A realistic study of the health effects of an airborne pollutant has to use
meteorological data of at least one continuous year in a given spatial
domain. Longer time periods may be averaged to one year or kept as they
are, but the prevalent weather phenomena in Europe do not allow to draw
detailed conclusions on computations based on smaller time periods. However,
according to a statement of the KMI (Koninklijk Meteorologisch Instituut van
België) [26], a data set consisting of the needed
observational variables and encompassing a period of one year can be made
available only for 
Belgian francs (ca. 8469 euro), and this amount
is independent of the fact that the data is needed for a scientific
noncommercial project.
Meteorological data has instead been obtained from the FASTEX archive [17] for free. The data consists of, among others, all the information necessary for the MESOPAC module (cmp. Subsection 2.1) and covers a time period from January 6, 1997, 0:00 GMT, to February 28, 1997, 18:00 GMT. While this time period is shorter than actually necessary, we feel confident that we can at least demonstrate the feasibility of the approach taken by solving an optimization problem as realistic as possible. A filter routine has been used to transform the data from standard WMO (World Meteorological Organization) format to CD-144 format. The 'present weather' code of field WW (WMO Table 020003) has been recast to CD-144 precipitation format according to Table 6.2.
Table 6.2: Transformation of WMO present weather code to CD-144
precipitation code. WMO codes missing have not been translated.
Missing data has been interpolated by a simple one-dimensional linear interpolation routine (in time). No attempt has been made to use other meteorological data in the interpolation process, although it has to be expected that the quality of the complete data set can be greatly improved in this way. A Voronoi diagram of the surface stations whose data has been used can be found in Figure 6.5, while the corresponding diagram for the upper air stations can be found in Figure 6.6. Note that not all upper air stations are within the boundaries of the computational domain, and that the data set in use is rather sparse.
Figure 6.5: Voronoi diagram of 84 surface air stations used.
Figure 6.6: Voronoi
diagram of the twelve upper air stations used. Note that three of the
stations are not located inside the boundaries of the computational grid.