Transcript NCAR_AGEDI_EAD_MODELING_V1

Title
Meeting Name/Presenter
Date
OVERARCHING GOAL
To develop a useful downscaled regional climate dataset to enable a variety of sectors to assess
the impacts of climate variability and change.
WHY DOWNSCALE?
Before
After
CHALLENGES TO ADDRESS
•Expense
•Downscaling properly can be expensive. How should we perform the downscaling?
•Model Uncertainty
•There are many global climate models (GCMs). Which climate model should we downscale?
•GCMs are imperfect. How do we address deficiencies in a given GCM?
ADDRESSING THE CHALLENGE OF EXPENSE
DOWNSCALING APPROCHES
•Statistical (uses empirical approach: cheap but cannot resolve processes well)
• Delta Method (add GCM mean change signal to present-day observations)
• Bias Corrected Spatial Disaggregation (BCSD)
•Dynamical (uses weather model: expensive but resolves processes well)
• Classic (directly downscale uncorrected GCM)
• Bias-corrected (correct the GCM with an observationally-based ‘reanalysis’ prior to dynamical downscaling)
•Hybrid
• Dynamically downscale lower-resolution outer domains continuously, and the higher-resolution (more
expensive) domains intermittently. Then, use statistical downscaling to fill in the gaps where high resolution
simulations were not dynamically downscaled.
• We will employ the Hybrid approach by dynamically downscaling bias-corrected GCM output and
employing BCSD to fill in the gaps.
ADDRESSING THE CHALLENGE OF MODEL UNCERTAINTY
There are approximately ~20 different GCM ‘families’ that support the newly released 5th Assessment
Report of the Intergovernmental Panel on Climate Change (IPCC). Which one do we downscale?
GCMs that best simulate presentday rainfall and temperature.
Knutti et al. (2013), GRL, doi:10.1002/grl.50256
To reduce uncertainty, we will downscale the NCAR CESM/CCSM4 model because it more accurately simulates
present-day climate than other global climate models. Additionally, we will bias-correct CESM/CCSM4 in order
to address uncertainty due to model deficiencies.
OUR MODELING APPROACH, PART 1
•Employ the Weather Research and Forecasting model (WRF) for dynamical downscaling
•Nested Domains: 36-km outer domain (D1); 12-km middle domain (D2); 4-km inner domain (D3)
MODEL DOMAIN
OUR MODELING APPROACH, PART 2
•25-year Historical ‘Truth’ simulation from 1980-2005. WRF driven with ERA-Interim
•120-year Historical+Future simulation with CCSM4 RCP8.5 simulation (‘business as usual’ emissions scenario)
•50-year Future “branch” simulation with CCSM4 RCP4.5 simulation (moderate emissions scenario)
MODLEING STRATEGY
OUR MODELING APPROACH, PART 3
•Employ a methodology that retains the more accurate ‘mean’ state from ERA-Interim, but retains the ‘eddy’
state from CCSM4.
•The ‘mean’ state is *always* a 25-year base period from 1980-2005, which ensures that the climate change
signal is included in the perturbation for CCSM4.
CCSM  CCSM  CCSM 
BIAS-CORRECTION METHOD
ERAINT  ERAINT  ERAINT 
CCSM R  ERAINT  CCSM 
=
ERA-Interim
+
CCSM4
EARLY RESULTS
In the following slides we show results from several case studies
• Tropical Cyclone Gonu, February 1-7 2007.
• A wintertime “Shamal” wind event, February 2008.
• A summertime month with a significant convective storm that brought rainfall, July 1995.
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
OVERVIEW
BACKGROUND
•Strongest Tropical Cyclone Ever
Recorded in Arabian Sea
•Extensive damage to Oman,
UAE, Iran and Pakistan
•Imperative that we can simulate
extremes such as GONU
MODIS Image courtesy NASA Earth Observatory
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
WIND TRAJECTORIES ANIMATION
THIS MOVIE
•Shows wind Trajectories for TC Gonu
•Colors = Wind Speed (blue = slower, red = faster)
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
RADAR REFLECTIVITY ANIMATION
THIS MOVIE
•Shows radar reflectivity estimate for TC Gonu
•Colors = Rainfall Intensity (blues = less; reds = more)
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
CLOUD TOP TEMPERATURE ANIMATION
THIS MOVIE
•Shows cloud top temperature estimate for TC Gonu
•Whiter colors = higher, colder clouds
•Greyer colors = lower, warmer clouds
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
WIND SPEED ANIMATION, 1000 M ASL
THIS MOVIE
•Shows wind speeds for TC Gonu
•Colors: blues = weaker winds; reds = stronger winds
WRF TEST SIMULATIONS: TROPICAL CYCLONE GONU, JUNE 1-7, 2007
OBSERVED SATELLITE INFRARED CHANNEL VERSUS WRF ‘PSEUDO’-INFRARED CHANNEL
WRF TEST SIMULATIONS: SHAMAL WIND EVENT, FEBRUARY 2008
WIND VECTORS ANIMATION, 250 M ASL
THIS MOVIE
•Shows vectors for Shamal wind event that occurred
during the Dubai Desert Classic golf tournament
•Colors: blues = weaker winds; reds = stronger winds
•Note the onset of the event from the northwest
MODIS Image courtesy of NASA Visible Earth
WRF TEST SIMULATIONS: SUMMERTIME CONVECTION, JULY 1995
IMAGE OF WRF RAIN WATER MIXING RATIO
THIS IMAGE
•Shows the rainwater mixing ratio in a WRF
simulation of the only convective rainfall event in
July 1995.
•This can be thought of as a 3-d depiction of rainfall
•Demonstrates that WRF can simulate these rare but
hydrologically important summer rainfall events.
WRF TEST SIMULATIONS: SUMMERTIME CONVECTION, JULY 1995
IMAGE OF MONTHLY TOTAL RAINFALL: SIMULATED (COLOR FILL) VERSUS OBSERVED (DOTS)
THIS IMAGE
•Overall WRF can simulate the patterns of rainfall in
the Oman Mountains, albeit with some biases due
to displacement.
CONCLUSIONS
•DY ??
[email protected]
AGEDI.ae