CAM - CGD

Sensitivity to the PBL and convective
schemes in forecasts with CAM
along the Pacific Cross-section
Cécile Hannay, Jeff Kiehl, Dave Williamson, Jerry Olson,
Jim Hack, Richard Neale and Chris Bretherton*
National Center for Atmospheric Research, Boulder
*University of Washington, Seattle
Joint GCSS-GPCI/BLCL-RICO Workshop, NASA/GISS, 18-21 September 2006
Motivation
• Using forecast runs to test new parameterizations during
the model development ?
• Is the GCSS-Pacific Cross-Section a good candidate to
do this ?
Outline
• Models: PBL and convective schemes
• Cross-section: climate runs versus observations.
• Forecast runs settings
• Forecast errors along the cross-section
• Examples: 3 cloud regimes
– ITCZ region
– Trade-Cumulus
– Stratocumulus
• Conclusion
Models: PBL and convective schemes
- Boundary layer: Holtslag-Boville (1993)
CAM
- Shallow convection: Hack (1993)
- Deep convection: Zhang-McFarlane (1995)
- Turbulence scheme: Grenier-Bretherton (2001)
includes explicit entrainment at the top of the PBL
CAM-UW
(Chris Bretherton) - Shallow convection: cloud-base mass flux based on surface
TKE and convection inhibition near cloud base
CAM-dilute
(Richard Neale)
- Deep convection: parcels are diluted by environment air
Observations along the cross-section (JJA 1998)
SWCF
LWCF
CERES
LWP
SSM/I
CERES
Low cloud
ISCCP, D2
Mid/high cloud
ISCCP, D2
Precipitation
GPCP
Model versus observations
SWCF
LWCF
CERES
--- Obs
--- CAM
--- CAM-UW
--- CAM-dilute
LWP
SSM/I
CERES
Low cloud
ISCCP, D2
Mid/high cloud
Precipitation
ISCCP, D2
GPCP
Forecast run specification
Initialize realistically
ERA40 reanalysis
CAM
5-day forecast
Starting daily at 00 UT
• Strategy
If the model is initialized realistically,
we assume the error comes from the
parameterizations deficiencies.
• Advantages
Full feedback  SCM
Deterministic  statistical
Look at process level
• Limitations
Observations
ERA40
Accuracy of the atmospheric state ?
Forecast errors and climate errors (CAM-ERA40)
Forecast T error (K), day 1
Forecast q error (g/kg), day 1
Forecast T error (K), day 5
Climate T error (K), JJA1998
Forecast q error (g/kg), day 5
Climate q error (g/kg), JJA1998
• Cloud regimes => range of error structures
• Climate bias appears very quickly in CAM
• Climate error ~ Forecast error at day 5
Forecast temperature errors at day 5
CAM
CAM-UW
CAMdilute
CAM-UW
Some improvement in the cumulus region
CAM-dilute
Reduces T bias near ITCZ
Error increases above 300 mb and in the lower troposphere.
Changes in regions where the deep convection is not active
Select a range of cloud regimes and forecast errors
3 locations
Forecast T error at day 5, CAM
Stratocumulus
Trade cumulus
ITCZ
ITCZ regime: forecast T error (JJA 1998)
CAM
CAM-dilute
ITCZ region: very sensitive to the deep convective scheme
ITCZ regime: Temperature equation
T
T RT
 V  T   ( 
)  Qphysics
t
p pc p
Total tendency
Advective tendency
Physics tendency
Select a range of cloud regimes and forecast errors
3 locations
Forecast T error at day 5, CAM
Stratocumulus
Trade cumulus
ITCZ
Stratocumulus: moisture and PBL (JJA 1998)
Specific humidity
CAM
day 0
day 1
day 2
day 5
PBL collapses
CAM-UW
PBL height
Stronger daily cycle
Stratocumulus: timeseries of T and q error
TCAM-TERA40
qCAM-qERA40
Stratocumulus: q equation (single forecast)
q
CAM
CAM-UW
Advective
tendency
Physics
tendency
Stratocumulus regime (Physics terms)
PBL
tendency
CAM
CAM-UW
Shallow
tendency
Prognostics cloud
water tendency
Conclusion
• CAM forecasts allows for diagnosing model errors in the
different cloud regimes.
• Climate bias appears very quickly in CAM
– Where deep convection is active, error is set within 1 day
– 5-day errors are comparable to the mean climate errors.
• New schemes: CAM-UW and CAM-dilute
- CAM-dilute: improves the warm bias in upper troposphere, but cold
bias increases in lower troposphere and near top of the model.
- CAM-UW: does not change the error structure but CAM-UW
operates very differently than CAM at the process level.
• Difficult to decide what is causing the errors in such a
coupled system => need observations.
=> Comparison along the A-train
Observations along the cross-section
SWCF
LWCF
CERES
LWP
SSM/I
CERES
Low cloud
ISCCP, D1
Mid/high cloud
ISCCP, D1
Precipitation
GPCP
Cumulus regime: Forecast q errors
CAM
CAM-UW
Cumulus regime: moisture budget terms
2 PBL/ShCu schemes operate in very different way.
ITCZ regime: Precipitation (JJA 1998)
- GPCP Dataset
Daily precipitation
- CAM
Loses water very
quickly during day 1.
- CAM-dilute
Precipitation increases
during day 1.
ITCZ regime: Temperature equation
Stratocumulus regime (Q, CLOUD, CLDLIQ)