Introduction - Climate Change: Identification and Projections - page 4

climate response function can simulate global average temperature trends of
the modern era and the past millennium” by Van Hateren [VAN 13].
However, nothing in this title refers to identification. The key word which
gets the closest is “modeling”, and none of the bibliographic references
given refer to the great Masters of identification theory (Aström, Ljung,
Soderström, etc.). It is quite possible that the author is unknowingly applying
identification, just as Jourdain used prose. With the exception of the
excellent paper mentioned above, we couldn't find any other significant
work on the global modeling of climatic process through identification.
Indeed, the IPCC has long been checking its models against the historical
climate data available: its large numerical models, based on the laws of
physics, as well as its simplified models, based on energy balances. Yet
identification is not involved. At most, these models involve partial
adjustments (closure parameters tuning) or fingerprinting (detection and
attribution of anthropogenic impact).
According to Hervé le Treut (2004), Director of the Institut Pierre Simon
Laplace 6 : “numerical models (i.e. large-scale physical models, simulated by
digital calculators) play a key role in studies of the greenhouse effect
because they are the only tool which can be used to evaluate future climates:
the analogy with climates of past eras which experienced different CO 2
levels and the extrapolation toward the future of climatic data collected
during the 20 th Century provide unarguably precious information, but can
only be interpreted with the help of physical models”.
In this work, we look instead to push forward with the logic of
identification, allowing the climatic data to speak for itself, using it as “black
box” input and output data (causes and effects), without constraining it to
any type of prior knowledge. This is not without its difficulties: the Earth's
climatic process is at the limit of what can be identified. To achieve this
goal, identification requires that input data be sufficiently accurate, with a
suitable number of significant events. In this case, the effects caused by the
input are partly obscured by the random fluctuations of the climate.
Regarding CO 2 , the first significant changes go back to less than a century
ago, and their effects are difficult to distinguish from natural variations, both
having the same order of magnitude, in terms of size and duration.
Furthermore, to observe relatively large-scale temperature variations, it is
6 Participant institute in the CMIP 5.