 |
| Titel |
Which boundary species are required? An evolutionary evaluation |
| VerfasserIn |
C. Bergemann, J. Meyer-Arnek, T. Erbertseder |
| Konferenz |
EGU General Assembly 2012
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 14 (2012) |
| Datensatznummer |
250069627
|
|
|
|
|
|
| Zusammenfassung |
| Analyses and forecasts of air pollutants at regional and local scales require air quality
modelling at high resolution therefore introducing the necessity of nesting into larger scale
models.
Currently within the FP7 GMES downstream service project PASODOBLE air
pollution forecasting services for regional and local applications are developed.
These activities are complementary to the large scale service currently developed
by the Atmospheric Core Service MACC (Monitoring Atmospheric Composition
Climate).
We report on our results regarding the question which boundary information from
MACC is needed for smaller-scale air pollution modelling and forecasting. The
major problems are expected to result from chemical incompleteness as well as
chemical inconsistency with the small-scale model. Currently, the MACC European air
quality ensemble product contains four single species (CO, NO2, O3 and SO2) and
one aggregate species (PM10). This is clearly much less than models require if
full nesting into the MACC-ensemble is desired. Here we study the problem of
chemical incompleteness and the question, which species should additionally be
provided.
We employ the POLYPHEMUS/DLR air quality model for a region covering Southern
Germany. In order to isolate the effect of chemical incompleteness, boundary conditions are
generated using the same model but covering a larger domain. In this way we obtain
boundary conditions for all chemical species.
We perform a large number of simulations with different subsets of the available
boundary species. An evolutionary feature selection approach allows us to estimate, which
species have the most important influence on the model results. Each simulation is compared
to a reference simulation using full boundary conditions.
In every generation of the evolutionary algorithm one boundary species selected from a
pool of available species is added. The five best setups, evaluated with a cost function, are
advanced to the next generation.
The largest improvement is obtained by adding PAN to the set of boundary conditions.
Other important species are HNO3, isoprene and methane. |
|
|
|
|
|
|