An Overview of New Developments with the NCEP Climate Forecast System

1
An Overview of New Developments with the NCEP
Climate Forecast System

• SURANJANA SAHA
• Environmental Modeling Center
• NCEP/NWS/NOAA
• 20th Annual Climate Diagnostics and Prediction
  Workshop
• State College, PA
• 24 October 2005

2
ACKNOWLEDGEMENTS SCIENTISTS AND STAFF OF THE
THE GLOBAL CLIMATE AND WEATHER MODELING
BRANCH ENVIRONMENTAL MODELING CENTER (EMC) AND
THE CONSIDERABLE HELP AND SUPPORT OF CLIMATE
PREDICTION CENTER (CPC) AND GEOPHYSICAL FLUID
DYNAMICS LABORATORY (GFDL)
3

• The NCEP Climate Forecast System (CFS) was made
  operational in August 2004
• Currently, two fully-coupled nine-month
  forecasts are made every day
• The present CFS operational system at T62L64
  resolution is frozen
• Development work is underway at EMC to improve
  the CFS
• We anticipate a new CFS implementation will take
  place in a few years

4

• For a new CFS implementation
• New upgrades to the CFS must lead to better
  performance
• Retrospective forecasts with the new CFS,
  covering a period of nearly 30 years, will then
  have to be made
• A global reanalysis of the atmosphere, land and
  ocean will have to be made, prior to that, to
  provide the initial conditions consistent with
  the new version of the CFS
• This is indeed an enormous challenge !!!

5
TESTING CHANGES IN THE OCEAN PART OF THE CFS
6
NEW OCEAN MODEL MOM4 (GFDL) Jiande Wang
Ocean-Only Simulation Run with MOM4 1981-2004
R-2 Daily Forcing (heat flux, E-P). Same
Resolution as operational MOM3, which has the
following configuration 1/3 degree at
equator, gradually decreasing to 1 degree at 30N
and 30S. Northern boundary is at 65N and southern
boundary is at 75S 40 Vertical layers, 10
meter interval in the top 220 m Depth to 5.5 Km
7

• Indonesian through flow is completely open in a
  larger area and the surrounding marginal seas are
  fully resolved as ocean (not land) points
• Use runoff data from NCAR (Dai and Trenberth)
• Use real fresh water flux, instead of virtual
  salt flux. This method gives more accuracy in
  the simulation of sea surface height

8
The bias everywhere is considerably less than
with MOM3
9

• NEW SEA ICE MODEL
• Xingren Wu
• Thermodynamics 3-layer (Winton, 2000)
• Dynamics EVP Model (Hunke and Dukowicz, 1997)
• The sea ice model was coupled to MOM4.
• The model was forced using R-2 climatology
• Sample results was for Year-20, March

10
Modeled sea ice thickness (Year-20, March)
The simulated sea ice distribution is reasonable,
but sea ice thickness may be not thick enough
11
Sea ice concentration (Year-20, March)
Observations
Model Simulation
Sea ice concentration is just a little bit too
high relative to the satellite observations
12

• Ocean Developmental work for the future
• MOM4
• Test higher horizontal resolution everywhere for
  MOM4 (¼ degree globally, or ¼ degree in the
  tropics decreasing to ½ degree at 30N and 30S)
• Increased number of vertical levels in the ocean
  mixed layer, from 40 to 50
• Use new, more efficient, coupler, with more
  frequent coupling to the ocean (than the present
  once a day)
• Include river (fresh water) runoff from
  climatology or NOAH land model

13

• Global Ocean Data Assimilation (GODAS)
• Explicit bias correction
• Geostrophic balance in the assimilation
• Altimetry and Argo salinity added to GODAS

14
TESTING CHANGES IN ATMOSPHERIC PART OF THE CFS
15
TEST THE CURRENT OPERATIONAL VERSION OF THE GFS
(USED FOR WEATHER PREDICTION) UPGRADES

• NOAH Land Model 4 soil levels. Improved
  treatment of snow and frozen soil
• Sea Ice Model Prediction of ice concentration
  and ice fraction
• Sub grid scale mountain blocking
• Reduced vertical diffusion
• RRTM long wave radiation

16

• GFS Developmental work for the future
• Test higher horizontal resolution (T126)
• Test new convection scheme (RAS)
• Test hybrid vertical coordinate (sigma-pressure,
  sigma-theta)
• Test improved boundary layer physics
• Test convectively forced gravity wave drag
• Test new short wave radiation parameterization

17
   TWO KINDS OF STUDIES CFS MONTHLY
RETROSPECTIVE FORECASTS CFS FREE COUPLED
SIMULATIONS (CMIP)
18
   CFS WEEKLY AND MONTHLY FORECASTS Higher
resolution, both spatial and temporal.  Spatial
T126L64 GFS Atmospheric model (as in
operational CFS) coupled to MOM3 Ocean
Model Temporal 4 CFS runs daily from
0Z,6Z,12Z and 18Z Atmospheric R2 initial
conditions coupled to the same Ocean GODAS
initial condition Period 2000 2004 (5
years) Summer 7 May 15 July (70 days) Winter
7 Nov 15 Jan (70 days) Integrations out to
65 days or more (covers the last full calendar
month)
19
POSTER OF AUGUSTIN VINTZILEOS WEDNESDAY,
9-1030 AM THE CFS 126 A DYNAMICAL SYSTEM FOR
SUBSEASONAL FORECASTS CHALLENGES IN PREDICTION
THE MJO U200 hPa forecasts averaged 20S-20N
and projected to the MJO EOFs obtained from R2
20
SUMMER
25 days
WINTER
13 days
Conclusions Good skill and for the reasons that
is shown in the poster,
we expect improvements....
21
POSTER OF HUA-LU PAN MONDAY, 9-1030
AM CLIMATE MODEL DIAGNOSES FROM A WEATHER
MODELERS POINT OF VIEW V850 hPa forecasts and
PRECIPITATION in the tropics from CFS
retrospective forecasts and free coupled runs are
examined
22
Conclusions Episodic nature of easterly waves
are well captured.
Tropical disturbances in the T126 are better
simulated
23
POSTER OF ÅKE JOHANSSON TUESDAY, 9-1030
AM PREDICTION SKILL OF NAO AND PNA FROM DAILY TO
SEASONAL TIME SCALES Skill in predicting NAO and
PNA indices (as deduced from geopotential at 500
hPa) from daily T126 and T62 retrospective
forecasts for 5 winters (DJF 2000-2004) are
examined, as well as the 24 winters of
operational T62 CFS
24
PNA
NAO
Conclusions There is lingering skill in both NAO
and PNA out to 45 days, albeit small. PNA has
higher skill than NAO in the short/medium range
while NAO has higher skill than PNA in the
intraseasonal time range
25
POSTER OF CATHERINE THIAW MONDAY, 9-1030
AM INDIAN SUMMER MONSOON PREDICTION IN THE NCEP
CLIMATE MODEL Using retrospective forecasts from
9-13 May, the CFS prediction of the onset of the
Indian summer monsoon is examined, using the wind
at 850 hPa, Precipitation and SST.
26
Conclusions The overall prediction of the Indian
monsoon rainfall is reasonable. The onset is a
little late and the rainfall is a little weak.
Higher horizontal resolution and better physics
may lead to improvements
27
PRECIPITATION (mm/day) JULY 2000-2004 XIE-ARKIN
CMAP
DIFF T126 - CMAP
DIFF T62 - CMAP
28
FREE COUPLED CMIP RUNS ATMOSPHERIC MODEL 1.
T62L64 CURRENT OPERATIONAL CFS
49 years (2002-2050) 2. T126L64
CFS OPERATIONAL VERSION 85 years
(2002-2084) 3. T126L64 GFS
OPERATIONAL VERSION 30 years
(2002-2031) OCEAN MODEL CFS OPERATIONAL MOM3
VERSION
29
T62 very regular T126 better variability,
weaker amplitude
30
POSTER OF CÉCILE PENLAND MONDAY, 9-1030
AM EL NIÑO IN THE CLIMATE FORECAST SYSTEM T62
vs T126 Prepare CFS output as we do COADS data
project SSTs onto a 4 x 10 grid, subject to a
3-month running mean, and then project onto 20
leading EOFs
31
Conclusions The resolution of the atmospheric
component of the model
matters a lot! The T126 does get the El Niño
spectrum about right
32
Good variability Good amplitude
33
Good variability Good amplitude
34
CDAS2
Amplitude CHI 200 hPa 107 m2/s T62 too
strong T126 better
35
CDAS2
Westward propagating Easterly waves
MJO
Phase Speed CHI 200 hPa m/s Eastward speed
T62 too slow T126 better Westward speed T62 too
strong T126 better
36
CDAS2
Semi-annual cycle in the observations not well
simulated in any run
Phase Speed CHI 200 hPa m/s At the
Equator Eastward speed T62 too slow T126
better Westward speed T62 too strong T126
better
37
Illinois 2m column soil moisture from CMIP runs
Courtesy Fan and van den Dool, Thursday 10.30
AM
Anomalies
Climatology
T126new has much reduced bias
38
POSTER OF SUDHIR NADIGA MONDAY, 9-1030
AM ENSO-RELATED SALINITY VARIABILITY IN
CFS Salinity simulation from a free coupled run
of the T62L64 operational CFS is compared to
salinity estimates from GODAS as well as
synthetic salinity dataset (Maes, 2000)
39
RED WARM EVENTS BLUE COLD EVENTS
Different water masses, caused by zonal advection
Conclusions Salinity contributes significantly
to the density variability in the global oceans.
The sub surface salinity shows a strong
ENSO-related signal in the western equatorial
Pacific Ocean
40
LAYOUTSEASONAL MEANS AVERAGED OVER 24
YEARS2007-2031

• CFS T126

• CFS T126 NEW

• CFS T62

• OBS

41
PRATE mm/day JJA
42
PRATE mm/day BIAS JJA
43
PRATE mm/day DJF
44
PRATE mm/day BIAS DJF
45
PRATE mm/day BIAS JJA
46
PRATE mm/day BIAS DJF
47
PRATE mm/day BIAS JJA
48
PRATE mm/day BIAS DJF
49
PRATE mm/day BIAS JJA
50
PRATE mm/day BIAS DJF
51
CONCLUSIONS A LOT MORE WORK NEEDS TO BE DONE
TO IMPROVE THE CFS. ACTIVITIES ARE IN PROGRESS IN
THE FOLLOWING AREAS INCREASED HORIZONTAL
RESOLUTION OF THE ATMOSPHERIC MODEL IMPROVED
PHYSICS AND NUMERICS IN THE ATMOSPHERIC MODEL NEW
OCEAN MODEL WITH INCREASED HORIZONTAL AND
VERTICAL RESOLUTION NEW COUPLER NEW LAND SURFACE
MODEL NEW ICE MODEL ESMF COMPATIBLE GLOBAL
ATMOSPHERE-LAND-OCEAN COUPLED REANALYSIS