Inferring instantaneous, multivariate and nonlinear sensitivities for the analysis
of feedback processes in a dynamical system: Lorenz model case-study
By FILIPE AIRES1;2¤
and WILLIAM B. ROSSOW3
1Columbia University, NASA Goddard Institute for Space Studies, USA
2Laboratoire de M´ et´ eorologie Dynamique du CNRS, France
3NASA Goddard Institute for Space Studies, USA
(Received 17 October 2001; revised 17 May 2002)

3. A NONLINEAR REGRESSION SCHEME FOR ESTIMATION OF SENSITIVITIES

To estimate the sensitivities of the dynamicalmodel in Eq. (1), we use a multivariate
nonlinear regression ? t to the statistics produced by observing the behaviour of the
system over a time period long enough to provide a good sample of the different states of
the system. Anymultivariate nonlinear regression technique, such as spline interpolation
or ARMAX¤ models, etc., could be used. For this purpose, we use a neural-network
(NN) technique because of its ability to process large-dimension datasets (which will
be helpful for further experiments on numerical models) and its capacity to integrate
a priori information about the problem (Aires 1999). This technique has been used
extensively to estimate physical relationships such as inverse radiative-transfer models
(see, for example, Aires et al. 2002b,c,d).

Falafulu Fisi Says:
September 28th, 2009 at 8:45 am

Steven said…
That’s either a mistake or a bare faced lie.

You call me a liar? Look Mr. Engineer, I read deep scientific peer review research papers in climate modeling frequently, because I can understand them. Second, I implement more advanced numerical models (published in other research journals that are not main domain of climate science ), that performed better than existing ones used by climate modelers of today , where those techniques are still unknown to climate scientists. It is equivalent to me talking about non-equilibrium economics which is something that is so unknown to or unfamiliar with our blog author Adolf Stroombergen. WHY? Because the topic is covered more in depth and published in issues of physics journals (eg : Physica A - Statistical Mechanics journal) rather than being published in an economic one.