Environmental Engineering Reference
In-Depth Information
3. Classification of Simulated Winds
MM5 was used to simulate Central California meteorology for December to
January 2000-2001 on a 4-km horizontal grid with 30 vertical layers. Simulated
hourly winds were interpolated to 10 m AGL using the lowest modeled level for
grid points closest to the SFBA monitors. Each day of model output was
statistically classified among the five SFBA clusters (wind patterns) using EOFs.
Table 1 s
hows the correspondence of each pair of cluster (observation) and
classification (simulation) labels using low- and high-order EOFs. Low-order
EOF classifications reflect large-scale (synoptic) influences, and high-order EOF
classifications reflect more localized winds. Several results are noted. First, MM5
could not produce the R3 flow pattern. Second, days with observations assigned to
R2 were often simulated as R1. For low-order EOFs, nearly half (16 days) of the
33 R2 days were correctly classified. For high-order EOFs, the R2 days were
usually classified as R1 (25 of 33 days). Third, for days with strong large-scale
pressure gradients (clusters R1, V, or Z), simulation- and observation-based labels
generally agreed. Three of five V days were classified as Z, the other cyclonic
pattern.
Table 1.
Observed cluster names and numbers of assigned days for December to January 2000-
2001 (left panel). Numbers of simulated days classified to each pattern, using low- (middle
panel) and high-order (right panel) EOFs
Obs. clusters
Model low-order classif.
Model high-order classif.
name
# days
R1
R2
R3
Z
V
R1
R2
R3
Z
V
1
7
30040 50040
R2
33
13
16
0
3
0
25
6
0
3
0
3
7
31020 50020
Z
1
100 10 200 10
V
5
10032 10031
The R2-R1 mismatch and the lack of R3 are consistent with MM5's tendency
to exaggerate influences of the large-scale pressure gradient and understate localized
flows. Anticyclonic conditions were generally simulated as R1, the anticyclonic
pattern with the strongest synoptic forcing. R2 days were simulated with realistic
low-order EOFs, but the model departed from reality at finer scales (high-order
EOFs). As most PM episodes involve R2 (exceedances are rare under R1), the R2-
R1 meteorological mismatch should result in CMAQ underestimating PM. Since
MM5 distinguishes between anticyclonic and cyclonic conditions, it should reproduce
the timings of the onsets and terminations of PM episodes.
