Environmental Engineering Reference
In-Depth Information
1
180
160
0.8
140
120
0.6
100
80
0.4
60
40
0.2
20
0
0
0 0 0 0 0 100 120 140 160 180
WD (deg)
0 0 0 0 0 100 120 140 160 180
WD (deg)
120
100
80
60
40
20
0
0 0 0 0 0 100 120 140 160 180
WD (deg)
Fig. 4
Probability P, expected mean wind direction EWD (degrees) and effective distance DIS
(m) against meteorological mean wind direction WD (degrees) at the station ST1 locating 7.3 m
from the road side.
Correspondingly, the expected wind direction (EWD) is equal to the meteorological
mean wind direction (WD) for perpendicular winds, while in other cases EWD turns
away from the road direction (Fig.
4b
). As a consequence, the effective distance
(DIS) has a finite value also in the case of mean wind direction WD being parallel
to the road as seen at the station ST1 in Fig.
4c
. Variations of the computed parameters
with a fixed meteorological mean wind direction associate with the changes of
wind speed.
The statistical agreement between the measured and modeled concentrations is
best at the nearest monitoring site (ST1), but decreases with the distance of monitoring
from the road. Similarly the slope decreases with distance, indicating increasing
overestimation at more distant sites. According to our analysis linear regression
seems not to be sensitive to wind direction, but depends strongly on wind speed.
The influence of
wind
sp
eed
on
the
predictions was studied by computing relative
bias RB defined as (
pred
-
obs
) /
obs
in four groups of wind speed. The results are
presented in Table
1
, where overestimation is limited to wind speeds <2 m/s, while
stronger wind speeds associate with underestimation at all monitors. Remarkably
high overestimation is found at the most remote monitor ST3 during very weak
winds (<1 m/s at 10 m height). Surprisingly, the best performance of the model is
observed just by the road side (ST1). The overestimated low wind speeds and
underestimated higher winds are responsible to the moderate constants k
0
and
slopes clearly below 1 in Table
2
.
The results suggest that the mixing by the traffic flow cannot be described
merely by an initial constant in vertical and lateral dispersion parameters as is the
