The War of the Graphs - Climate Change: Identification and Projections - page 38

The second comes at the high level of random fluctuations, which can be
seen in all the figures above. Observations which are limited to several
decades - or even to one or two centuries - are not long enough to be able to
use them for drawing definitive conclusions.
The third is that these overlays, regressions and correlations are only
significant if the global climate machine is a process with no memory, where
output depends only on the input at that precise moment, without taking past
events into account. This is far from being the case, due to thermal inertia
caused by the oceans. Therefore, even with the most sophisticated statistical
techniques, the conclusions will be inconsistent, regardless of the theory
being defended.
The only way of tackling the issue is by using a dynamical causal model.
Indeed, correlations need to be established, not only between input and
output at a given moment, but also between output and all or part of past
input data. The technique for determining - explicitly or implicitly - such
correlations is none other than that used for the identification of dynamical
systems. Only this is able to establish whether, for example, the lines in
Figure 3.5 contradict the theory of solar activity contribution or not. The
signals still need to contain the necessary information, requiring that their
duration be sufficiently long with regard to their spectrum; at least a
millennium. It is also necessary to take into account all potential causes
simultaneously, which multivariate identification automatically ensures.
3.3. Usable data
Despite its poor demonstration capability, Figure 3.4 may lead us to the
worrying conclusion that the results of identification could differ greatly
depending on the data used, and the battle of the graphs would give rise to
questions over the overall validity of paleoclimatic data. It is not a question
of challenging anybody's scientific rigor; when a fact appears to be perfectly
established, it is admissible, in a pedagogical context, to select the
observations which illustrate the argument in the clearest way, and even to
emphasize the point. However, this is not justifiable when scientific fact is
not yet established.
Finally, what credit should be given to the reconstructions available? In
the AR5 (Figure 5.7), the IPCC presents the “plates of spaghetti”, which are