JCSDA Outlook

1
Next Generation of NR
Michiko Masutani January 2009
2
Joint OSSE will be concentrate on existing T511
and T799 Nature run produced by ECMWF for next
few years. There are several global meso scale
models are being developed. Atmosphere and
Ocean coupled model is becoming mature Climate
model with more than 20 aerosols and chemical are
also available Joint OSSE will keep posting
making good nature as one of the goals for
development of global models.
3
Strong demands for high resolution NR
Integrations of meso/regional OSSE effort into
Joint OSSEs
Note There are global meso-scale model (NICAM,
GFDL, ESRL) and relatively low resolution
regional OSSEs are considered.
Good hurricanes and storms in T799 run even for
meso scale OSSEs. Before producing regional
NR, it is highly recommended to perform regional
OSSEs (40-60km resolution) with T799 global NR.
Mesoscale NR must be another Joint OSSE NR
which will be shared within Joint OSSE Regional
OSSEs are affordable to Universities. Simulation
of observations may be difficult. Regional OSSE
must present evaluation of effect of lateral
boundary conditions.
4
Potential candidates for the next Nature Runs
5
ECMWF Plans to make T1279 model operational in
2009. The vertical resolution would remain
unchanged (L91). A non-hydrostatic version of
the IFS dynamics has been developed by
Meteo-France and partners. This is not envisaged
for operations until after 2018 (10 years from
now). ECMWF plans are to change from L91 to
L150 a year or so after the horizontal
resolution change. So that is currently planned
for 2010. Reference Presentation by Thomas
Jung http//www.emc.ncep.noaa.gov/research/JointOS
SEs/record/NR_diag/Jung_ECMWF_Nature_Run_TJ.ppt
6
Realistic MJO
Observed
From Fig 15 in Bechtold et al. (TM556) T159
horizontal resolution (80km) 91 level
7
NICAM Icosahedral atmospheric model Non
hydrostatic, cloud resolving model 14 km model
with limited coupling with ocean can be run for
several month (one year?) once the new computer
at Earth Simulator become available. (resource
to run 3.5km model for only one week)
Realistic MJO in 14km model. 500hPa height
forecast skill may not be so good. Out put can be
process to a regular lat lon grid. Vertical
coordinate number of levels ??
8
6 day forecast from NCEP analysis26DEC2006 2100
JST 3.5 Km model
NICAM
Observed
9
ESRL FIM (FIMS) FvGCM Icosahedral grid with
NCEP physics Hydrostatic version Vertical
coordinate Number of levels 64 Hybrid vertical
coordinate hybrid sigma-theta coordinate with
low level sigma pressure configured similar to
those of GFS        30 km resolution FIMS is
running forecast in real time              For
OSSE nature run, it is possible to run FIM with
15 km resolution. Targeting a 4km resolution
model in 2010NIM non hydrostatic version of
FIM Resolution targeting at a few km but it
heavily depends on the computer resource
        Number of levels 100 height coordinate
(?).         Time line of development uncertain
mainly due to the computer resource 2-D model
tested
10
NASA-GFDL model (GEOS-6 ) Target 2011 Dynamical
core - FvGCM, cubic globe non hydrostatic
Vertical coordinate and number of levels
?? Current version uses NCEP physics (Clouds are
parameterized) 1km 32 level is being
tested, C360 (25km) produce 70m/s in hurricane
One year run with 30km AMIP type run is
possible. With more resource one year run with
7km-12km resolution is possible If more
resource availableone year run with
Non-hydrostatic global cloud-resolving run at 4-5
km
11
GFDL Atmospheric model with coupled chemistry and
ocean(CM3) All chemistry in CRTM are included 1
degree 48 level model will be mature in
2012 72 levels require retuning of physics. There
is a demand from IPCC for 0.5 degree model.
Further detail plan become clearer over the next
12 month. NCEP CFS Ocean Atmosphere coupled
model Forecast skill by T386 (40km) coupled
model is not less than the uncoupled operational
model. Weaker but realistic MJO and ENSO in T62
coupled model For the detail pan_convection_2008
_08.ppt Posted at http//www.emc.ncep.noaa.gov/res
earch/JointOSSEs/record/2008/Nov08/
12
Realistic MJO
NCEP CFS coupled model with T62 resolution
V850 at 15N Coupled CFS T62
V850 at 15N Observed (GDAS)
13
Global version of NCEP NMM

• Rancic et al. (1996) Purser and Rancic (1998
  1989) introduced numerically generated
  quasi-uniform, quasi-conformal gridding of the
  sphere

• A global version of NCEP's Eta model has been
  developed at the University of Maryland, using a
  framework of optionally quasi-uniform cubic and
  octagonal grids (Rancic and Zhang 2006). 
• There is a new NSF sponsored project at the
  University of Maryland with the goal to further
  advance the developed quasi-uniform grid
  framework and combine it with the non-hydrostatic
  NCEP's NMM

14
NCAR/MMM New global weather and climate model
building from experience with the WRF and CCSM
models. Nonhydrostatic cloud resolving
model Horizontal grid Icosahedral hexagonal
grid with C-grid staggering Capability for
selective mesh refinement Finite volume numerics
with good conservation properties Intended to be
able to run at convection-permitting resolution
(1-2 km grids) for at least some experimental
simulations. At present, a number of technical
issues associated with this approach have been
resolved and the numerics is being implemented on
the sphere. It will be a couple years before we
have a fully functioning system capable of
simulating these high-resolution Nature Runs.
15
Planned NWP model changes at UK Met Office
16
Tropics and hurricanes
Intercomparison of African easterly wave
representation in operational numerical weather
prediction modelsGareth Berry and Chris
Thorncroft http//wwwt.emc.ncep.noaa.gov/seminars
/presentations/2008/Gareth_emc_081118.pdf TC
forecast evaluation at UK Met Office http//www.me
toffice.gov.uk/weather/tropicalcyclone/verificatio
n.html Go to the final table for the seasonal
mean track errors. Attached isthe northern
hemisphere 5-year running mean chart up to the
current date. It is showing a continued long term
downward trend in errors. As for AEWs, our model
is good at analysing them and propagating those
that are analysed, but is less good at developing
new waves during the forecast. This is one
conclusion from a joint paper with ECMWF,
NCEP,Meteo France and SUNY, which has just gone
for review to Weather and Forecasting. This
shouldn't have any impact on tropical cyclone
predictions in the Atlantic, since once a storm
forms it is initialised in the model and it is
then more down to the model's interaction of the
initialised vortex with the synoptic scale
environment which determines the quality of the
forecast track. (UK Net Office)
17
NOAA/AOML Aiming for 1km hurricane model. HWRF
model working close with ESRL, MMM and
NCEP. http//storm.aoml.noaa.gov/projects/hrs/wiki
High resolution model at by Earth Simulator
center in Japan 2-10km Global model 100
levels 100m-2km regional model Physics are
rather simple
18
Requirement for the meso scale (can be global)
Nature run - sample suggestions-
?Could be either global or regional. ?The NWP
model must have good forecast skill Great
visualization does not guarantee good forecast
skill. However, standard good forecast score
does not imply good Nature run. We need to work
on variation metrics for a good Nature run. ?Must
have a good TC or a severe storm in the nature
run period. ?Sufficient number of vertical
levels. Minimum 91 levels. ?Some degree of
coupling with ocean and land surface ?If it is
regional, the effect of the lateral boundary must
be evaluated. ? A list of verification method
must be produced by Joint OSSE. ? Need NR to be
shared within Joint OSSE ? User friendly
archive ? At least three month is required for
OSSE