Model assessment and verification - Air Pollution Modeling and Its Application - page 289

6. Questions and Answers
Question: In the SARMAP study for the Central Valley of California, the MM5
developers carried out many sensitivity studies. They found that a high vertical
resolution was needed. Did you employ a high vertical resolution? Could the
stated problem be due to model resolution? What was the grid resolution in
your study (especially the lowest vertical grid)?
Answer: The MM5 configurations used in our study built upon a wealth of
knowledge from previous modeling and field study efforts. The lowest modeled
level in our MM5 simulations was around 24 m, and horizontal grid size was
4 × 4 km. We believe that this grid resolution is sufficiently fine to represent
three dimensional PM transport and dispersion processes. The major deficiencies
identified when validating the simulations related to localized flow features
(slope flows and land/sea breezes) that could not be reproduced by MM5. The
inability of this model to accurately simulate meteorological features at the
intra-day time scale (less than about 12 h) has been documented by a number of
researchers for various study domains.
Question: How much of the discrepancy in simulated PM levels was from
meteorological insufficiency under stagnant conditions, and how much was
from chemistry?
Answer: Our validation method only considers the fidelity with which meteoro-
logical conditions are simulated. The air quality simulation used the same
chemistry model for every day, and only the meteorology was changed.
Meteorology does affect PM precursor levels, which could indirectly affect the
chemistry. The proposed method cannot apportion biases in simulated PM
levels between these components.
Question: How is the “microscale” defined? Is it sub-grid scale, or defined physically
based on monitoring network size? What is the exact cutoff value?
Answer: The scales are based on EOF modeling of surface wind monitoring
network measurements. A large number of EOFs were estimated, and they
could be ranked in order of decreasing scale. (This is not true of EOF modeling
in general, but it has been verified for this particular study domain.) EOFs
associated with the microscale appeared to capture random fluctuations. The
fluctuations are not necessarily at the subgrid scale. The EOFs representing the
fluctuations were ranked higher than the 14 EOFs retained for the model
validation. There is no exact cutoff value based on the physical characteristics
of the measured winds.