Identification Method - Climate Change: Identification and Projections - page 68

to expect that numerous attempts have already been made since the
appearance of paleoclimatic reconstructions. In fact, the very first EBMs
were not carried out for the purposes of identification, but rather to perform
abstract theoretical analysis. They very quickly became mini-models of
general circulation, with large numbers of parameters and therefore
unidentifiable. For instance, when Crowley [CRO 00] compares the
MAGGIC model to paleoclimatic data, he adopts the parameters assumed by
the IPCC. Even though some parameters are subject to variations designed to
study their effect on the models' behavior, this is not true identification.
However, there is one case where “official” techniques of identification
(Ljung, 1999) are applied to EBMs, but in a context of simplifying complex
models by identification [WEL 14]. This is carried out, not by using real
historical data, but with data obtained by simulation of GCMs. The goal is
therefore to obtain more flexible models, whose global behavior is identical
to that seen in GCMs, and also to extract from them values for global climate
parameters; the first being the fundamental parameter: climate sensitivity.
And yet, why not carry out identification using authentic historical data?
Firstly, because they are less rich in terms of information compared to
simulated signals on which one can “experiment” repeated variations for
forcing input, and also possibly because the results do not comply with the
intents: they do not clearly confirm the anthropogenic influence, or they
show the predominant impact of solar activity. In summary, the AR5 makes
no reference to any identification of a climate model, whether black box or
grey type, neither in Chapter 9, regarding the evaluation of models (EBM or
GCM) nor elsewhere.
Nevertheless, we are taking a leap into the unknown. The model - or
class of models - that we propose to identify is intended to be completely
free of any a priori knowledge, except the general structure of the model
(Chapter 4). Of course, we will make the connection between the parameters
of the model and the climatic parameters familiar to all climatologists. We
will also occasionally make the link between some features of the model and
general behavior of thermal processes, or even to propose related
explanations, for instance, for thermohaline circulation. However, in no case
will the identified models rely on the accuracy of our knowledge of climatic
processes, which is an area in which we do not claim to have any expertise.