Geoscientific Model Development (Dec 2021)

Concurrent calculation of radiative transfer in the atmospheric simulation in ECHAM-6.3.05p2

  • M. R. Heidari,
  • Z. Song,
  • Z. Song,
  • E. Degregori,
  • J. Behrens,
  • H. Bockelmann

DOI
https://doi.org/10.5194/gmd-14-7439-2021
Journal volume & issue
Vol. 14
pp. 7439 – 7457

Abstract

Read online

​​​​​​​The scalability of the atmospheric model ECHAM6 at low resolution, as used in palaeoclimate simulations, suffers from the limited number of grid points. As a consequence, the potential of current high-performance computing architectures cannot be used at full scale for such experiments, particularly within the available domain decomposition approach. Radiation calculations are a relatively expensive part of the atmospheric simulations, taking up to approximately 50 % or more of the total runtime. This current level of cost is achieved by calculating the radiative transfer only once in every 2 h of simulation. In response, we propose extending the available concurrency within the model further by running the radiation component in parallel with other atmospheric processes to improve scalability and performance. This paper introduces the concurrent radiation scheme in ECHAM6 and presents a thorough analysis of its impact on the performance of the model. It also evaluates the scientific results from such simulations. Our experiments show that ECHAM6 can achieve a speedup of over 1.9× using the concurrent radiation scheme. By performing a suite of stand-alone atmospheric experiments, we evaluate the influence of the concurrent radiation scheme on the scientific results. The simulated mean climate and internal climate variability by the concurrent radiation generally agree well with the classical radiation scheme, with minor improvements in the mean atmospheric circulation in the Southern Hemisphere and the atmospheric teleconnection to the Southern Annular Mode. This empirical study serves as a successful example that can stimulate research on other concurrent components in atmospheric modelling whenever scalability becomes challenging.