By Christopher Monckton of Brenchley
The IPCC’s forthcoming Fifth Assessment Report continues to suggest that the Earth will warm rapidly in the 21st century. How far are its projections short of observed reality?
A monthly benchmark graph, circulated widely to the news media, will help to dispel the costly notion that the world continues to warm at a rapid and dangerous rate.
The objective is to compare the IPCC’s projections with observed temperature changes at a glance.
The IPCC’s interval of temperature projections from 2005 is taken from the spaghetti-graph in AR5, which was based on 34 models running four anthropogenic-forcing scenarios.
Curiously, the back-projections for the training period from 2005-2013 are not centered either side of the observational record (shown in black): they are substantially above outturn. Nevertheless, I have followed the IPCC, adopting the approximate upper and lower bounds of its spaghetti-graph.
The 34 models’ central projection (in yellow below) is that warming from 2005-2050 should occur at a rate equivalent to approximately 2.3 Cº/century. This is below the IPCC’s long-established 3 Cº centennial prediction because the models expect warming to accelerate after 2050. The IPCC’s upper-bound and lower-bound projections are equivalent to 1.1 and 3.6 Cº/century respectively.
The temperature scale at left is zeroed to the observed temperature anomaly for January 2005. Offsets from this point determine the slopes of the models’ projections.
Here is the outturn graph. The IPCC’s projections are shown in pale blue.
The monthly global mean UAH observed lower-troposphere temperature anomalies (vortex.nsstc.uah.edu/data/msu/t2lt/uahncdc.lt) are plotted from the the beginning of the millennium in January 2001 to the latest available month (currently April 2013).
The satellite record is preferred because lower-troposphere measurements are somewhat less sensitive to urban heat-island effects than terrestrial measurements, and are very much less likely to have been tampered with.
January 2001 was chosen as a starting-point because it is sufficiently far from the Great El Niño of 1998 to prevent any distortion of the trend-line arising from the remarkable spike in global temperatures that year.
Since the 0.05 Cº measurement uncertainty even in satellite temperature anomalies is substantial, a simple least-squares linear regression trend is preferred to a higher-order polynomial fit.
The simplest test for statistical significance in the trend is adopted. Is the warming or cooling trend over the period of record greater than the measurement error in the dataset? On this basis, the zone of insignificance is shown in pink. At present the trend is at the upper bound of that zone and is thus barely significant.
The entire trend-line is beneath the interval of IPCC projections. Though this outcome is partly an artefact of the IPCC’s unorthodox training period, the slope of the linear trend, at just 0.5 Cº/century over the past 148 months, is visibly below half the slope of the IPCC’s lower-bound estimate of 1.1 Cº/century to 2050.
The principal result, shown in the panel at top left on the graph, is that the 0.5 Cº/century equivalent observed rate of warming over the past 12 years and 4 months is below a quarter of the 2.3 Cº/century rate that is the IPCC models’ current central projection of warming to 2050.
The only moment when the temperature anomaly reached the IPCC’s central estimate was at the peak of the substantial el Niño of 2010.
The RSS dataset, for which the April anomaly is not yet available, shows statistically significant cooling since January 2001 at a rate equivalent to 0.6 Cº/century.
Combining the two satellite temperature datasets by taking their arithmetic mean is legitimate, since their spatial coverage is similar. Net outturn is a statistically insignificant cooling at a rate equivalent to 0.1 Cº/century this millennium.
The discrepancy between the models’ projections and the observed outturn is startling. As the long period without statistically-significant warming (at least 17 years on all datasets; 23 years on the RSS data) continues, even another great el Niño will do little to bring the multi-decadal warming rate up to the IPCC’s least projection, which is equivalent to 1.1 Cº/ century to 2050.
Indeed, the maximum global warming rate sustained for more than a decade in the entire global instrumental record – equivalent to 1.7 Cº/century – is well below the IPCC’s mean projected warming rate of 2.3 Cº/century to 2050.
This discrepancy raises serious questions about the reliability of the models’ projections. Since theory would lead us to expect some anthropogenic warming, its absence suggests the models are undervaluing natural influences such as the Sun, whose activity is now rapidly declining following the near-Grand Maximum of 1925-1995 that peaked in 1960.
The models are also unable to predict the naturally-occurring changes in cloud cover which, according to one recent paper echoing a paper by me that was published three years ago, may have accounted for four and a half times as much warming from 1976-2001 as all other influences, including the influence of Man.
Nor can the models – or anyone else – predict el Niños more than a few months in advance. There is evidence to suggest that the ratio of el Niño to la Niña oscillations, which has declined recently, is a significant driver of medium-term temperature variation.
It is also possible that the models are inherently too sensitive to changes in radiative forcing and are taking insufficient account of the cooling effect of non-radiative transports
Furthermore, the models, in multiplying direct forcings by 3 to allow for allegedly net-positive temperature feedbacks, are relying upon an equation which, while applicable to the process engineering of electronic amplifiers for which it was designed, has no physical meaning in the real climate.
Without the Bode equation, net feedbacks may well be vanishingly different from zero, in which event the warming in response to a CO2 doubling, which is about the same as the centennial warming, will be equivalent to the IPCC’s currently-predicted minimum warming rate, equivalent to 1.1 Cº/century.
Be that as it may, as the above graph from the draft Fifth Assessment Report shows, in each of the four previous IPCC Assessment Reports the models have wildly over-projected the warming rate compared with the observed outturn, and, as the new outturn graph shows, the Fifth Assessment Report does the same.
I should be interested in readers’ reactions to the method and output. Would you like any changes to the monthly graph? And would it be worthwhile to circulate the monthly-updated graph widely to the news media as an answer to their dim question, “Why don’t you believe in global warming?”
Because there hasn’t been any to speak of this millennium, that’s why. The trouble that many of the media have taken to conceal this fact is shameful. This single, simple monthly graph, if widely circulated, will make it very much harder for them to pretend that the rate of global warming is accelerating and we are to blame, or that the “consensus” they have lazily accepted is trustworthy.
The climate scare has only lasted as long as it has because the truth that the models have failed and the world has scarcely warmed has been artfully hidden. Let it be hidden no longer.
What might the mathematical symbol be for faith, and is it treated as a function or an operator? If there is nonesuch may I suggest an addition sign modified with an exaggerated descender.
I think too there should be boundaries of incredibility beyond which plots are discontinued for being meaningless. In the noodle plot above that happened during the training period.
So – how can the ‘training period’ graphs lie 90% above the actual outturn..?
‘Rubbish in; rubbish out…’
Barry, my point is this:
The GCMs are composed of layer upon layer of assumptions about how the earth’s climate system operates, assumptions which are translated into model inputs, computational algorithms, internal data transformations, and eventually, the model run final outputs.
If we were to examine every input, every initialized parameter, every computational algorithm, and every internal data transformation, could we honestly say that we have enough observational knowledge about how the climate system actually works to believe that we have all the elements needed to accurately model the earth’s climate system, and that each element is reasonably within the bounds of what might occur in the physical operative system — individually as single elements, and collectively as an aggregate of the operative real-world physical processes?
If we haven’t yet reached critical mass in having enough of a basic understanding of the actual climate system as it works in the physical world to properly construct a computational model for research purposes, then my fundamental question remains …… what do we gain in terms of a better understanding of the real-world climate system if we spend the greatest part of our time and energy making changes to computer models in order to fit a curve — changes to inputs, process algorithms, and data transformations which may or may not have anything in common with what actually happens in the operative physical system?
Beta Blocker,
But you focus on only one part of the enedeavour. Most of the modeling in GCMs is physics, not statistics based. The equations for hydrological operation, conservation of mass and energy, radiative transfer (generalised) and many basic physical processes underpin the models, and these are not left untouched but refined as understanding of the physics – not curve fittng – grows. For some components where the physics is not well-bounded, yes the processes are parametrised by referring to obs – not trends, by the way, just the average behaviour. No processes are trained to match long-term trends. It’s not a curve-fitting exercise in that regard.
Why should there be anything wrong in trying to imrove the models in every way that is reasonable? Why spurn experimental science? Why are we so absolute about the physics in climate science, but not in the other sciences (eg medicine)? In truth, there is no model in any science that is a perfect mirror of nature. What we gain by parametrising certain phenomena in climate models is a short-cut to representing the climate system when we can’t account for every molecule.
Let me make a very important point again – the models are not trained to trends. That is a common misconception. The ‘curve-fitting’ is done to approximate average conditions, not trends over time.