improving simulations and more effectively forecasting climate change impacts over decades or even millennia.”
This study was supported by E3SM and the Exascale Computing Project (ECP), both sponsored by the Biological and Environmental Research program in DOE’s Office of Science. The Advanced Scientific Computing Research program in DOE’s Office of Science funds the OLCF and NERSC.
Frequently Asked Questions (FAQs) about Climate Modeling Software
What is the significance of the new solver algorithm for MPAS-Ocean?
The new solver algorithm developed for the MPAS-Ocean model is a groundbreaking advancement in climate research. It significantly reduces computational time while improving the accuracy of simulations. This enhancement is achieved through the integration of Fortran and C++ programming, representing a major step forward in efficient and reliable climate modeling.
How does the MPAS-Ocean model contribute to climate research?
MPAS-Ocean, as part of the Department of Energy’s Energy Exascale Earth System Model (E3SM), plays a crucial role in climate research. It helps in forecasting weather and studying climate patterns by simulating ocean surface currents. The ocean, along with the atmosphere, is a key component in Earth system models, and accurate simulation of ocean currents is vital for reliable climate projections.
What are the computational improvements in the MPAS-Ocean model?
The new solver algorithm for the MPAS-Ocean model has reduced the total run time of simulations by 45%. This improvement is made possible by a semi-implicit solver for the barotropic system, which is unconditionally stable and allows for larger time steps without sacrificing accuracy. This leads to significant time and computing power savings.
What is the role of Trilinos and ForTrilinos in this research?
Trilinos is a database of open-source software, written in C++, used for solving scientific problems on supercomputers. ForTrilinos is a related software library that incorporates Fortran interfaces into C++ packages. The research team used ForTrilinos to design the new solver algorithm for MPAS-Ocean, enabling the use of C++ components in Fortran without the need for translation, thus streamlining the development process.
What are the future goals for MPAS-Ocean’s development?
Future enhancements for MPAS-Ocean include optimizing processor communications and adapting the solver for GPU usage. The team is also working on updating the time stepping method for the baroclinic system to further improve efficiency. The aim is to make climate predictions faster, more reliable, and more accurate, which is crucial for climate security and high-resolution projections.
More about Climate Modeling Software
- MPAS-Ocean Model Overview
- Department of Energy’s E3SM Project
- Trilinos Software Library
- ForTrilinos: Bridging Fortran and C++
- Ocean Waves in Climate Modeling
- High-Performance Computing in Climate Research
- The International Journal of High Performance Computing Applications
- Advances in Modeling Earth Systems
- Computational Efficiency in Climate Models
- DOE Office of Science Programs