A New Approach for Coupled Regional Climate Modeling Using More than 10,000 Cores - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2015

A New Approach for Coupled Regional Climate Modeling Using More than 10,000 Cores

Marcus Thatcher
  • Fonction : Auteur
  • PersonId : 983348
John Mcgregor
  • Fonction : Auteur
  • PersonId : 983349
Martin Dix
  • Fonction : Auteur
  • PersonId : 983350
Jack Katzfey
  • Fonction : Auteur
  • PersonId : 983351

Résumé

This paper describes an alternative method for coupling atmosphere-ocean regional climate models that communicates momentum, radiation, heat and moisture fluxes between the atmosphere and ocean every time-step, while scaling to more than 10,000 cores. The approach is based on the reversibly staggered grid, which possesses excellent dispersive properties for modeling the geophysical fluid dynamics of both the atmosphere and the ocean. Since a common reversibly staggered grid can be used for both atmosphere and ocean models, we can eliminate the coupling overhead associated with message passing and improve simulation timings. We have constructed a prototype of a reversibly staggered, atmosphere-ocean coupled regional climate model based on the Conformal Cubic Atmospheric Model, which employs a global variable resolution cube-based grid to model the regional climate without lateral boundary conditions. With some optimization, the single precision, semi-implicit, semi-Lagrangian prototype model achieved 5 simulation years per day at a global 13 km resolution using 13,824 cores. This result is competitive with state-of-the-art Global Climate Models than can use more than 100,000 cores for comparable timings, making CCAM well suited for regional modeling.
Fichier principal
Vignette du fichier
978-3-319-15994-2_61_Chapter.pdf (4 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01328612 , version 1 (08-06-2016)

Licence

Paternité

Identifiants

Citer

Marcus Thatcher, John Mcgregor, Martin Dix, Jack Katzfey. A New Approach for Coupled Regional Climate Modeling Using More than 10,000 Cores. 11th International Symposium on Environmental Software Systems (ISESS), Mar 2015, Melbourne, Australia. pp.599-607, ⟨10.1007/978-3-319-15994-2_61⟩. ⟨hal-01328612⟩
86 Consultations
250 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More