Overall Results - Climate Change: Identification and Projections - page 91

8
Overall Results
8.1. Preliminary comments
We will now combine the four temperature reconstructions (in order:
Moberg, Ljungqvist, Loehle, Jones-Mann, all spliced with HadCRUT4),
with the four irradiance reconstructions (in order: Usoskin-lean0, Usoskin-
timv15, Be10-lean0, Be10-timv15).
Some results in this chapter will pose interpretation problems which
should be anticipated. Indeed, it may be of some surprise that some
confidence regions and intervals show an empty intersection, both between
one another and with the IPCC assumptions. As such, it will be noted that
the identified sensitivities to solar irradiance are mutually incompatible,
depending on whether the reconstructions are based on lean0 or tim15. More
specifically, the regions of confidence from Jones & Mann will be
systematically remote or disconnected from all the others. Even though these
are confidence regions of only 90%, and not 100% certainties, we are led to
the conclusion that certain data, hypotheses or conclusions are incompatible,
without being able to ascertain which are correct and which are not.
Let us look back here at the notion of confidence regions. In
identification, they are traditionally determined using the input/output data
itself, and not deduced from a priori statistical characteristics, like variance
in measurement errors. Indeed, these characteristics themselves are
identified, in this case, through the function of autocovariance of the output
error. This is why we have systematically ignored uncertainty indicators
which often accompany the collected data.