This new paper allows great headlines to proclaim that the warming “pause” in global surface temperature is explainable by climate models. As is often the case in climate reporting the details do not back up the headline.
Risbey et al (2014) in Nature Climate Change is yet another paper suggesting that the global surface temperature hiatus of the last 15-years or so is due to changes in the character of the ENSO. But they go a little further and say that once the observational timing of ENSO changes is included in climate models they do a good job. Unfortunately, whilst an interesting and though-provoking paper, it does not support its own conclusion that “climate models have provided good estimates of the 15-year trends for recent periods.”
Climate models have many uses and are essential tools to discover what is going on and, with major caveats, suggest future possibilities. It is well-known that as a whole the CIMP5 ensemble of models does not represent reality that well with only two models coming anywhere near reflecting the hiatus in global surface temperature seen in the last 15-years or so.
With a climate model ensemble that is mainly unrepresentative of reality there are several possibilities for further action. One is to have faith in the models that over longer timescales realities departure from them is temporary. Another is to select those models that best simulate reality and concentrate on them, and the other is to refine the models. Risbey et al (1014) carry out both the latter options.
They selected 18 out of 32 CIMP5 models choosing the ones that had sea surface temperature as a model output. In itself this introduces a selection effect whose influence on subsequent selections of “suitable” models is unknown. Out of those 18 they selected the four best and four worst. The best included ENSO parameters that are in phase with observations. They argue that when the phase of ENSO is got right climate models do represent reality. Unfortunately the evidence they provide for this is not convincing.
If the ENSO with La nina dominant is having the effect of flattening the global surface temperature of the past 15 years or so then the converse must also be true. ENSO with El Nino dominant would have contributed to the warming seen since about 1980. Couple this with other decadal factors such as stratospheric water vapour and the Atlantic Oscillation and it is easy to see the implication of judging most of the 1980 – 97 increase to be mostly anthropomorphic. It would be to overestimate future warming, which is precisely what the CIMP5 ensemble does show the recent flatlining.
Our lack of understand of the ENSO process also affects the stated conclusions of this paper. We cannot predict these events with any certainty and we cannot simulate them to any degree of great accuracy. So while there are ENSO components in a climate model, to say that those in the right phase do better could mean nothing. In addition there are other semi-regular changes such as the Atlantic oscillation that might, or might not, be in phase with the observations.
Supplementary information would have helped understand this paper, especially the selection of the models, but unfortunately there are none. This means that given the information in this paper alone it would not be possible to retrace the author’s footsteps.
This paper allows great headlines to be written proclaiming that the “pause” in global surface temperature is explainable by climate models. As is often the case in climate reporting the details do not back up the headline.
What this paper has really done is to draw attention to the limitations of the climate models. One can select subsets of them and argue that they are better than others but the real test is if the Risbey et al (2014) paper has predictive power. In science looking forward is always more powerful than looking back and adjusting models to fit the data.
Risbey et al (2014) say they expect the observed trend to bounce back. So do many others for different reasons. If it does how will we know who is right?