Environmental Engineering Reference
In-Depth Information
3. Operational Purposes: Forecasting and Emission Control
Scenario Analysis
The use of regional air quality models for forecasting has evolved from a
prototypal stage, about 10 years ago, to a currently fully operational stage. Several
centres now produce routine forecasts of air pollutants. The evaluation of the
forecast skill of prototype (Tilmes et al., 2002) or operational systems (Honoré et
al., 2008) has been made. A common platform of air quality has been set up within
the integrated GEMS/MACC European projects.
As an example, daily operational forecasts have been carried out since 2004,
using the models MOCAGE and CHIMERE within the PREV'AIR forecasting
platform, on the request of the French ministry of environment. This system is
evaluated on an annual basis and results for the first 3 years of operations are now
published (Honoré et al., 2008). The results (see
Fig. 3)
indicate in particular that
the fair skill of ozone forecasts slowly decreases with lead time, showing that the
accuracy of weather forecasts are not a major limitation to ozone forecasts.
0.85
0.8
Rural Stations
0.75
Suburban Stations
Urban Stations
0.7
-1
0
1
2
Lead Time of the Forecast (day)
Fig. 3.
Mean correlation between observed and predicted ozone daily maxima over most ozone
monitoring stations in France, as a function of lead time. Negative lead times stand for hindcasts.
The skill is calculated using the operational forecasts over three summer seasons
In the beginning of last decade a prospective vision of long-term emission
control has been built and models have been used to predict the expected air
quality improvement in 2010. This prediction has been coordinated at the European
level, within the Clean Air For Europe strategy, through the CITYDELTA and
EURODELTA projects (Cuvelier et al., 2007; van Loon et al., 2007), by the use of
several models.
The characterization of uncertainty in our knowledge is reflected in the spread
of the simulations carried out by an ensemble of air quality models for the same
weather conditions. The extent to which model ensembles spread correctly charac-
terize uncertainty has been discussed (Vautard et al., 2006). Ensembles have also
been used for producing improved forecasts or simulations of air pollutants, by the
use of advanced statistical techniques (Riccio et al., 2007).
