With the data available, we decided to use a computational grid of size
. Given the sparsity and the resolution of the
population data, land use data, and meteorological data available, this
seemed to be a comfortable compromise between computational accuracy and data
accuracy. The grid spacing chosen was 9086.9m. This a compromise between
several values, since 
in north-south direction has a length of
m on the surface of the earth, but the same angle in west-east
direction has a length of 
m on 
latitude
and 
m on 
latitude.
All settings in the air dispersion model used are standard ones as
recommended in [6], i. e. the time step length is one hour, four
puffs are released per hour, twice per hour a sampling of pollutant
concentration takes place, etc. No air chemistry has been included in the case
study, and no precipitation data has been used. The emission rate of the
hypothetical pollutant described in Section 6.1.4 was set to
1000 gs
. Note that changing this value does not represent a scaling
of the objective function, due to the nonlinear Michaelis-Menten-term for
metabolism in the liver (see next section). The temperature of the exhaust gas
was set to 434 
K, while the exhaust speed was set to 50 ms
for a opening diameter of 4m. No other pollution emitter has been placed on
the computational grid.