Why climate models are failing - Climate Change: The Facts - page 18

While noting that the sensitivity is far from certain, Annan and Hargreaves used
probabilistic estimates themselves rather than a 'uniform prior' subject to perturbation by
multiple GCM runs, noting that the standard methodology 'has unacceptable properties'
that condition it to produce unrealistically high probabilities for large warming. 15 By
using the more realistic approach, the 95 per cent confidence limit for a large warming
dropped by one-third, and they noted that '[t]hese results also impact strongly on projected
economic losses due to climate change.' 16 Incorporating a prior probability distribution
based on expert opinion (rather than the oft-used and ill-founded uniform distribution) into
aBayesiananalysis,AnnanandHargreavesfindamedianvaluefortheequilibriumclimate
sensitivity of 1.9°C. Incorporating a Cauchy distribution for the Bayesian prior results in a
median climate sensitivity estimate of about 2.2°C.
Van Hatteren found, using a retrospective approach, a climate sensitivity of 2.0 +/-
0.3°C, and noted 'it is at the lower end of the range considered likely [by the IPCC].' 17
In this they were referring to the 2007 IPCC 'Fourth Assessment' report. The 2013 report
reduces the lower limit from 2.0 to 1.5°C, with no 'best estimate'. This will provide the
IPCC with some cover when the low sensitivity becomes obvious even to those currently
defending the established, failing paradigm.
One of the most important papers in this tranche is that of Ring et al., 18 if only because
the fourth author, Michael Schlesinger of the University of Illinois, has long been an
extremelyvocaladvocateforahigh-endsensitivity.Byadjustingtheirmodelwithobserved
temperatures(whichincludesmuchof'thepause'),theyarriveatasensitivityof1.5-2.0°C.
While they claim that this is 'on the low end of the estimates in the [2007] IPCC's Fourth
Assessment Report ' it is in fact clearly beneath that 19 —yet another example of Kuhn's
observation that what turns out to be obvious is initially ignored.
Hargreaves et al. used a new determination of the cooling during the Last Glacial
Maximum to derive estimates of the climate sensitivity. 20 Their estimates use two different
statistical techniques, one employing regression relationships between tropical
temperatures during the Last Glacial Maximum and climate model climate sensitivity, and
another using a Bayesian approach weighting each climate model based on how well it
matches the new Last Glacial Maximum data. The two methods produced very similar
results, with a mean equilibrium climate sensitivity of about 2.5°C with a 90 per cent
confidence range of about 1°C to 4°C.
Aldrin et al. also fit model results to observed temperatures but in addition included
observed changes in oceanic heat content and found a mean sensitivity of 2.0°C, which is
over 40 per cent below the mean of the model family used in the 2013 IPCC report. 21
Spencer and Braswell used a simple climate model to simulate the global temperature
variations averaged over the top 2000 meters of the global ocean during the period