How does rising atmospheric CO2 affect marine organisms?

Click to locate material archived on our website by topic


Analyzing Observed vs CMIP5 Model Simulations of Global Temperature

Paper Reviewed
Chylek, P., Folland, C., Klett, J.D. and Dubey, M.K. 2020. CMIP5 climate models overestimate cooling by volcanic aerosols. Geophysical Research Letters 47, e2020GL087047.

Chylek et al. (2020) introduce their work by noting climate model simulations of global mean surface temperature diverge from the observational record around the turn of the 21st century, with "the model warming being somewhat too high." Two possible explanations for this discrepancy that they go on to investigate include (1) a failure of the models to properly account for aerosol forcings from volcanic eruptions and (2) a general underestimation of solar forcing.

The bulk of their analysis was focused on the first of these possible sources of error, where Chylek et al. compared the observed and model-simulated hemispheric mean temperature projections using a set of influencing factors, including anthropogenic greenhouse gases and aerosols, natural solar variability, volcanic eruptions and internal climate variability. Observed temperature data for the period 1890-2017 were obtained from the HadCRUT4.6.0.0 global data set whereas model simulated temperatures were acquired from 108 individual simulations performed by 43 CMIP5 models, supplemented with RCP4.5 projections after 2005.

In describing their findings, the authors report that "climate models overestimate the cooling effect of volcanic activity and underestimate the effect of the variability of solar radiation." Figure 1a illustrates this discrepancy using box and whisker plots of volcanic aerosol regression coefficients of the Northern and Southern Hemispheres that resulted from regression analyses of observed HadCRUT temperatures and ensemble means of CMIP5 model simulations. The probability that the observed differences occur by chance (p-value) was reported to be less than 10-16. Additional discontinuity between models and observations in this regard is shown in Figure 1b, which presents similar box plots for summer and winter seasons for the Northern Hemisphere.

Altogether, Chylek et al. report "the volcanic aerosol regression coefficients of the CMIP5 simulations are consistently significantly larger (by 40-49%) than the volcanic aerosol coefficients of the observed temperature." The hypothesized source of this discrepancy, they say, likely originates in faulty model parameterization of aerosol-cloud interactions within ice and mixed phase clouds.

Lastly, although it was not the major focus of their work the four scientists provide estimates of the individual contribution of the several key forcing factors examined in their study on both observed and simulated hemispheric temperatures. Examination of the equations displayed in Figure 2 reveals in the authors' words that, in addition to overestimating volcanic aerosol effects, "the CMIP5 climate models also significantly underestimate the effect of solar variability on the hemispherical and global temperature." In fact, Figure 2 reveals that "in CMIP5 models there is effectively no influence of solar variability on temperature, while the analysis of the observed temperature suggests quite a significant effect, especially on the Southern Hemisphere, consistent with the global results of Folland et al. (2018)."

Such large and significant discrepancies between observed and simulated hemispheric temperatures identified by Chylek et al. demonstrate that current state-of-the-art models are still not ready for primetime. Much more investigation and analysis must be conducted to identify and then correct these and likely many more other problems of which scientists are currently unaware before their projections can be trusted in policy formation.


Figure 1. (a) Box and whisker plots of volcanic aerosol regression coefficients of the Northern and Southern Hemispheres resulting from regression analyses of observed HadCRUT temperatures (blue shading) and ensemble mean CMIP5 model simulations (green shading). (b) Box plots of volcanic aerosol regression coefficients of HadCRUT temperature data and CMIP5 ensemble mean simulations in the summer and winter seasons of the Northern Hemisphere. Whiskers represent the minimum and maximum values (not counting outliers) and the box shows the first quartile, median, and third quartile. Source: Chylek et al. (2020).


Figure 2. Equations illustrating the influence of multiple forcing factors on Northern (NH) and Southern (SH) Hemisphere observed and CMIP5 model-simulated temperatures as presented in Chylek et al. (2020). GHGA = anthropogenic greenhouse gases and aerosols, SOL = solar irradiance, VOLC = volcanic aerosols, NINO = El Niño-La Niña variability, AMO = Atlantic Multidecadal Oscillation index. All uncertainties in brackets represent one standard deviation.

Reference
Folland, C., Boucher, O., Colman, A. and Parker, D. 2018. Causes of irregularities in trends of global mean surface temperature since the late 19th century. Science Advances 4, eaao5297.

Posted 15 July 2020