loading page

Open-Source High-Performance Flux Transport
  • +4
  • Ronald Caplan,
  • Lisa Upton,
  • Jon Linker,
  • Charles Arge,
  • Raphael Attié,
  • Cooper Downs,
  • Carl Henney
Ronald Caplan
Predictive Science Inc.

Corresponding Author:caplanr@predsci.com

Author Profile
Lisa Upton
SSRC
Author Profile
Jon Linker
Predictive Science Inc.
Author Profile
Charles Arge
NASA Goddard Space Flight Center
Author Profile
Raphael Attié
NASA Goddard Space Flight Center
Author Profile
Cooper Downs
Predictive Science Inc.
Author Profile
Carl Henney
Air Force Research Laboratory Kirtland AFB
Author Profile

Abstract

The main objective of the NASA-NSF SWQU “A New-generation Software to Improve the Accuracy of Space Weather Predictions” effort is to develop a data-driven time-dependent model of the solar corona and heliosphere. This model will provide coronal and solar wind predictions and be made available to the public. One key component of this model is the use of a data-assimilation flux transport model to generate an ensemble of synchronic radial magnetic field maps to use as boundary conditions for the coronal field model. While flux transport models have long been established in the community, they are not open source or available for public use. We therefore are developing a new Open-source Flux Transport (OFT) software suite. The computational core of the OFT is the High-Performance Flux Transport code (HipFT). HipFT implements advection, diffusion, and data assimilation for the solar surface on a logically rectangular non-uniform spherical grid. It is written in Fortran and parallelized for use with multi-core CPUs and GPUs using a combination of OpenACC/MP directives and Fortran’s standard parallel ‘do concurrent’. To alleviate the strict time-step stability criteria for the diffusion equation, we use a Legendre polynomial extended stability Runge-Kutta super time-stepping algorithm (RKL2). The code is designed to be modular, incorporating various differential rotation, meridianal flow, super granular convective flow, and data assimilation models. Multiple realizations of the evolving flux will be computed in parallel using MPI in order to produce an ensemble of model outputs for uncertainty quantification. Here, we describe the initial implementation of the HipFT code and demonstrate its validation and performance. We use an analytic solution of surface diffusion and rigid rotational longitudinal velocity to validate the advection and diffusion implementations. We also compare realistic flux transport test problems against the established AFT flux transport code.