Environmental Engineering Reference
In-Depth Information
The free-running model is thus already better than the statistical model in a
general sense. It could be improved further by using data-assimilation of ground-
based PM10 observations from the Dutch air quality monitoring network and
surrounding countries to improve the initial conditions. Furthermore, satellite data
from for example MODIS could be used to detect and include the contribution of
incidental sources like dust storms and forest fires, which cannot be included in
emission databases. These possibilities will be investigated in the near future.
Measured
PROPART
LOTOS-EUROS_BiasCor
70
60
50
40
30
20
10
0
Mar/1
Apr/1
May/1
Jun/1
Date
Fig. 1.
Time series of observations, the bias-corrected LOTOS-EUROS and the statistical model
PROPART for a station in the south-east of the Netherlands
Table 1.
Performance of bias-corrected LOTOS-EUROS (LE_bc) as compared with persistence,
the statistical model PROPART, LOTOS-EUROS without bias correction and LOTOS-EUROS
with an alternative bias correction (LE_altbc)
Parameter
LE_bc
Persist PROPART
LE
LE_altbc
Mean (μg/m
3
)
26.45
26.51
27.22
13.25
25.21
Bias (μg/m
3
)
−0.03
0.02
0.82
−13.23
−1.28
3
stde (μg/m )
9.06
10.99
10.50
9.60
9.47
Skillvariance
0.70
1.00
0.99
0.50
0.89
Correlation R
0.70
0.63
0.66
0.68
0.70
Residue (μg/m
3
)
6.49
7.64
7.53
13.40
6.94
rmse (μg/m
3
)
9.06
10.99
10.54
16.34
9.56
Hit rate
49.21
46.32
45.43
8.30
44.47
# Predicted 50-200 (μg/m
3
) 380
991
1047
1
725
# Observed 50-200 (μg/m
3
) 1003
983
949
1003
1003
%Correct 50-200 (μg/m
3
)
61.32
42.48
40.21
100.00
50.34
