Some possible future global modeling developments at NCEP

1
Some possible future global modeling developments
at NCEP
Where the Nations climate and weather services
begin
2
Upgrades and/or options available in GFS

• A restructured GFS is under test
• A two time-level semi-Lagrangian semi-implicit
  dynamics following ECMWFs approach (adapted by
  Late Dr. Sela) is an option
• This SLSI option is being tested at T1148 with
  linear grid and takes less resource compared to
  the operational T574 Eulerian model.
• Generalized hybrid coordinate option which can be
  used with or without enthalpy as a prognostic
  variable.
• NDSL semi-Lagrangian option by Dr. Henry Juang
• Near surface sea-temperature model (NST) option
  by Dr. Xu Li

3
Upgrades and/or options available in GFS

• Option of using Tiedtkes shallow convection with
  modification to get good marine stratus
• Ferrier microphysics (used in NAM)
• Relaxed Arakawa-Schubert cumulus parameterization
• Near sea surface temperature model (NST) is an
  option
• With these available options, multi-model
  ensemble with a single executable is possible.

4
(No Transcript)
5
Comparing grid-scale microphysics schemes
FEATURE Zhao Carr (1997) Modified version in GFS Ferrier et al. (2002) In Eta, WRF option
Prognosticvariables Water vapor, cloud condensate (water or ice) Water vapor, total condensate (cloud water, rain, cloud ice, snow/graupel/sleet)
Condensationalgorithm Sundqvist et al. (1989) Asai (1965)used in high res models
Precip fluxesand storage Top-down integration of precip, no storage, instantaneous fallout. Precip partitioned between storage in grid box fall out through bottom of box
Precip type Rain, freezing rain, snow Rain, freezing rain, snow/graupel/sleet (variable rime density for precip ice)
Mixed-phaseconditions No coexistence of supercooled cloud water ice, simple melting eqn. Mixed-phase at gt-10C, includes riming, more sophisticated melting/freezing
6
Flowcharts of Sundqvist-based schemes
Sundqvist et al. (1989)
7
(No Transcript)
8
Differences in Condensation/Deposition

• Zhao Carr (1997)
• RH bRHs (1-b)RHe (1)
• RH ?grid-averaged relative humidity
• RHs ? in-cloud, saturated RH (1.0)
• RHe ? relative humidity in cloud-free environment
• b ? cloud fraction (0 to 1)
• Assume RHe RH0 b(RHs-RH0) (2)
• Combining (1) (2) ?
• In free atmosphere, (RH0)atm0.75 (land), 0.80
  (sea)
• In lowest 10 model levels,RH0 increased from
  0.95 at sfc to (RHo)atm 10th level above sfc
• Partition moisture convergence between
  increasing RHe , b, qcw

• Asai (1965)
• Originally from Asai (1965)
• Adjust to target RH - condensation
• DQ q -qs
• qwater vapor mixing ratio
• qssaturation mixing ratio
• DQDQ1DQ2 (1)
• DT (L/Cp)?DQ1 (2)
• DQ2 L?qs?DT/(Rv?T2) (3)
• Putting (2) into (3) (4)
• Putting (2) (4) into (1)
• Rapidly convergesneeding at most 3 iterations
  for accuracy lt0.1

9
Precipitation Sedimentation
(Dh rk-1?VK-1 ? Dt) ? qKN Dh ? qKN PK-1N1
(Note PRODKN is a function of qKN)
(Dh rK ?VK ? Dt) ? qKN1 (Dh rK ? VK-1 ?
Dt) ? qKN Dh ? PRODKN
10
Deriving hydrometeor species from total condensate
11
1st-guess size of precipitation ice (snow) as a
function of temperature
0.01
Observed size distributions of ice as functions
of temperature, fit to (M-P) exponential spectra
as
From Ryan (1996)
D (mm)
N(D)Noexp(-l?D),
0.1
No is the intercept, l is the slope, and D
l-1 is the mean diameter

• HHHP (Washington state)SMPC (California)GM
  (California)PLATT (multiple locations)AWSE
  (Australia)YL (China)B, M (Europe)

1.0
10.0
Adjust D so that 0.1L-1 ? Ns ? 20L-1
12
Global Ice Spectra (Ryan, BAMS, 1996)
13
Global Ice Properties (Ryan, BAMS, 1996)
14
Other features of NGCP01 scheme

• Algorithm discriminates between cloud ice and
  snow (precip ice)
• No cloud ice if subsaturated (prevents too much
  sublimation)or if Tgt0?C (melting) ? only snow
  (precip ice) is present
• Ns 0.2Ni if at or above ice saturation T lt
  -8?C, -3?C lt T lt 0?C(Ni is number of cloud ice
  crystals, Ns is number of snow)
• Ns 0.1 Ni if at or above ice saturation and
  -8?C lt T lt -3?C

• Variable rime density ? assumes accreted liquid
  water fills holes of ice w/o changing volume ?
  (Total growth)/(Depositional growth)

• Efficient lookup tables store solutions for
  various moments (ventilation, accretion, mass,
  precipitation)
• Composite of multiple velocity-diameter
  relationships
• Increase in fall speed of rimed ice (Böhm, 1989)

15
What Happens in Areas of Strong Ascent?
Snow

• Sequence of more heavily rimed precip ice1.0
  (unrimed snow) ? RF ? 46.4 (sleet at 0C)
• When Ns(Ns)max DDmax, then increase RF to
  accommodate large ice mixing ratios
• (Ns)max20 L-1, Dmax1 mm at 0C? (rqs)1.2 g
  m-3 is max for unrimed snow

Graupel
Sleet
16
Cool Images of Rimed Snow Graupel(Electron
Microscopy Unit at the Beltsville Agricultural
Center)
17
Impacts of Riming Assumption

• Electron microscope images indicate rime can
  build up along outside of ice lattice, and not
  necessarily filter into the air holes.
• The sponge model assumed for riming will lead
  to a high estimation of ice-particle density
  (high rime factor, RF) when compared to real
  rimed ice particles.
• But my strong suspicion is that the ice-particle
  densities in this scheme produce much less
  graupel than those produced from Rutledge-Hobbs
  (1983) and Lin et al. (1983) 3-class ice schemes
  that predict cloud ice, snow, graupel (RH) or
  hail (Lin)

18
THE PHYSICS WHEEL OF PAIN
19
NSST NCEP GFS

• Analysis
• Introduction of NSST (T-Profile)
• Mixed Layer T ? T(z) Well-defined SST
• Obs. Treated as z-dependent as there originally
  are
• Observation depths for In Situ and satellite data
• Use of more observations
• Available in GSI already
• New data sets
• Assimilation
• 3DVAR in GSI ? Direct assimilation of radiances
• Forecasting
• NSSTM coupled to GFS Atmospheric Forecasting
  Model (AFM) in forecasting mode

20
NSST and NWP Interaction
FCST
ANAL
IC
Atmospheric Forecasting Model (AFM)
Atmospheric Analysis (GSI)
NSST Analysis (NSSTAN)
BG
NSST Model (NSSTM)
Radiative Transfer Model (CRTM)

Observation operator (relate T-Profile to the
radiance)
Jacobi (the sensitivity of the radiance to
T-Profile)
21
What is NSST?
NSST is a T-Profile just below the
sea surface. Here, only the vertical
thermal structure due to diurnal thermocline
layer warming and thermal skin layer cooling is
resolved Assuming the linear
profiles, then, 4 parameters are enough to
represent NSST
Diurnal Warming Profile
T
Mixed Layer
z
Thermocline
Skin Layer Cooling Profile
Deeper Ocean
z
z
5
22
NSST Analysis variable
Analysis variable a defined reference
temperature, currently the foundation
temperature. Therefore, .
Observation operators and their Jacobi Relate
the depth dependent data, Satellite , ch
(channel) dependent with skin depth of 0.1 1.0
mm in sub-layer. In Situ with depth of
0.2 15.0 m, to with CRTM and NSSTM for
direct assimilation.
Analysis increments
Tr
k time index
SST
Products
The boundary condition for GFS_AM
The boundary condition for CRTM
Will be used to combine NSSTM and OGCM
23
The NOAA Environmental Modeling System at
NCEPNEMS
23
24
What is NEMS?

• NEMS stands forNOAA Environmental Modeling
  System
• A shared, portable, high performance software
  superstructure and infrastructure
• For use in operational prediction models at
  National Centers for Environmental Prediction
  (NCEP)
• National Unified Operational Prediction
  Capability (NUOPC) with Navy and Air Force
• Eventual support to community through
  Developmental Test Center (DTC)
• http//www.emc.ncep.noaa.gov/NEMS/

24
25
NEMS motivation

• Develop a common superstructure for all NCEP
  models.
• Modularize large pieces of the models with ESMF
  components and interfaces.
• Divide atmospheric models down into Dynamics and
  Physics components but no further.
• Take history file I/O outside the science parts
  and into a common Write component.
• Keep science code and parallelization code in the
  respective models the same as before.

25
26
NEMS core developers
Ed Colon makefiles, scripts, regression
Nicole McKee documentation, web, testing
Ratko Vasic upgrades, regression, atmos coupling
Jun Wang IO, post, configuration
Weiyu Yang ensemble, earth coupling, ESMF
26
27
NEMS project developers
Tom Black Dusan Jovic Jim Abeles NAM
S Moorthi Henry Juang GFS
Jesse Meng Jim Geiger Land
Sarah Lu Arlindo da Silva GOCART
Tom Henderson Jim Rosinski FIM
Eugene Mirvis DTC
27
28
NEMS Component Structure
MAIN
NEMS
NEMS LAYER
All boxes represent ESMF components.
Ensemble Coupler
EARTH(1NM)
Atm
Ice
Ocean
GFS
FIM
NMM
Domains(1ND)
Wrt
Wrt
Dyn
Phy
Wrt
Dyn
Phy
Chem
Dyn
Phy
2
28
Below the dashed line the source codes are
organized by the model developers.
29
NEMS implementation plans

• 2011 implementation
• NMMB with nests
• 2012 implementation
• NEMS GFS Aerosol Component (NGAC)

29
30
NMMB with nests

• 12 km NAM will still run to 84 hr, with current
  output
• Fixed domain nests run to 60 hr
• 4 km CONUS
• 6 km Alaska
• 3 km HI PR
• Single locatable 1.33 km (CONUS) or 1.5 km
  (Alaska) nest to 36hr
• Nests
• Static, 1-way
• Boundaries from parent every timestep
• Nest is grid-associated with parent (same
  orientation w.r.t. earth)
• Moving nests and 2-way interaction under
  development

30
31
NEMS GFS Aerosol Component (NGAC)
Atmosphere
Color Key
Generic Component
unified atmosphere Including digital filter
Generic Coupler
Completed Instance
Physics
Dynamics
Dyn-Phy Coupler
NAM Phy
NMM-B
Phy-Chem Coupler
GFS Phy
GOCART
Spectral

• Dynamics, physics and chemistry run on the same
  grid in the same decomposition
• GOCART does not own aerosol tracers (i.e, do not
  allocate aerosol tracer fields)
• PHY2CHEM coupler component transfers/converts
  data from physics export state to GOCART import
  state
• Convert units (e.g., precip rate, surface
  roughness)
• Calculations (e.g., soil wetness, tropopause
  pressure, relative humidity, air density,
  geopotential height)
• Flip the vertical index for 3D fields from
  bottom-up to top-down
• CHEM2PHY coupler component transfers data from
  GOCART export state to physics export state
• Flip vertical index back to bottom-up
• Update 2d aerosol diagnostic fields

31
32
NEMS delivery plans

• 2011 deliveries
• GFS
• GEFS
• Postprocessor
• FIM
• Multimodel ensemble
• GRIB2 output
• 2012 deliveries
• NMM nested in GFS
• Moving nests
• Coupled ocean atmosphere
• Tiled land model
• netCDF output
• ARW

32
33
NEMS GFS Aerosol Component

• Status Update by Sarah Lu

NCEP is developing a common modeling framework
using Earth System Modeling Framework (ESMF)
infrastructure
NCEP is developing a common modeling framework
using Earth System Modeling Framework (ESMF)
infrastructure
Earth
Earth
Atmosphere
Atmosphere
NMM-B
GFS
FIM
NMM-B
GFS
FIM
DYN
PHY
CHEM
DYN
PHY
DYN
PHY
DYN
PHY
CHEM
DYN
PHY
DYN
PHY

• Community-based development on-going efforts to
  integrate new ESMF-based components into NEMS,
  including GOCART from NASA/GSFC, FIM from
  NOAA/ESRL, and MOM4 from NOAA/GFDL
• One unified atmospheric component that can invoke
  multiple dynamics (spectral, NMM-B, FIM) and
  physics (GFS, NAM)
• FY11 operational implementation for NEMS NMM-B
  (for regional applications)

• Community-based development on-going efforts to
  integrate new ESMF-based components into NEMS,
  including GOCART from NASA/GSFC, FIM from
  NOAA/ESRL, and MOM4 from NOAA/GFDL
• One unified atmospheric component that can invoke
  multiple dynamics (spectral, NMM-B, FIM) and
  physics (GFS, NAM)
• FY11 operational implementation for NEMS NMM-B
  (for regional applications)

34
Team efforts toward building global aerosol
forecast capability at NCEP

• Mark Iredell (NEMS framework)
• Shrinivas Moorthi (physics)
• Yu-Tai Hou (radiation-aerosol)
• Henry Juang (dynamics)
• Jun Wang (I/O)
• Hui-Ya Chuang (post)
• Weiyu Yang (replay capability)
• Ho-Chun Huang (GSI, verification)
• GSFC collaborators (Arlindo da Silva and Mian
  Chin)
• Downstream application (Xu Li, Jeff McQueen,
  Youhua Tang)

35
Implementation of prognostic aerosols in the NEMS
GFS
Physics with prognostic aerosols
Physics
Radiation
Radiation
Direct effect
Land surface processes
Land surface processes
Vertical diffusion
Vertical diffusion
Gravity wave drag
Gravity wave drag
Convective transport Tracer scavenging
Convection
Convection
Large-scale condensation
Large-scale condensation
Cloud scheme
Cloud scheme

Aerosol sources
Aerosol chemistry
Color Key
Dry deposition

GOCART grid component
Sedimentation
New routine
Coupler Transfers/converts data between physics
and GOCART convert units (e.g., precip rate),
calculations (e.g., soil wetness, relative
humidity) flip the vertical index
Wet deposition
Modified routine
Aerosol diagnostics
Unchanged routine
36
Off-line
In-line
Chemistry
Chemistry
Emission
Emission
Vertical Diffusion
Vertical Diffusion
Settling
Settling
?t 1 hour
?t DTF
Dry Deposition
Dry Deposition
Wet Deposition
Wet Deposition
Cloud Convection
Cloud Convection
Transport
Transport
This flowchart is taken from Ho-Chun Huangs ppt
37
Resources
Off-line System In-line System
Configuration 60-hr once per day (00Z), hourly output. 96-hr once per day (00Z) output every 6 hr The re-play mode
Resolution 1X1, L 64 T126 L64
Resource (memory) 1.24GB (max) 649 Mb
Resource (wall-time) 12531 sec (032851) with single processor 786 sec ( 13 min), using 60 tasks
Input files 15.4GB (raw GFS output), 3.4 GB (input ready meteorology, HPSS archive) 77 Mb siganl (63.8 Mb) and sfcanl (13.7 Mb)
Output files Raw output 18.7GB (peak 21.8GB), HPSS archived output 3.25GB 1.48 GB 1787 Mb sig sfc flx (8.8 Mb) aer (3.6 Mb)

The re-play mode meteorological fields are taken
from analysis (oper GDAS) and aerosol fields are
from previous day NEMS forecasts. The siganl is
blended from operational siganl and NEMS sigf24
38
The outcomes of GOCART aerosol fields
Off-line System In-line System
Provide dynamic dust/smoke LBCs for regional AQ forecasts YES YES
Provide global volcanic particulates transport tracking capability and regional LBCs YES YES
Radiation feedback in GFS NO YES
Atmospheric correction in SST retrievals NO YES
Include aerosol effects in GSI/CRTM NO YES
Aerosol data assimilation NO YES
Aerosol-cloud interaction in GFS/CFS NO YES

dust/smoke
volcanic ash
Full package
39
Global annual total aerosol emission, annual
averaged aerosol burden, and lifetime for dust
species. GFS GEOS4 offline
GOCART AeroCom Emissions (Tg/yr) 641.97 1970
3242 1789 541-4036 Burden (Tg) 32.5 31.6
38.4 19.2 1.4-33.9 Lifetime
(days) 18.18 5.85 4.33 4.22
0.92-18.4 Note The first column is the
result of NEMS/GFS-GOCART simulations, the second
column is the result of GEOS4-GOCART on-line
simulations Colarco et al., 2010, the third
column is the result of the offline GOCART model
Chin et al., 2009, and the final column is the
average/range of the AeroCom models.

• On-going and planned activities
• Refine and optimize the system preliminary
  results show weak emissions and removals in NEMS
  GFS
• Aerosol verification system (AERONET, MODIS,
  CALIPSO)
• Real-time system (proposed configuration T126
  L64, 4-day forecast from 00Z GDAS (for met
  fields) and NEMS (for aerosol fields)

in coordination with NRL, ECMWF, GSFC, UKMO,
JMA
From my last presentation
40
Global annual total aerosol emission, annual
averaged aerosol burden, and lifetime for dust
species. GFS HYB-10 GEN-10 HYB-07 GEN-07 GOE
S5 Emissions (Tg/yr) 641.97 688.75 629.11 690.68
617.51 1970 Burden (Tg) 32.5 29.26 20.21 33.01 20
.36 31.6 Lifetime (days) 18.18 15.5 11.7 17.5 12.
1 5.85 Note The first column is the result of
NEMS/GFS-GOCART simulations, the second column is
the result of GEOS4-GOCART on-line simulations
Colarco et al., 2010, the third column is the
result of the offline GOCART model Chin et al.,
2009, and the final column is the average/range
of the AeroCom models.

• GFS physics in NEMS has been updated to be
  consistent with operational GFS (R11579,
  committed on 23 Dec 2010)
• Four 13-month NEMS experiments are conducted
  (sigma-P and sigma-theta-P 2007 and 2010 RAS
  scheme with tracer scavenging / convective
  transport)
• NEMS vs GEOS5
• Emissions in NEMS are 1/3 of emissions in GEOS-5
• Lifetime in NEMS is 2-3 times longer than GEOS-5
• Need to adjust tunable parameters in GOCART as
  host AGCM is changed from GEOS-5 to GFS

41
Dust Source Function

• Function of surface topographic depression,
  surface wetness, and surface wind speed (Ginoux
  et al. 2001)
• S Source function sp fraction of
  clay and silt size
• u10 wind speed at 10 m ut threshold wind
  velocity
• A constant6.5 wt surface wetness

This slide is taken from Ho-Chun Huangs ppt
42
Global annual total aerosol emission, annual
averaged aerosol burden, and lifetime for dust
species. GFS GEN HYB HYBx GOES5 Emissions
(Tg/yr) 641.97 629 689 1288 1970 Burden
(Tg) 32.5 20.2 29.3 57.9 31.6 Lifetime
(days) 18.18 11.7 15.5 16.5 5.85 Note The
first column is the result of NEMS/GFS-GOCART
simulations, the second column is the result of
GEOS4-GOCART on-line simulations Colarco et al.,
2010, the third column is the result of the
offline GOCART model Chin et al., 2009, and the
final column is the average/range of the AeroCom
models.

• HYBx increase source function (from 0.175e-9 to
  1.80.175e-9) and wet scavenging (from 0.2 to
  0.8)
• Next steps
• Enhance removal efficiency in NEMS/GFS-GOCART
• Evaluate the model by comparing with observations
  (AERONET, MODIS, CALIOP, MISR) and modeling
  results (off-line GFS-GOCART, GEOS5),
• Determine model configuration and set up a NRT
  system

43
International Cooperative for Aerosol Prediction
(ICAP) Multi-model ensemble dust forecasts
Total AOD at 550 nm(march 07)
NAPPS (NRL)
NGAC (NCEP) and other centers
MACC (ECMWF)
GEOS-5 (GSFC)
44
Dakar, Senegal
Cape Verde
Kuwait
45
(No Transcript)