Image Processing Reference
In-Depth Information
•
Its proximity to emission sources
•
Morphology of the buildings around the area, impinging on air flow
•
Climatic and meteorological conditions of the area under study
Each element of the town, including the measuring stations, is identified by an ID
card. Then, the relationships existing between the ID card and the pollutant concen-
tration at each measuring station can be studied. The ID card is built from three
sources of data: the measuring stations, the geographical database and remotely
sensed data. Combined exploitation of this dataset allows the study of the morpho-
logical configuration of the city and the characterization of the measuring stations
(Ung et al.
2002a, b
; Weber et al.
2002
).
The sitting of measuring stations is done according to objectives of air pollution
control, their neighborhoods, population density, and sources of air pollution.
Due to the cost of a measuring station, the agencies in charge of air quality control
have a restricted number of stations. The ID cards can be used to detect areas
of the city similar to those surrounding the measuring stations. These areas are
called the pseudo-stations; their attributes are similar to those of the measuring
stations. The similitude is defined from the ID cards. At this stage, a hypothesis
can be suggested regarding the possibility to model the air pollution concentrations
for these areas through a combination of satellite images according to a relationship
between these images and the measuring stations. If this hypothesis is valid, then
it is possible to obtain a densification of the measuring network with virtual
stations. The estimation of air pollutant concentrations is based on a law linking
satellite measures with atmospheric transmittance and then to air pollutant concen-
tration measured by actual stations (Wald and Baleynaud
1999
; Retalis et al.
1999
;
Sifakis et al.
1992, 1998
; Finzi and Lechi
1991
; Basly
2000
). The establishment of
such a law is only possible for a restricted set of pseudo-stations; a virtual station
is such a pseudo-station.
Once the virtual stations are generated, a surface interpolation can be applied
based on both the actual and virtual stations. This produces a surface map of the
concentration for each pollutant. Actually, the fusion process is fully achieved by
imposing some constraints on the interpolation process. The result should repro-
duce what is obtained by the numerical models of airflow at a resolution of 10 km
and, of course, it should reproduce what has been measured at the actual stations
and assessed at the virtual stations. There are several techniques for fusing gridded
data and point measurements with constraints at both ends of the multiresolution
pyramid. Here, below we show an example in which we adopted that of Beyer et al.
(
1997
) and Lefèvre et al. (
2004
).
11.5.3
Illustrated Example
The area of interest is the urban community of Strasbourg, in the Northeast of
France, close to Germany. The data available are concentrations of pollutants provided
