Search

link to homepage

Institute for Advanced Simulation (IAS)

Navigation and service


MPAS-A

Model for Prediction Across Scales – Atmospheric core

Scientific area

An atomspheric solver for fully compressible non-hydrostatic equations of motion on unstructured Voronoi meshes for a wide range of horizontal resolutions.

Short description

The Model for Prediction Across Scales is a collaborative project for developing atmosphere, ocean and other earth-system simulation components for use in climate, regional climate and weather studies. The primary development partners are the climate modeling group at Los Alamos National Laboratory (COSIM) and the National Center for Atmospheric Research. The defining features of MPAS are the unstructured Voronoi meshes and C-grid discretization. The unstructured Voronoi meshes, formally Spherical Centriodal Voronoi Tesselations (SCVTs), allow for both quasi-uniform discretization of the sphere and local refinement, while the C-grid discretization is especially well-suited for higher-resolution, mesoscale atmosphere and ocean simulations.

The MPAS atmospheric dynamical core (MPAS-A) solves the fully compressible nonhydrostatic equations of motion. The horizontal Voronoi mesh used by MPAS-A is depicted below. The variable resolution meshes a predominantly comprised of hexagons, but pentagons and septagons are occasionally present. The fully compressible non-hydrostatic equations are cast in terms of a geometric-height vertical coordinate, and the solver makes use of a split-explicit time integration scheme. The time-integration scheme employs a 3rd-order Runge-Kutta method, and large time step, for the meteorologically significant modes and a forward-backward method with smaller time steps for the acoustic modes. The numerical schemes used in the atmospheric component of MPAS are very similar to those employed in the Advanced Research WRF model. MPAS-A uses different versions of netCDF as primary format for parallel file I/O operations and was extended recently to support SIONlib for applications at extreme scale.

MPAS-C gridC-grid staggered variables on the Voronoi mesh. Horizontal velocities are defined on the cell faces, all other scalar variables are defined at the cell centers. The dual of the Voronoi mesh is shown in dashed lines in the figure.

Scalability

  • 458,752 cores on BlueGene/Q (JUQUEEN)
  • 131,072 cores on IBM System X iDataPlex, x86 (SuperMUC, LRZ)
  • 131,072 cores on Cray XC30, x86 (Edison, NERSC)

Performance of MPAS-A on JUQUEENScaling of time-integration (dynamics, file output) of uniform-resolution 2km mesh with 147,456,002 grid columns (cells) for an idealized Jablonowski and Williamson baroclinic wave experiment (physics is turned off).

Programming language and model

  • Fortran, C, C++
  • MPI, OpenMP
  • SIONlib for file I/O

Tested on platforms

  • BlueGene/Q
  • x86

Application developers

National Center of Atmospheric Research
Mesoscale and Microscale Meteorology Laboratory
Boulder, CO, USA

Los Alamos National Laboratory
Climate, Ocean and Sea Ice Modeling Group
Los Alamos, NM, USA

Code contributors

Karlsruhe Institute of Technology
Institute of Meteorology and Climate Research
Garmisch-Partenkirchen, Germany

Contact

Michael G. Duda
National Center of Atmospheric Research
Mesoscale and Microscale Meteorology Laboratory
Boulder, CO, USA
duda@ucar.edu

Dom Heinzeller
Karlsruhe Institute of Technology
Institute of Meteorology and Climate Research
Garmisch-Partenkirchen, Germany
heinzeller@kit.edu


(Text and images provided by the developers, c.f. Technical Report on the 2017 Extreme Scaling Workshop)


Servicemeu

Homepage