by Judith Curry
What are the implications of climate model deficiencies in simulating multi-decadal natural internal variability for IPCC’s climate change detection and attribution arguments?
Two recent papers raise some significant concerns in this regard:
The Atlantic Multidecadal Oscillation in twentieth century climate simulations: uneven progress from CMIP3 to CMIP5
Alfredo Ruiz Barrados, Sumant Nigam, Argyro Kavvada
Abstract (excerpts) Decadal variability in the climate system from the Atlantic Multidecadal Oscillation (AMO) is one of the major sources of variability at this temporal scale that climate models must properly incorporate because of its climate impact. The current analysis of historical simulations of the twentieth century climate from models participating in the CMIP3 and CMIP5 projects assesses how these models portray the observed spatiotemporal features of the sea surface temperature (SST) and precipitation anomalies associated with the AMO. The structure and evolution of the SST anomalies of the AMO have not progressed consistently from the CMIP3 to the CMIP5 models. While the characteristic period of the AMO (smoothed with a binomial filter applied fifty times) is underestimated by the three of the models, the e-folding time of the autocorrelations shows that all models underestimate the 44-year value from observations by almost 50 %. Variability of the AMO in the 10–20/70–80 year ranges is overestimated/underestimated in the models and the variability in the 10–20 year range increases in three of the models from the CMIP3 to the CMIP5 versions. Spatial variability and correlation of the AMO regressed precipitation and SST anomalies in summer and fall indicate that models are not up to the task of simulating the AMO impact on the hydroclimate over the neighboring continents.
Published in Climate Dynamics [link].
Excerpts from the Introduction provides some background:
Decadal climate prediction has taken a prominent role for the first time in the experiments of the Coupled Model Intercomparison Project Phase 5 (CMIP5). The need for useful decadal predictions has been made not only from scientific papers, but also from the impact of climate-related events like the current melting of the Greenland glaciers, the ongoing drought in northern Mexico and central US, as well as past decade-long droughts over the same region in the recent twentieth century and over western Africa. Therefore, if one aspires to have reliable decadal predictions, climate models have to properly incorporate the processes that give rise to decadal variability in specific components of the climate system, in addition to the mechanisms through which these processes impact the surface climate affecting human societies.
Phenomena with defined decadal variability that climate models must properly include are the Pacific Decadal Oscillation (PDO) and the Atlantic Multidecadal Oscillation (AMO). Decadal control of hydroclimate from the AMO over North America and Africa is one of the main reasons to worry about having this phenomenon properly incorporated in climate models. Multi-year, summer and fall droughts over North America and Africa have been observationally linked to decadal SST variability in the Atlantic.
Excerpts from Concluding Remarks:
Decadal variability in the climate system from the AMO is one of the major sources of variability at this temporal scale that climate models must aim to properly incorporate because its surface climate impact on the neighboring continents. If climate models do not incorporate the mechanisms associated to the generation of the AMO (or any other source of decadal variability like the PDO) and in turn incorporate or enhance variability at other frequencies, then the models ability to simulate and predict at decadal time scales will be compromised and so the way they transmit this variability to the surface climate affecting human societies.
The key point from the paper is this:
Variability of the AMO in the 10–20/70–80 year ranges is overestimated/underestimated in the models and the variability in the 10–20 year range increases in three of the models from the CMIP3 to the CMIP5 versions.
This is more completely stated in the text (note: Fig 3 is the key figure IMO, which I couldn’t figure out how to extract):
It is clear that both sets of CMIP3 and CMIP5 models underestimate low frequency variability in the 70–80 and 30–40 year ranges while overestimate variability in the 10–20 year range. Variability in the higher 10–20 year range increases from CMIP3 to CMIP5 in three of the models surpassing the variability in this range from observations.
Implications for detection and attribution of climate change
While the Barrados et al. paper focuses on the implications related to regional hydroclimates, I think perhaps a more significant implication of their analysis is on the fitness-for-purpose of climate models for detection and attribution of climate change over the last century. IPCC AR4 defines detection and attribution in the following way:
‘Detection’ is the process of demonstrating that climate has changed in some defined statistical sense, without providing a reason for that change.[T]he methods used to identify change in observations are based on the expected responses to external forcing, either from physical understanding or as simulated by climate models. An identified change is ‘detected’ in observations if its likelihood of occurrence by chance due to internal variability alone is determined to be small.
[W]hen fingerprints from Atmosphere-Ocean General Circulation Models (AOGCMs) are used, averaging over an ensemble of coupled model simulations helps separate the model’s response to forcing from its simulated internal variability.
Detection does not imply attribution of the detected change to the assumed cause. ‘Attribution’ of causes of climate change is the process of establishing the most likely causes for the detected change with some defined level of confidence . As noted in the SAR and the TAR, unequivocal attribution would require controlled experimentation with the climate system. Since that is not possible, in practice attribution of anthropogenic climate change is understood to mean demonstration that a detected change is ‘consistent with the estimated responses to the given combination of anthropogenic and natural forcing’ and ‘not consistent with alternative, physically plausible explanations of recent climate change that exclude important elements of the given combination of forcings’.
Both detection and attribution require knowledge of the internal climate variability on the time scales considered, usually decades or longer. The residual variability that remains in instrumental observations after the estimated effects of external forcing have been removed is sometimes used to estimate internal variability. However, these estimates are uncertain because the instrumental record is too short to give a well-constrained estimate of internal variability, and because of uncertainties in the forcings and the estimated responses. Thus, internal climate variability is usually estimated from long control simulations from coupled climate models. Subsequently, an assessment is usually made of the consistency between the residual variability referred to above and the model-based estimates of internal variability; analyses that yield implausibly large residuals are not considered credible (for example, this might happen if an important forcing is missing, or if the internal variability from the model is too small). Confidence is further increased by systematic intercomparison of the ability of models to simulate the various modes of observed variability, by comparisons between variability in observations and climate model data and by comparisons between proxy reconstructions and climate simulations of the last millennium.
Model and forcing uncertainties are important considerations in attribution research. Detection and attribution results based on several models or several forcing histories do provide information on the effects of model and forcing uncertainty. Such studies suggest that while model uncertainty is important, key results, such as attribution of a human influence on temperature change during the latter half of the 20th century, are robust.
The approaches used in detection and attribution research described above cannot fully account for all uncertainties, and thus ultimately expert judgement is required to give a calibrated assessment of whether a specific cause is responsible for a given climate change. The assessment approach used in this chapter is to consider results from multiple studies using a variety of observational data sets, models, forcings and analysis techniques. The assessment based on these results typically takes into account the number of studies, the extent to which there is consensus among studies on the significance of detection results, the extent to which there is consensus on the consistency between the observed change and the change expected from forcing, the degree of consistency with other types of evidence, the extent to which known uncertainties are accounted for in and between studies, and whether there might be other physically plausible explanations for the given climate change. Having determined a particular likelihood assessment, this was then further downweighted to take into account any remaining uncertainties, such as, for example, structural uncertainties or a limited exploration of possible forcing histories of uncertain forcings. The overall assessment also considers whether several independent lines of evidence strengthen a result.
In summary, the IPCC’s detection and attribution arguments do not seem to account for the substantial underestimation by climate models of natural internal variability on time scales of the main multi-decadal variability of the AMO and PDO, this general result was known at the time of AR4, see figure 9.7. I have discussed this issue at length in previous posts: Overconfidence in IPCC’s detection and attribution: Part I, II, III, IV.
Further, the flat blade of the hockey stick, which arguably under represents variability on multi-decadeal time scales, is used to lend credence to the climate model simulations.
Von Storch on climate models and natural internal variability
Hans von Storch et al. have a recent paper Can climate models explain the recent stagnation in global warming? From the abstract:
In recent years, the increase in near-surface global annual mean temperatures has emerged as considerably smaller than many had expected. We investigate whether this can be explained by contemporary climate change scenarios. In contrast to earlier analyses for a ten-year period that indicated consistency between models and observations at the 5% confidence level, we find that the continued warming stagnation over fifteen years, from 1998 -2012, is no longer consistent with model projections even at the 2% confidence level. Of the possible causes of the inconsistency, the underestimation of internal natural climate variability on decadal time scales is a plausible candidate, but the influence of unaccounted external forcing factors or an overestimation of the model sensitivity to elevated greenhouse gas concentrations cannot be ruled out. The first cause would have little impact of the expectations of longer term anthropogenic climate change, but the second and particularly the third would.
From the main text:
What do these inconsistencies imply for the utility of climate projections of anthropogenic climate change? Three possible explanations of the inconsistencies can be suggested: 1) the models underestimate the internal natural climate variability; 2) the climate models fail to include important external forcing processes in addition to anthropogenic forcing, or 3) the climate model sensitivities to external anthropogenic forcing is too high,.
The first explanation is simple and plausible. Natural climate variability is an inevitable consequence of a slow system (climate) interacting with a fast system (weather) (10). The forcing of the slow system by the (white noise) low-frequency components of the fast system produces a “Brownian motion” of the slow system, represented by a red variance spectrum – in qualitative agreement with observations. However, the details of the response depend strongly on the internal dynamics of the slow system in the time scale range of interest – in the present case, on decadal time scales. It is long known, from successive reports of the Intergovernmental Panel on Climate Change(4), that contemporary global climate models have only limited success in simulating many such processes, ranging from the variability of the ocean circulation, ENSO events, various coupled ocean-atmosphere oscillation regimes, to changes in sea ice, land surface, atmospheric chemistry and the biosphere. The inability to simulate the statistical internal climate variability may have been artificially compensated in the past by tuning the models to prescribed external forcings, such as volcanic eruptions and tropospheric aerosols.
This would explain why simulations with historical forcing by different GCMs tend to be very similar and follow closely the observed record. This artificial “inflation” of forced variability at the expense of unpredictable natural variability works, however, only in the period of tuning, and no longer in the post-tuning phase since about 2000. The net effect of such a procedure is an underestimation of natural variability and an overestimation of the response to forced variability.
Fitness for purpose ?
While some in the blogosphere are arguing that the recent pause or stagnation is coming close to ‘falsifying’ the climate models, this is an incorrect interpretion of these results. The issue is the fitness-for-purpose of the climate models for climate change detection and attribution on decadal to multi-decadal timescales. In view of the climate model underestimation of natural internal variability on multi-decadal time scales and failure to simulate the recent 15+ years ‘pause’, the issue of fitness for purpose of climate models for detection and attribution on these time scales should be seriously questioned. And these deficiencies should be included in the ‘expert judgment’ on the confidence levels associated with the IPCC’s statements on attribution.