Competitive examination for admission to two 24-month research grants:
"Statistical models for remote sensing of air quality"
Bergamo University, Department "Ingegneria dell'informazione e metodi
matematici"
Tutor: Prof. Alessandro Fassò
Deadline for application: November 17, 2010.
Oral exam: November 19, 2010 (skype teleconference allowed)
For more information contact: [log in to unmask]
***************************************************************
*General objectives of the program:*
The objective is to propose a statistical model for the dynamical
spatio-temporal
mapping of fine particles capable to monitor atmospheric pollution both
in the
Lombardia region and in Europe. The integration of data from the
monitoring network
and satellite data will provide a uniform spatial representation which
will be
independent of policy of investment in monitoring networks of the
different nations
and regions in Europe.
It will also streamline the existing monitoring networks, with
significant savings at the
regional level. From a methodological point of view, the innovative
aspect is the
estimation of a parametric or semi-parametric model estimation by an
extension of
the EM algorithm.
The several issues, that will be jointly addressed, refer to the missing
spatial data,
different spatial scales, data defined on different grids and large
arrays of data.
The project involves collaboration with the Dept. of Statistics,
Viadrina University,
Frankfurt Oder (DE), il Dept. of Statistics, University of Glasgow (UK),
il Dept. of
Mathematics, University Jaime I, Castillon (S) e il Dept. of
Mathematics, Universidade
Técnica de Lisboa (PT).
*Requirements*:
PhD or a degree attained under the laws previously in force for the
reform of the
autonomy university education or degree in Statistics (Class 92 / S) or
Mathematics
(Class 45/S) or degree Master equivalent, or equivalent degree
qualification obtained
abroad.
*Matters on which the candidates' qualifications should address*:
Geostatistics, spatial-temporal models, hierarchical models, estimates
for complex
models, computational statistics, and compute-intensive.
*Research Program:*
Researchers will contribute to the main team objective, namely the
construction of
the calibration model space-time satellite data which is divided into
several stages.
The first is methodological and theoretical, will follow a significant
part consisting of
encoding and implementation in a distributed computing environment.
COSP: Preparing the module for managing data on different scales. We
compare the
Bayesian approach and one based on the likelihood.
Heterotopic Coregionalization: Extension of the algorithm of Zhang
(2007) for
completely heterotopic and autocorrelated data.
Module for missing data: Preparing the module for the management of
missing data.
This will form an integral part of the model. The missing data are not
estimated by
interpolation techniques, but they are an output of the same model.
EM Algorithm: The optimization of the likelihood for the valuation of
maximum
likelihood estimates can be a problem because of the dimensionality of
the parameter
vector. The EM algorithm is generalized here using an approximation that
will handle
the case said "completely heterotopic and autocorrelated".
Large covariance matrices: The use of tapering and reverse techniques of
covariance
matrices of large size will require a theoretical study on the
asymptotic behavior of
estimators derived from it.
Lombardia Region data: The collection of data on the Lombardia Region,
includes:
AOT, PM data, weather, altitude, land classification. This phase is
shortened by
knowledge of the regional sources. Ready georeferenced database for
space-time
analysis.
Lombardia Region Model: Identification, estimation and validation through
crossvalidation and bootstrap techniques. This phase also includes the
systematic
quantification of the uncertainty of the model
European data: Collection of data relating to Central or Atlantic.
Includes: AOT, PM
data, weather, altitude, land classification. Much more expensive than
RL for the
multiplicity of sources. Ready georeferenced database for space-time
analysis.
EU model: Identification, estimation and validation through
crossvalidazione and
bootstrap techniques. This phase also includes the systematic
quantification of the
uncertainty of the model.
*Selection:*
Selection is based on curriculum and publications assessment and an oral
exam. The
oral exam will take place on 19 November 2010 at 14:00 at the Department
"Ingegneria dell'Informazione e Metodi Matematici", Via Marconi 5,
Dalmine BG.
Applicants with domicile outside of the Lombardia Region can be remotely
interviewed via skype with webcam capability. It is the responsibility
of the applicant
to express his/her intention to do this by contacting by email the
tutor, Prof.
Alessandro Fassò, [log in to unmask], at least 7 days before the
interview.
*Subject matter of the interview:*
The interview will focus on the issues of the project.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|