82
the approach of FR2011 to the derivation of AAWR using both the older version of the
CRU ΔT used in their study and the more recent version used in our analysis, and
arrive at 0.166 °C/decade for the older version and 0.183 for the latest version.
The difficulty in the approach used by FR2011 is that their value of AAWR is
based upon analysis of a residual found upon removal of all of the natural processes
thought to influence ΔT. If an unaccounted for natural processes happens to influ-
ence ΔT over the period of time upon consideration, such as the Atlantic Meridional
Overturning Circulation, then the value of AAWR found by examination of the
residual will be biased by the magnitude of the variation in ΔT due to this process
over the period of time under consideration.
Quantitative analysis of the CRU data record reveals the cause of the difference
of these two apparently disparate estimates of AAWR for the 1979–2010 time
period. The fifth rung of the Fig. 2.5 ladder plot indicates AMOC may have contrib-
uted 0.043 °C/decade to the rise of ΔT, over the time period 1979–2010. Upon use
in our EM-GC framework of the same version of CRU ΔT that was analyzed by
FR2011, we compute AAWR = 0.109 °C/decade and a slope of 0.058 °C/decade for
the contribution of AMOC to ΔT over 1979–2010. Thus, natural variation of cli-
mate due to variations in the strength of the Atlantic Meridional Overturning
Circulation accounts, nearly exactly, for the difference between the FR2011 esti-
mate of AAWR (0.170 °C/decade) and our value (0.109 °C/decade).^28
There is considerable debate about whether North Atlantic SST truly provides a
proxy for variations in the strength of AMOC. An independent analysis conducted
using different methodology (DelSole et al. 2011 ) supports our view that internal
climate variability contributed significantly to the relative warmth of latter part of
the time series examined by FR2011. Analysis of a residual to quantify a process,
rather than construction and application of a model that physically represents the
process, violates fundamental principles of separation of signal from noise (Silver
2012 ). The estimates of AAWR shown in Figs. 2.4 and 2.5 yield similar values,
0.111 °C/decade versus 0.109 °C/decade, whether or not AMOC is considered,
because our determination of AAWR is built upon a physical model for the human
influence on climate (Eq. 2.4) and does not rely on analysis of a residual.
If there is one word that best summarizes the present state of climate science in
the published literature, it might be confusion. Alas, the argument put forth in the
prior paragraphs, that a value for AAWR from 1979 to 2010 of ~0.10 °C/decade is
inferred from the climate record whether or not variations in the strength of AMOC
are considered in the model framework, is in direct contradiction to Zhou and Tung
( 2013 ) (hereafter ZT2013). ZT2013 examined version 4 of the CRU ΔT data record,
using a modified residual method,^29 and concluded AAWR is 0.169 °C/decade if
temporal variation of AMOC is ignored, but drops to 0.07 °C/decade if variations in
(^28) That is, 0.109 + 0.058 °C/decade is nearly equal to 0.170 °C/decade.
(^29) The method used by ZT13 is similar to that of FR2011, except ZT13 include a model for ΔTHUMAN
in their calculation of regression coefficients that are used to remove the influence of ENSO, vol-
canic, and solar variations from ΔT (their case 1) or remove the influence of ENSO, volcanic, solar
variations, and AMOC from ΔT (their case 2). For both cases, their model of ΔTHUMAN is a linear
function from 1860 to 2010.
2 Forecasting Global Warming