Environmental Engineering Reference
In-Depth Information
Prediction) whereas the air quality run has been nested in the results of the
CHIMERE chemical transport model, provided by the KNMI. The left panel of
Fig. 1.
shows the map for the city of Shenyang for July 2007. By artificially
increasing or decreasing the percentage urban green vegetation cover, one can
investigate its impact on air quality. The middle and right panels show artificial
vegetation cover maps by decreasing and increasing the real maps, respectively.
The effect of the vegetation change on the air pollution can be very interesting for
the local government when they define abatement measures for air pollution.
123.10
123.20
123.30
123.40
123.50
123.10
123.20
123.30
123.40
123.50
123.10
123.20
123.30
123.40
123.50
123.10
0
123.20
123.30
123.40
123.50
123.10
0
123.20
123.30
123.40
123.50
123.10
0
123.20
123.30
123.40
123.50
1
1
1
Fig. 1.
Map of fractional vegetation for the city of Shenyang: satellite-derived data for July 2007
(left panel); artificially decrease in vegetation (middle panel); and artificially increase in
vegetation abundance (right panel)
3. Results
validation gives good results with positive correlations between measurements and
modeled results in almost all stations, periods and discussed pollutants. It is also
clear that the January pollution levels are much higher than the October pollution
levels, especially for SO
2
. Looking specifically at January 2004, we can see that
the model has too high peak concentrations for both SO
2
and PM
10
. However, the
correlation and regression of the model on the measurements are quite good
(correlation
1
of SO
2
= 0.46 with the slope of the regression
1
= 0.92; correlation
1
of
PM
10
= 0.58 with the slope of the regression
1
= 1.21). This shows that our model is
quite capable of reproducing the winter time variability of the Shenyang pollution.
The modelled pollution is more important in stations where the measured pollu-
tion is higher, giving confidence in the spatial variability of the model.
For the October results, the same conclusions as for January can be made for
PM
10
, although the correlation is slightly lower (correlation
1
of PM
10
= 0.52 with
the slope of the regression
1
= 0.67) and the spatial spread in the concentrations is
strongly underestimated in the model. This underestimation can also be found in
January but is less clear there. The October SO
2
concentrations are much less well
represented in the model than the previous discussed pollutants and months
(correlation
1
of SO
2
= 0.20 with the slope of the regression
1
= 0.40). This is pro-
Using measurements and model results averaged over all the available stations.
1
