Influence of parallel computational uncertainty on simulations of the Coupled General Climate Model

Abstract

This paper investigates the impact of the parallel computational uncertainty on climate simulations using the Community Climate System Model Version 3 (CCSM3). A series of sensitivity experiments have been conducted and the analyses are focused on the Global and Nino3.4 sea surface temperatures. It is shown that the amplitude of the deviation induced by the parallel computational uncertainty is the same order as that of the climate system change. However, the ensemble mean method can reduce the influence and the ensemble member number of 15 is enough to ignore simulated errors. For climatology, the influence can be ignored when the climatological mean is calculated by using more than 30-yr simulations. It is also found that the parallel computational uncertainty has no effect on the simulated periods of climate variability such as ENSO. Finally, it is suggested that the influence of the parallel computational uncertainty on Coupled General Climate Models (CGCMs) can be a quality standard or a metric for developing CGCMs

Machine dependence as a source of uncertainty in climate models: The HadGEM3-GC3.1 CMIP Preindustrial simulation

“…Additionally, Figure 4d-9d indicate that model intercomparison analyses conducted on climatological (30 years) or decadal (10 years) timescales may be substantially influenced by the diversity of HPC platforms used to perform the simulations. This result is in contrast with what previously found by Song et al (2012), who stated that for (30-year) climatological means machine dependence uncertainty is negligible.…”

“…The question to answer is whether the spread between the same set of ensemble members run on the ARCHER and MO platforms is any different. Song et al (2012) addressed the same question (using the Community Climate System Model Version 3) showing that a minimum of 15 ensemble members are needed to make machine dependence uncertainty negligible.…”

“…To the extent of our knowledge, only a few authors discussed the existence of machine dependence uncertainty and highlighted the importance of bit-for-bit numerical reproducibility in the context of climate model simulations. Song et al (2012) and Hong et al (2013) investigated the uncertainty due to the round-off error in climate simulations. Liu et al (2015b) and Liu et al (2015a) discussed the importance of bitwise identical reproducibility in climate models, highlighting the lack of attention that the climate modelling community gives to the subject.…”

FIO‐ESM Version 2.0: Model Description and Evaluation

“…According to CMIP6 rules, for each numerical experiment, the ensembles are distinguished as “rKiLpMfN” by different numbers K, L, M, and N, which indicate different initial conditions, initialization methods, perturbed physics parameterizations and forcing, respectively, for the model configuration. The three realizations of historical run only differ in terms of initial conditions and are named r1i1p1f1, r2i1p1f1, and r3i1p1f1, respectively, and they have almost no effects on the climatological distribution and ENSO results (Song, Qiao, Lei, Wang, 2012). Therefore, in this paper, the time series analysis of historical simulation is applied to the ensemble mean and three realizations, while the first realization (r1i1p1f1) is analyzed for the evaluation of the climatological distribution and ENSO.…”