GMAO Modeling and Data Assimilation System Plans. Possible contribution to VOCALS

1
GMAO Modeling and Data Assimilation System Plans.
Possible contribution to VOCALS
Julio Bacmeister, Max Suarez, Lawrence Takacs,
Philip Pegion, and Caterina Tassone
2
Current status of GMAO global models. Plans for
the near-medium term
Coupled model(s) Operational model still NSIPP1
(now 7 years old) 2x2.5 atmosphere, 0.66x5/8
reduced gravity Ocean (old Poseidon) Diagnostic
clouds, First order Louis PBL scheme, RAS
convection Average Fluxes coupled once per
day 12-month coupled forecasts run every month
GEOS5 AGCM (description follows) to be coupled
to MOM, MIT and Poseidon ocean models. 1x1
atmos, 0.33x0.33 ocean EnKF ocean data
assimiliaton Readiness/utility for
VOCALS Results from NSIPP1 forecast runs can be
made available GEOS5 CGCM is still in
development. Plumbing is finished for MOM.
Analysis of runs/tuning has not begun.
3
Current status of GMAO global models. Plans for
the near-medium term
Atmospheric model and data assimilation system
(DAS) GEOS5 AGCM coupled to NCEP GSI using
incremental analysis updates (IAU) Modern Era
Re-analysis for Research and Applications (MERRA)
covering period 1979-present will begin in July
2007. Will run at 0.5x0.66 resolution using
frozen system. Completion expected in about 18
months. Emphasis hydrological cycle, changes in
satellite observing systems. Forward processing
stream covering . Runs at 0.5x0.66 resolution,
but uses evolving latest-greatest
system. Hurricane-season forecast projects
expected to occur every season, e.g., MAP05
July-December 2005, MAP06 June-October 2006. TC4
May?? 2007-??. Forecast runs conducted at
0.25x0.33 resolution. Initialized with NCEP
analyses 2005, 2006, in-house analyses in 2007.
4
Components of GEOS-5 AGCM and DAS
Dynamical Core Lin-Rood FV
Turbulence Lock et al 2000 w/ mods
Moist Physics Prognostic clouds RAS convection
Radiation M. D. Chou
GWD Boville/Garcia/Sassi
Land Surface Koster et al. catchment based
Analysis NCEP GSI w/ IAU
Plumbing/ infrastructure ESMF
Resolution 0.66x0.5 analysis/frcst, 0.33x0.25 frcst 72-levels
Aerosols/chemistry Da Silva, Colarco (CARMA, Aerochem)
5

• GEOS-5 Atmospheric Data Assimilation System
  (ADAS)
• NCEP-GMAOs Gridpoint Statistical Interpolation
  (GSI) analysis
• Radiance-based assimilation
• Adaptive observational bias correction for
  radiances
• JCSDAs CRTM

• Conventional Data
• Radiosonde u, v, T, q, Ps
• Pibal winds u, v
• Wind profilers u, v
• conventional aircraft reports u, v, T
• ASDAR aircraft reports u, v, T
• MDCARS aircraft reports u, v, T
• NEXRAD radar winds
• Dropsondes u, v, T, Ps
• GMS, METEOSAT, cloud drift IR and vis winds u,
  v
• GOES cloud drift IR winds u, v
• GOES water vapor cloud top winds u, v
• Surface land observations PS
• Surface ship and buoy observations u, v T, PS,
  q
• SSM/I precipitation
• TMI precipitation
• SSM/I wind speed

• GSI Input data streams
• Conventional data
• TOVS 1B radiances
• AMSU-A AMSU-B N15, N16, N17
• HIRS3 N17
• HIRS2 N14
• MSU N14
• SSM/I radiances
• GOES sounder TB
• GOES-10, GOES 12 chan 1-18
• AIRS (250 channels)
• EOS AMSU-A
• SBUV2 ozone N16

6
NCEP GSI used with Incremental Analysis Update
(IAU)
3 hours of free running model
Oval represents analysis. Increment applied over
6 hours as forcing term
3Z
9Z
12Z
15Z
0Z
6Z
18Z
Schematic diagram of Incremental analysis update
cycle
NOTE This is a continuous AGCM integration with
evolving forcing terms for u, v, T, q, and ps
7
NCEP GSI used with Incremental Analysis Update
(IAU)
Continuous AGCM integration with evolving forcing
terms for u, v, T, q, and ps e.g. for
temperature during tana-3hrs lt t lt
tana3hrs System structured such that
analyses (e.g. Tana) can be easily replaced with
any fields at AGCM resolution, replay mode,
e.g. 0.5o GEOS5 analyses interpolated to 0.25o,
external analyses. Impact of physics
modifications can be assessed without re-running
full analysis system
8
Prognostic cloud scheme Two families - 1)
convective source and 2) RH-based, large-scale
source equivalent to simple, bi-modal
prognostic PDF of total water
qtot,an
qsat
Constraints on PDF are provided by prognostic
values of ql,ls, ql,an , qi,ls qi,an and
fan Cloud Condensate sinks liquid autoconversion
(Sundquist type), ice sedimentation (Lawrence and
Crutzen), accretion Rain and snow subject to
reevaporation
9
Convection scheme

• RAS with enhanced microphysics, loosely following
  Sud and Walker (1999)
• Each call invokes spectrum of plumes
• Choose cloud-base, cloud-top.
• Find necessary plume entrainment
• Calculate mass fluxes
• Estimate vertical motion w for microphysics
• Apply trigger functions/adjustment timescales
• Adjustment timescales are ad hoc, but exert major
  control on simulations.
• uniform 1800s, 6hr, 24hr
• height dep 1800s for shallow
• to 24hr for deep

In-cloud total water profile
cloud
precip
Dz
vapor
10
PBL Scheme
Lock et al. 2000 turbulence scheme. Invoked in
unstable or cloud-topped PBLs. (obtained from
S.A. Klein)
Stable situations w/ no or weakly-cooling PBL
cloud still use Louis et al. 1982 1st order scheme
Non-locally determined K-profiles via test parcels
Strongly cooling PBL cloud.
Unstable surface layer (shown here added to cloud
top -driven Kzz.)
GEOS-5 extension unstable surface parcel
calculation includes moist heating and
entrainment. Parcel entrainment chosen to
represent smaller diameters than represented in
RAS.
11
Current status of GMAO global models.
Readiness/Utility for VOCALS

• Atmospheric model and data assimilation system
  (DAS)
• Available as part of existing GMAO plans
• MERRA and Forward processing streams from DAS
• IAU increments (Fu,v,T,q) contain information
  about model deficiencies
• Additional VOCALS outputs could be added
• High resolution runs for Hurricane-season
  forecasts
• Possible - contingent on manpower/computer
  resources
• Short runs assimilating VOCALS data, using higher
  resolution SST data, and/or high resolution
  replays
• Physics sensitivity experiments parameter
  tweaks to parameterization swaps
• GEOS5 Single Column Model is available to
  facilitate implementation

12
Questions

• How realistic are diurnal cycle, day-to-day SEP
  PBL and cloud structures in assimilation?
• Upsidence wave?
• Synoptic variability?
• Vertical structure?
• If not realistic why not?
• Parameterizations simply too far off?
• Not enough positive influence from data
• If realistic then
• How do relatively weak analysis tendencies keep
  run in line?
• Do analysis tendencies hint at problems in model
  physics?

Look at preliminary results for January 2006
13
Mean January 2006 precipitation from two GEOS5
analyses
Uniform RAS timescales 1800s
Height dep RAS timescales
GEOS5
GPCP
diff
Notice drizzle (0.4 to 0.8 mm/d) in SEP and
other stratus regions. Not in obs
14
Mean January 2006 precipitation from two GEOS5
analyses
Uniform RAS timescales 1800s
Height dep RAS timescales
GEOS5
GPCP
diff
Double ITCZ intensifies
Notice drizzle (0.4 to 0.8 mm/d) in SEP and
other stratus regions. Not in obs
15
GEOS5 Analyzed cloud fields compared with MODIS
images
Jan 4 2006 1930Z
Wm-2
Jan 18 2006 1930Z
Wm-2
16
GEOS5 Analyzed cloud fields compared with MODIS
images
Jan 4 2006 1930Z
Wm-2
Jan 18 2006 1930Z
POCs?
Wm-2
17
GEOS5 DAS Longitude-height cross-sections along
18-20S 5-day averages January 20-24 2006
fields
Physics tendencies
analysis tendencies
18
GEOS5 DAS Longitude-height cross-sections along
18-20S 5-day averages January 20-24 2006
19
Liquid water path too high in SEP (and
generally) Is deck too thick, or too wet? More
drizzle needed?
20
Current status of GMAO global models.
Readiness/Utility for VOCALS

• Atmospheric model and data assimilation system
  (DAS)
• Available as part of existing GMAO plans
• MERRA and Forward processing streams from DAS
• IAU increments (Fu,v,T,q) contain information
  about model deficiencies
• Additional VOCALS outputs could be added
• High resolution runs for Hurricane-season
  forecasts
• Possible contingent on manpower/computer
  resources
• Short runs assimilating VOCALS data, using higher
  resolution SST data, and/or high resolution
  replays
• Physics sensitivity experiments parameter
  tweaks to parameterization swaps
• GEOS5 Single Column Model is available to
  facilitate implementation

21
Questions/Summary

• How realistic are diurnal cycle, day-to-day SEP
  PBL and cloud structures in assimilation?
• Upsidence wave?
• Synoptic variability?
• Vertical structure?
• If not realistic why not?
• Parameterizations simply too far off?
• Not enough positive influence from data
• If realistic then
• How do relatively weak analysis tendencies keep
  run in line?
• Do analysis tendencies hint at problems in model
  physics?
• Are GEOS5 clouds ready for microphysics/aerosol
  improvement?
• Are mean SEP PBL and clouds more realistic in
  assimilation runs than in climate runs
• Systematic/coherent/coordinated approach needed
  to compare models in forecast/analysis mode with
  observations in SEP.
• VOCALS Rex
• Cloudsat/CALIPSO, GLAS?

22
Moist Physics for GEOS-5 Moist physics consists
of RAS convection and large-scale cloud
condensate scheme. Two condensate phases
liquid and ice Two condensate families
convective source, statistical source Moist
Physics State Variables (vapor, 4 condensates, 2
fractions qs and fs) RAS outputs detrain
ed mass and cloud condensates, precipitating
condensate profiles, updated T,q,u,v,
tracers Prognostic Cloud scheme inputs T, u,
v, qs, fs, convective mass and condensate
profiles outputs updated T, qs, fs, rain
rates (post-processed for radiation)
ftot, qvap, qice, qliq, Rice, Rliq
Effective radii