Just how reliable is the IPCC AR5 advice to policy makers?
The recent IPCC report stated that climate scientists are 95-100% certain that the observed temperature rise since 1850 is anthropogenic. The headline attribution statement in Chapter 10 was
“It is extremely likely that human activities caused more than half of the observed increase in GMST from 1951 to 2010.The best estimate of the human-induced contribution to warming is similar to the observed warming over this period.”
World political leaders are basing policy on the validity of this statement, which is entirely based on comparing CMIP5 models to global surface temperature data. These ‘fingerprinting’ studies are described in chapter 10, which I find all but impossible to comprehend. The underlying assumption in AR5 is that natural climate variability has essentially played no role in warming since 1950. However is this actually true ?
Figure 1 shows a comparison between the ensemble of CMIP5 models and observations.
Agreement until 2000 would appear to be reasonably good. However we then read in Chapter 9 Box 9.1 that in reality:
Model tuning aims to match observed climate system behaviour and so is connected to judgements as to what constitutes a skilful representation of the Earth’s climate. For instance, maintaining the global mean top of the atmosphere (TOA) energy balance in a simulation of pre-industrial climate is essential to prevent the climate system from drifting to an unrealistic state. The models used in this report almost universally contain adjustments to parameters in their treatment of clouds to fulfil this important constraint of the climate system.
So the models are tuned so as to describe past observations. Furthermore periods of cooling are explained by volcanoes which are simulated by something called ‘EMICS’ – Earth System Models of Intermediate Complexity up until 2005. Please don’t ask me what EMICS are but there is also no doubt in my mind that these are also tuned so that aerosols can then match the global temperature response to volcanic erruptions such as Pinatubu!
Science of Doom has written a detailed analysis of AR5 attribution studies and even he is not convinced. He writes:
Chapter 10 of the IPCC report fails to highlight the important assumptions in the attribution studies. Chapter 9 of the IPCC report has a section on centennial/millennial natural variability with a “high confidence” conclusion that comes with little evidence and appears to be based on a cursory comparison of the spectral results of the last 1,000 years proxy results with the CMIP5 modeling studies.
He proposes an alternative summary for Chapter 10 of AR5:
It is extremely likely [95–100%] that human activities caused more than half of the observed increase in GMST from 1951 to 2010, but this assessment is subject to considerable uncertainties.
The pause in warming since 1998 is clearly evident in Figure 1 above. This pause undermines the statement that natural variability is unimportant. Whichever way you look at it, the lack of warming for the last 18 years must be natural whatever the final explanation may be. So how well do the models actually perform at simulating such natural varaibility. This from Knudson et al. ( again thanks to SoD).
The model control runs exhibit long-term drifts. The magnitudes of these drifts tend to be larger in the CMIP3 control runs than in the CMIP5 control runs, although there are exceptions. We assume that these drifts are due to the models not being in equilibrium with the control run forcing, and we remove the drifts by a linear trend analysis (depicted by the orange straight lines in Fig. 1). In some CMIP3 cases, the drift initially proceeds at one rate, but then the trend becomes smaller for the remainder of the run. We approximate the drift in these cases by two separate linear trend segments, which are identified in the figure by the short vertical orange line segments. These long-term drift trends are removed to produce the drift corrected series.
In other words, any ‘natural’ trends generated by models in the temperature data are assumed to be an artifact and simply removed.
Five of the 24 CMIP3 models, identified by “(-)” in Fig. 1, were not used, or practically not used, beyond Fig. 1 in our analysis. For instance, the IAP_fgoals1.0.g model has a strong discontinuity near year 200 of the control run. We judge this as likely an artifact due to some problem with the model simulation, and we therefore chose to exclude this model from further analysis
This certainly does not sound to me that the CMIP3(5) models are correctly simulating natural variability. Many models have drifts which are assumed to be anomalous and need to be corrected into zero dependency. Then any resulting model structure is discarded as being an artifact. However the observed cooling from 1940-1970 and the pause in warming post 1998 are certainly not artifacts, so why reject those models that produce something similar? The drifts and structure of these models are of the same order as observed warming. Do models correctly simulate the actual climate or do they just simulate the anthropogenic component? The AR5 attribution study depends critically on them being able to simulate both anthtopogenic forcing and natural variability. This is clearly evident from Fig 10.5 in chapter 10.
Now suppose there actually is a natural 60 year oceanic heat oscillation present – AMOC as described by Chen and Tung. This would necessarily imply that there has been a natural warming component of magnitude of about 0.2C since 1950. This is also clearly visible in a simple fit to the Hadcrut4 data including such a term.
In this case the (Internal Variability) term in Figure 10.5 turns out to be non-zero and the result now looks rather different. The models now run way too warm as shown below.
The anthropogenic term ANT is now no longer in agreement with the observations and the attribtion statement should be modified to account for the non-zero natural component. As a consequence it becomes clear that CMIP5 models are running about 50% too hot. Therefore the attributrion statement should now probably read instead.
“It is very likely that human activities caused about half of the observed increase in GMST from 1951 to 2010. The best estimate of the human-induced contribution to warming is 55% of the observed warming over this period.”
This could also explain discrepancies in the spatial distribution of warming. In general models and data agree fairly well that most warming occurs over land and arctic regions. However van Oldenborgh et al. show there is significant disagreement mostly over the oceans, overestimating warming in the northern Pacific and underestimating warming in the southern tropical oceans. Figures 4 and 5 show simple CMIP5 modle comparisons for the two hemispheres.
The southern hemisphere is dominated by oceans and warms slower. Overall the spatial agreement on a grid spacing level is fairly good except for the last 15 years. However there are discrepencies both in overestimating warming in about one quarter of the world and underestimating it in about one third.
WG1 Box 9.2 AR5 admits that “Hiatus periods of 10/15 years can arise as a manifestation of internal decadel climate variability, which sometimes enhances and sometimes counteracts the long term forcing trend” followed by a long explanation as to why this does not affect ever increasing confidence in climate models.
I am not convinced – nor do I suspect are some of the authors of AR5. One can only imagine the intense pressure that has been placed on them to increase the certainty of human warming, thereby maintaining the various green/governmental interests and funding lobbies for another 4 years.
I wonder who will eventually jump first ?