Environmental Engineering Reference
In-Depth Information
(a)
(b)
Average temperature at representative stations
(ASOS+RAWS)
Average dew point at representative stations
(ASOS+RAWS)
40
25
observed (ASOS+RAWS)
observed (ASOS+RAWS)
TD
35
20
30
15
25
10
20
Noah
15
5
0
6
12
18
24
0
6
12
18
24
Hours after OOZ 1 August 2000
Hours after OOZ 1 August 2000
Average temperature at representative stations
(observed and best WRF run-WRF-55))
(c)
(d)
Average temperature at representative stations
(ASOS vs.RAWS)
40
40
observed RAWS
RAWS stations
RAWS subset
35
35
Obs
Model
30
30
observed ASOS
ASOS stations
25
25
20
20
15
15
0
6
12
18
24
0
6
12
18
24
Hours after OOZ 1 August 2000
Hours after OOZ 1 August 2000
Fig. 2.
Time series for August 1, 2000, observations and simulations. See text
4. Model Physics Issues
Difficulties were encountered in WRF runs with some data sources and physics
combinations, manifested as substantial afternoon cold biases. Although the cold
bias appeared to develop “bottom-up” from the surface after sunrise on August 1,
nudging to the analysis “top-down” from the free atmosphere succeeded in con-
trolling the problem.
Figure 3 s
hows the difference 2 m temperature between
unnudged and nudged runs using RRTM and Dudhia radiation. This is a 48 h
forecast valid 00Z August 2. In many places, the unnudged forecast is up to 12°C
colder than the nudged forecast, which is more skillful relative to the observations.
In the Pacific Northwest, cold biases were negatively correlated with 2 m
moisture, suggesting an issue with the land surface model (
Fig. 4)
. Of particular
concern for this study, however, were predicted Central Valley temperatures that
were, on average, 7°C too low on the afternoon of August 1, even when RAWS
stations were excluded. The Central Valley problem appeared as nests were
implemented, and thus a longstanding bug in WRF may have been exposed. Both
problems were very largely mitigated through the employment of CAM longwave
and shortwave radiation, which removed the need for nudging.
