seminari a Roma "La Sapienza"

[Giovanna Jona Lasinio <jona@pow2.sta.uniroma1.it>]

GRUPPO DI RICERCA GRASPA (Gruppo di Ricerca per le Applicazioni della
Statistica ai Problemi Ambientali)
SEMINARI ORGANIZZATI DALL'UNITA' LOCALE DI ROMA
(www.uniroma3.it/politiche/lagona/ambiente)
 
******************************************************************
Il giorno 20 luglio 2001 alle ore 10:30 presso il Dipartimento di
Statistica, Probabilitŕ e Statistiche Applicate (P.le Aldo Moro 5, Sala
34)

i professori HANS WACKERNAGEL (Ecole des Mines, Paris) e DAN GRIFFITH
(Syracuse University) terrano due seminari di cui si allegano titolo ed
abstract:

**********************************************************************
DATA ASSIMILATION INTO PHYSICAL MODELS: A GEOSTATISTICAL PERSPECTIVE ON
KALMAN FILTERING

Hans Wackernagel 

Operational forecasting systems are composed of an observational net- 
work providing continuously data in time, a dynamical (physical, chemical 
or ecological) model and a data assimilation scheme. The modern way is 
to sequentially assimilate data from the network into the physical model 
using a non-linear Kalman filter. Geostatistical concepts can be applied 
in this space-time context to improve the procedure. The talk will report 
about some recent work on the topic [2, 3, 1] and examples from
operational 
oceanography will be discussed. 

References 
[1] Bertino, L., Evensen, G., and Wackernagel, H. Combining geostatistics 
and Kalman filtering for data assimilation in an estuarine system. Inverse 
problems (2001), 36p. accepted. 
[2] Sénégas, J., Wackernagel, H., Rosenthal, W., and Wolf, T. Error 
covariance modeling in sequential data assimilation. Stochastic
Environmental 
Research and Risk Assessment 15 (2001), 65{86. 
[3] Wolf, T., Sénégas, J., Bertino, L., and Wackernagel, H. Applica- 
tion of data assimilation to three-dimensional hydrodynamics: the case of
the 
Odra lagoon. In geoENV III { Geostatistics for Environmental Applications 
(Amsterdam, 2001), P. Monestiez, D. Allard, and R. Froidevaux, Eds.,
Kluwer. 
1 


A SPATIAL FILTERING SPECIFICATION FOR THE AUTO-POISSON MODEL
Daniel A. Griffith
Department of Geography
Syracuse University

ABSTRACT The central role a Poisson model plays with respect to the
analysis
of counts is analogous to the position of the normal distribution in the
context of models for continuous data. Its extension to the auto-Poisson
model is to describe georeferenced data consisting of counts exhibiting
spatial dependence. The conventional specification of this auto- model is
plagued by being restricted to only situations involving negative spatial
autocorrelation, and an intractable normalizing constant. Results
circumventing these two restrictions will be presented that account for
spatial autocorrelation in the mean response specification by
incorporating
latent map pattern components. First the conceptual background for this
work
will be outlined. Then, illustrative findings will be summarized for seven
empirical datasets available in the literature. These empirical analyses
involve comparisons with the logistic and Gaussian approximations to a
Poisson model. One important finding is that the revised auto-Poisson
specification is able to employ a standard Poisson regression estimation
procedure.







                                                           

Giovanna Jona Lasinio
Dipartimento di Statistica, Probabilita' e Statistiche Applicate
Universita' di Roma "La Sapienza"
P.le Aldo Moro 5
00185 Roma

tel. Office: (39)0649910473
fax: (39)064959241
web page: http://apegate.roma1.infn.it/~gJona/giohome/