Partial Results - Climate Change: Identification and Projections - page 78

Immediately, we continue with the examination of the output simulated
by the identified model, and the influence of various inputs on this output. In
each graph in Figure 7.2, the grey line indicates the historic temperature
observed. Every line is lightly smoothed for better legibility.

It can be noted that the influence of solar activity on recent warming is
greater than that of CO 2 . Volcanic activity plays a role which is marginal,
but not negligible.

Output error is high. Deviations from one extreme to the other
reach 0.5 ° C. These come firstly from natural random fluctuations of
temperature, which cannot be explained by input variables (CO 2 , TSI
and volcanism). They also arise from uncertainties on measurements
of paleoclimatic data, in both input and output. With identification,
it is impossible to specify which explanation has the greater effect,
but the examination of the period 1850 to now, where we have more
reliable data, indicates that the influence of internal climatic variations is
essential.

Even by making the model more complex and giving it more degrees of
freedom, the error would not significantly change: the inputs do not have
components which are correlated enough with the output over the whole
duration of the identification. However, if we greatly restricted the period of
identification, for instance to 1975 to 2000 (Figure 3.6), the output error
over this period would be reduced to annual fluctuations, rather than
reaching the shown by Figure 7.2 over this period. In order to

achieve this result, the identification algorithm would apply its six degrees of
freedom by taking full advantage of the correlation observed during this
period, between the rise in CO 2 and global temperature. The entirety of the
warming observed during the second half of the 20 th Century could then be
attributed to the single anthropogenic factor. In reality, a period of 25
years is much too short to be sure that the correlation observed is not
caused by random climatic fluctuations: this would be a blatant example of
“cherry picking” performed solely to support the argument of a given
hypothesis. To a lesser extent, this could also be applied to simulations
limited to the period 1850 to now, which avoids going back to the LIA and
the MWP.

±° 0.2 C