Community Atmosphere Model

Dynamical Downscaling of CCSM Using WRF
Yang Gao1, Joshua S. Fu1, Yun-Fat Lam1,
John Drake1, Kate Evans2
of Tennessee, USA
Ridge National Laboratory, USA
9th Annual CMAS Conference, Chapel Hill, North Carolina
October 11, 2010
 The
Community Earth System Model (CESM) is
being used to simulate IPCC AR5 scenarios.
 To
study climate change on regional and local
scales, downscaling becomes an important
technique to link global and regional models.
 There
are high uncertainties in regional climate
downscaling. Different sensitivity cases are
needed to optimize the regional climate
Linkage from Global model to Regional Model
Community Earth System Model
Regional Climate Model
CESM 1.0
WRF 3.2.1
Community Atmosphere
Model (CAM)
Land Model
Ocean component
Sea Ice Model
36 km by 36km CONUS
1 degree by 1 degree
D1: 12 km by 12km domain
3 hourly resolution
D2: 4km by 4km Eastern US domain
Global and Regional Model Configurations
Most of the physics schemes are different in
CESM and WRF except radiation scheme
Sensitivity Scenarios
3.5-day overlapping run segments in January, April, July
and October in 2002
Typical Analysis Nudging vs. No nudging
Horizontal wind components (U and V) in all layers
Temperature (T) and Water vapor mixing ratio (Q) above the PBL
2) Scheme options comparison:
CAM/CAM vs. RRTMG/RRTMG for shortwave/longwave radiation
WRF: CAM 3 (A spectral-band scheme used in the NCAR
Community Atmosphere Model (CAM 3.0) for climate simulations.)
CESM (CAM4): (Parameterizations of shortwave and longwave
Radiation in CAM3 and CAM4 are the same)
Initial condition integrity
(Temperature) METGRID
(Relative Humidity)
(Wind Vector)
Patterns and Spatial distribution are similar
Time: 2002-04-01-00:00
Comparison between CESM and WRF
We mainly focus on the following parameters comparison on
the surface layer:
10 m Wind speed*, 10 m Wind direction*, 2 m temperature, 2 m specific
humidity and precipitation
(*Note: CESM does not output 10 meters wind speed and wind direction, so the lowest
model layer (around 60 m) values are used for the comparison)
Temperature at 2 meters
Overall, temperature
with RRTMG rw/lw
radiation scheme and
with nudging has lower
bias than the other two
Dashed time series represent
2m temperature from CESM
Correlation between temperature bias and temperature
There is high relationship between
the temperature and the bias. Bias
tends to change from positive to
negative when the temperature
from CESM increases.
Specific Humidity at 2 meters
schemes perform similar
with each other. Nudging
performs better for most
of the sub-region than no
nudging case. The biases
are mainly ranging from
-2 to 2 g/kg.
Wind Speed bias at 10 meters
Most of the cases, wind
speed has negative bias.
The two radiation
schemes have quite
similar performance on
wind speed.
Wind direction bias at 10 meters
Overall, nudging case has
much lower bias than no
nudging for wind
direction. Radiation
schemes does not have
much impact on the wind
Precipitation bias
Small bias in
January, April and
October. In July,
WRF predicts more
precipitation than
Comparison of WRF output with observational data
Climate data may not represent a specific year. We try to
evaluate how far the WRF downscaling simulations compared
with Meteorological Assimilation Data Ingest System (MADIS)
observational data.
Overall, RRTMG radiation scheme with nudging performs
better than the other two cases, so we only compare WRF output
with observational data for the RRTMG/NUDGING case.
Comparison between MADIS and WRF OUTPUT
Wind Speed at 10 meters
Mean obs
Mean prd
Wind Direction at 10 meters
Mean obs
Mean prd
Compared with observational data, the biases of wind speeds are
within 2 m/s for most of the sub-region.
Wind direction is also comparable with observational data.
Comparison between MADIS and WRF OUTPUT
Temperature at 2 meters
Mean obs
Mean prd
Specific humidity at 2 meters
Mean obs
Mean prd Bias
The bias of temperature and specific humidity ranges from -2 to 2
degree and -1 to 1 g/kg, respectively.
• Overall, nudging case performs better than no nudging case.
• RRTMG radiation scheme with nudging case shows the
lowest bias for temperature compared with CAM radiation
scheme and no nudging case.
• There is high relationship between the temperature and the
bias. Bias tends to change from positive to negative when the
temperature from CESM increases.
• WRF simulations driven by CESM are comparable to the
observational data, and the range of biases for temperature,
wind speed, specific humidity are from -2 to 2 degrees , from
1 to 2 m/s and from-1 to 1 g/kg, respectively.
Future work
Downscaling from CESM to WRF for12km by 12km
CONUS domain and 4km by 4km Eastern US domain
Chemistry Downscaling from CESM to CMAQ
Thanks for your attention!