Environmental Engineering Reference
In-Depth Information
segments from a “cold start” from 3 June 2008 at 00 UTC to 19 June at 00 UTC.
Seventy-two 48-h forecasts were then made 12 h apart from 19 June at 12 UTC to
24 July at 00 UTC, alternating between 12 and 00 UTC starts.
period 20 June to 23 July predicted by CHRONOS and GEM-MACH15. The
overall PM
2.5
spatial patterns are similar for the two models, but GEM-MACH15-
predicted PM
2.5
magnitudes are generally higher, especially over eastern North
America, and PM
2.5
transport over the Atlantic Ocean is greater.
Fig. 1.
Mean Day-1 20-UTC PM
2.5
concentration fields (μg m
−3
) for the 20 June-23 July 2008
period from averaging 12 (left) CHRONOS and (right) GEM-MACH15 00-UTC runs
Table 1
presents some objective scores for both CHRONOS and GEM-MACH15
against measurements for a similar period. Canadian and U.S. O
3
and PM
2.5
ambient
measurements were obtained in near-real time from Canadian stations and
the AIRNow meta-network (see http://airnow.gov). For both O
3
and PM
2.5
,
CHRONOS predictions are on average biased low for this period whereas GEM-
MACH15 predictions are biased high. Overall, the GEM-MACH15 scores for
hourly O
3
are slightly better than those for CHRONOS. For hourly PM
2.5
the
GEM-MACH15 mean bias and correlation coefficient (r) scores are better than the
CHRONOS scores but the unbiased RMSE score is worse. Except for the O
3
correlation coefficient, all of these differences are statistically significant based on
a bootstrapping calculation.
Table 1.
Objective scores for 72 CHRONOS and GEM-MACH15 runs for all hours of period
from 19 June (12 UTC) to 26 July 2008 (00 UTC)
Statistics
Hourly O
3
(ppbv)
Hourly PM
2.5
(μg m
−
3
)
CHRONOS
GEM-MACH15
CHRONOS
GEM-MACH15
Mean bias
2.5
0.7
−4.3
−2.1
Unbiased RMSE
15.4
15.3
12.8
13.5
R
0.68
0.68
0.30
0.40