Upload
justin-horton
View
217
Download
0
Tags:
Embed Size (px)
Citation preview
Comparison of Different Approaches
NCAR Earth System LaboratoryNational Center for Atmospheric Research
NCAR is Sponsored by NSF and this work is partially supported by the Willis Research Network and the Research Program to Secure Energy for America
Greg Holland
Note: This discussion applies only to regional climate simulations and interpretations
Summary of Different Approaches• Model Dynamics and Physics• Spectral Nudging• Horizontal Boundary Conditions• Surface Boundary (ocean and land)• Length of Forecasts
Potential Impacts of Different Approaches:• Level I: results in scatter, but does not substantially change
the overall outcome or interpretation;• Level II: may substantially impact the overall interpretation
but this cannot be fully quantified;• Level III: probably will substantially impact the overall
interpretation.
Holland Comparison of Approaches RPSEA 0310
Model Dynamics• Dynamical Core
– Discretization of the equations– Grid solvers– Accuracy and long-term drift
• Level I impacts on changes
Holland Comparison of Approaches RPSEA 0310
Cloud Physical Processes• For coarse grids, dx>15-20 km,
require cumulus parameterization• Issues
– Changing the parameterization scheme, or even tuning the inherent parameters can make a substantial difference to the results
– This is sorted empirically by experience and comparison with known climate (we did this in NRCM Phase I)
• Level I impacts on changes.
Holland Comparison of Approaches RPSEA 0310
Cloud Physical Processes ctd• Issues:
– For very fine grids, dx<4-5 km, cloud physical processes directly (called resolvable convection)
– From 4-20 km there is no good approach.– GFDL and others have used explicit clouds
and no parameterization in the 15-20 km zone– NCAR does not generally do simulations for
4<dx<12 km, above 12 km we use cumulus parameterization, occasionally with some explicit clouds also.
• Level II impacts, requires some further investigation to determine actual impacts
Holland Comparison of Approaches RPSEA 0310
Other Physical Processes
• Boundary-layer transfers, radiation budget, etc
• All are different, but all are also carefully compared to “reality”
• Probably Level I impacts.
Holland Comparison of Approaches RPSEA 0310
Spectral Nudging• Nudging the interior domain to force it to
follow the global model, with a spectral cut-off– e.g. GFDL use global wave numbers 0,1 and 2
for spectral nudging
• Issues may result in a false sense of accuracy when a good fit is obtained for current climate using global analyses, but it then can introduce substantial errors when forced to follow the global climate model for predictions
• Level II and Level III impacts.Holland Comparison of
Approaches RPSEA 0310
Example Impact of Spectral Nudging
2005 Aug-Sept-Oct(mm dd)
Observation15 Storms
Spectral Nudging9 Storms
Control12 Storms
Horizontal Boundary Conditions• The global model forces the regional
model across the horizontal boundaries, different approaches include:– Hand all information from the global domain– Use current weather and add climate
increments of , e.g. temperature, humidity, mean wind conditions, etc
– Bias correct the global model for known errors
• Issues: small domains, use of combination of current analysis and climate perturbation
• Level II and Level III impacts.Holland Comparison of
Approaches RPSEA 0310
WRF12 kmWRF
36 km Imag
e by
Ste
ve D
ayo
@U
CA
R
CCSM ~ 150 km
CC
SM ~
150
km
Nested Regional Climate Model Specifics
Holland Comparison of Approaches RPSEA 0310
Outer domain size chosen to maximize internal generation of relevant weather systems, such as easterly waves. And thus to minimize impacts of known biases in the global climate models.
Surface Boundary Condition
• Fixed Ocean (defined by bias-corrected global model)– Careful testing leads us to the conclusion that
this is not a major problem in the Atlantic….Level I Impact
• However– Eastern Pacific climate model bias definitely
leads to Level III impacts on the Atlantic….
Holland Comparison of Approaches RPSEA 0310
Holland Comparison of Approaches RPSEA 0310
NCEP/NCAR ReanalysisCCSM
Windshear Bias in CCSM
(Vecchi and Soden 2007)
Yamada et al (2010)
Their Conclusion: “Consistent with recent studies, frequency is reduced over the North Atlantic due to intensified vertical wind shear.”
Holland Comparison of Approaches RPSEA 0310
Length of Forecasts
• Computing limitations mean that compromises have to be made on the length of climate predictions at high resolution:– We decided on 3x11-y time slices– GFDL use 3-months Aug-Sep-Oct for each year– Yamada et al (2010) did a single 5-month
global simulation.
• Level I-III impacts:– For NRCM, Level I
Holland Comparison of Approaches RPSEA 0310
Summary
Holland Comparison of Approaches RPSEA 0310
The choice of model configuration and the way in which model biases are handled can have a substantial impact on the prediction. For existing studies the level of impacts are:• Model Dynamics: Level I• Model Physics: Level I-III• Spectral Nudging: Level III• Horizontal Boundary Conditions:
Level I-III• Surface Boundary (ocean and land):
Level III• Length of Forecasts: Level I