Environmental Engineering Reference
In-Depth Information
4. Ozone Forecasting
For ozone forecasting the operational statistical model performs well. But since it
includes only temperature as meteorological variable, some situations cannot be
handled well, for example when only part of the country is covered in cloud. LOTOS-
EUROS is better equipped to include meteorological variables. Furthermore, it
desrcibes the hour to hour development of the concentrations over the whole
country and provides the timing of ozone maximum values which enhances the
possibility to communicate the results to the general public.
The free-running LOTOS-EUROS model does not outperform the statistical
ozone model for a 1-day forecast, but it does so for longer forecast times. Average
concentrations are well modelled, but the variability of the model's extreme values
is underestimated.
A data-assimilation procedure was set up, using hourly ground-based obser-
vations from the Dutch national air quality monitoring network and networks in
Germany and Belgium, upwind of the Netherlands when conditions for smog are
favourable. The emissions of NOx, VOC, the top ozone boundary conditions and
deposition velocities were provided with random noise to create an ensemble. The
EnKF was used with 15 ensemble members, which appeared to be the best trade-
off between convergence and computational efforts. Indeed, the analysis was
improved substantially
(Fig. 2
).
Analysis
Analysis
300
300
200
200
100
100
0
0
0
100
200
300
0
100
200
300
LML daily 03 max [ug/m3]
LML daily 03 max [ug/m3]
1 day forecast
1 day forecast
300
300
200
200
100
100
0
0
0
100
200
300
0
100
200
300
LML daily 03 max [ug/m3]
LML daily 03 max [ug/m3]
Fig. 2.
Modelled (LE) versus observed (LML) daily ozone maxima. Upper panels: analyses of
free-running model (left) and model with data assimilation (right). Lower panels: 1-day forecast
of free-running model (left) and model with data assimilation (right)
