Environmental Engineering Reference
In-Depth Information
observed value to modeled value for each parameter or scale by the average value
for each parameter of each cluster. One would expect the performance to improve
from the first (cruder) correction to the second (finer) correction. Yet as pointed
out in the last part of the paper, when we tried these two corrections with the
meteorological module, the match between modeled and observed clusters
improves in one and deteriorates in the other relative to the case without any
bias correction. We are still in the process of understanding such behaviour.
Question:
How are
parameters
(wind, temperature, mixing depth … etc.)
selected
for use in cluster analysis? Is there any
systematic
approach that has been used?
Answer:
We try to look at the match between output of an air quality modeling
system and observation as a whole with the matches of individual modules as
part of the analysis. As such, consideration of both meteorological and air quality
parameters to be used should not be done separately as meteorological para-
meters and air quality parameters. Since one step in this technique is to maximize
the match between observed meteorology and air quality clusters (Box 5 in
Fig. 1)
as a bench mark for the modeling system, whatever we think may enhance such
match is considered. One example is the trial inclusion of mixing height, a
derived meteorological parameter which has implications on air pollution
concentration. Another example is the trial inclusion of composite parameters
like photochemical smog (O
3
+ NO
2
) in air quality which may respond better to
meteorological forcing than the individual pollutants on the averaging time
scale used here. As such, there is no a priori way to determine the best list of
parameters, as this will vary from place to place given its dependence on the
peculiar combination of pollutants in conjunction with the climate. Guided by
our initial understanding of the air quality phenomenon, we will test different
combinations of meteorological and air quality parameters and then combine
the two to determine which combination of meteorological and air quality lists
of parameters will produce the best match in the observation data. The list of
meteorological and air quality parameters (separate) will be adopted.
