Lessons Learned from Nested Regional Climate Model Experiments

Transcription

Lessons Learned from Nested Regional Climate Model Experiments
Lessons Learned from Nested
Regional Climate Model Experiments James Done
NCAR Earth System Laboratory Na3onal Center for Atmospheric Research Cindy Bruyère, Greg Holland
and Asuka Suzuki-Parker
NCAR is Sponsored by NSF and this work is par3ally supported by the Willis Research Network and the Research Program to Secure Energy for America The Nested Regional Climate Model:
Rationale and Approach •  An initiative of NCAR with university, government and private
industry colleagues.
•  Test bed for high-resolution climate modeling:
- combine weather and climate to utilize the best of both,
- gain experience with high resolution climate simulation,
- produce useful initial predictions of regional changes in highimpact weather: drought in the Western US, and hurricanes.
Approach: Nest the Weather Research and Forecasting (WRF)
model within the Community Climate System Model (CCSM).
Phase 1: Tropical Channel
45S - 45N
36 km resolution
51 levels
10 mb TOA
2000 - 2005
NNRP input data
Reynolds SST data
Periodic EW boundary
Tropical Channel: Results
Impact of Resolution
36km domain
18 storms
12km Nest
28 storms
27 storms observed
Asuka Suzuki-Parker
Phase 2: North American Climate
Image by Steve Dayo @UCAR
CCSM ~ 150 km
Nested Regional Climate Model Setup
WRF
WRF 12 km
36 km
CCSM ~ 150 km
Phase 2: North American Climate
•  CCSM3
–  1 current climate member (1950-1999)
–  3 ensemble members from 2000-2060 under two A2
and one A1B scenarios (IPCC-AR4)
•  NRCM:
–  1 ensemble member:1995-2005, 2020-2030 and
2045-2055.
–  A2 scenario at 36 km, and 12 km grid spacing (in
progress)
–  Lateral and surface boundary conditions from CCSM.
–  Surface is forced from CCSM data - no feedback to
ocean.
–  No nudging.
Regional Model: Results
Cat 3 Hurricane
October 2046 . . .
Regional Model: Results
Cat 3 Hurricane
October 2046 . . .
. . . . but
High Vertical Wind Shear
NCEP/NCAR Reanalysis
Aug-Sept-Oct
Average 1996
Channel
NRCM
Similar Bias in CCSM
CCSM
NCEP/NCAR Reanalysis
Sensitivity Studies
NCEP/NCAR Reanalysis
NRCM + channel configuration
NRCM
NRCM + Reynolds SST
Bias Correction
•  Describe any 6-hourly CCSM data set as an average
annual cycle plus a perturbation term:
- applied to variables: U,V,Z,T,RH,Surface T and
PMSL.
•  Do the same for NCEP-NCAR Reanalysis data:
•  Replace
with
to represent a bias
corrected future 6-hourly forecast:
Base climate provided by NCEP-NCAR Reanalysis
data and the weather and climate change signal
provided by CCSM
Choice of Base Period
•  The 20-year period of 1975-1994 was chosen based
on:
–  need to smooth out influences of El Niño
–  quality of data in early period is poor
–  exclude apparent climate shift in 1995
•  Caution: Multi-decadal variability, climate trends and
shifts.
–  1975-1994 had an average 8.9 TCs/yr
–  1995-2005 had an average 14.3 TCs/yr
Shear after Adjustment
NCEP/NCAR Reanalysis
NRCM
Adjusted N. A. Regional Climate
Simulated Satellite Imagery, Sept 2047
Results: Cyclone Tracks
1995-2005
7.6/yr
2020-2030
8.5/yr
2045-2055
10.4/yr
•  Increasing trend in annual storm counts
Results: SST Changes (IPCC A2)
Absolute
Relative
Results: Areal Frequency
Equatorward shift
Results: Maximum Intensity Location
Results: Intensity Change
1995-2005
2020-2030
2045-2055
Wind speed
distribution
shifting to right
Projected increase in
TC intensity, but 36 km
model cannot resolve
intense hurricanes.
Assessing Extreme Storms using
Extreme Value Theory
We utilize the Weibull distribution:
b ⎛ x ⎞
PDF = f (x) = ⎜ ⎟
a ⎝ a ⎠
b−1
e
⎛ x ⎞
− ⎜ ⎟
⎝ a ⎠
b
where parameters a and b determine the scale and the
shape, respectively. Weibull fit to normalized,
smoothed observed intensities
Results: Intense Hurricanes
Application of Extreme Value Theory
60%
PE69=Cat5;
PE58=Cat4,5
PE48=Cat3,4,5
PE32=All Hurricanes
30%
Conservative and consistent with
almost all other studies.
23 Summary
•  Climate model data required bias removal for the regional
model to reasonably reproduce hurricane characteristics.
•  North Atlantic frequency increase, with a 25% increase by
2050.
•  Observed equatorward shift over past decade to continue.
•  Intensity increases, with a conservative estimate of 30%
increases in major hurricanes by 2055.
Holland, G.J., J. Done, C. Bruyère, C. Cooper and A. Suzuki, 2010: Model Investigations of the
Effects of Climate Variability and Change on Future Gulf of Mexico Tropical Cyclone Activity.
OTC Metocean 2010. http://www.mmm.ucar.edu/people/holland/files/
OTC2010_Future_Hurricanes_Submitted.pdf
24 Next Steps
•  Two-way couple regional atmospheric model to a regional
ocean model.
•  Determine the contribution of internal model variability to
storm variability.
25 Computational Demands
Dedicated time on NCAR IBM Power 6 (bluefire):
•  24 nodes (~20% of total number of processors)
for ~ 2.5 months
•  36 (12) km simulations use 128 (256) processors per job
•  3.9M processor hours
•  ~300 Tb of data
•  Performed analysis as the model runs