Modelling Activities at CAWCR, 2014 Presenter: Gary Dietachmayer, on behalf of colleagues throughout CAWCR WGNE-29 Melbourne, 10-13 March 2014 The Centre for Australian Weather and.

Download Report

Transcript Modelling Activities at CAWCR, 2014 Presenter: Gary Dietachmayer, on behalf of colleagues throughout CAWCR WGNE-29 Melbourne, 10-13 March 2014 The Centre for Australian Weather and.

Modelling Activities at CAWCR, 2014
Presenter: Gary Dietachmayer, on behalf of
colleagues throughout CAWCR
WGNE-29
Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
Major forecast systems summary
• POAMA-M24 (Seasonal prediction system) – March 27, 2013
• APS1 ACCESS-R – Apr-17 2013
• APS1 ACCESS-C – Oct-8 2013
• Solar – Ngamai switch – Dec 2013
• APS1 ACCESS-TC – Dec-19 2013
• (NWP) Ensembles
• NCI / NeCTAR (Tuesday – Tim Pugh)
• SREP / FDP (Wednesday Hi-res NWP)
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
A reminder on forecast system names ….
ACCESS – G / R / C / TC
WGNE-29, Melbourne, 10-13 March 2014
“POAMA” – Predictive Ocean
Atmosphere model for Australia
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
POAMA P24 to M24
• Core elements largely unchanged (BAM3, T47/250km, L17, ACOM2, two-deg,
L25)
• More emphasis on shorter (multiweek, instraseasonal) FCs, better early spread
• Introduction of atmospheric perturbations via breeding method (3 x 11)
P24, M24
Upper-tercile
precip
Significant MJO
improvement – 0.5 RMM
correl 17 -> 22.5 days
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-R (changes)
• Regional model rationalisation: “A0” + “R0” -> “R1” (12km, L70, large-domain)
• Component upgrades (eg., UM 6.4 -> 7.5)
• Additional data:
• IASI, GPS-RO, locally-processed ATOVS (RARS)
R0 (left),
R1 (right)
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-R (performance)
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-R (performance)
Best, middle, worst
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-R (performance)
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-C (changes)
• Component upgrades (eg., UM 6.4 -> 7.6)
• Resolution (0.05 -> 0.036 deg, L50 -> L70)
• Nesting in APS1 ACCESS-R
• New (trial) Darwin domain
• Still FC-only (no DA), now with two-step reconfig
• Operational-timing driven in part by SC upgrade
• Is still “overly convective” for tropical stability
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-C (performance)
• For many variables, similar to APS0
• Forecasters report improved wind-changes (topog/res?)
• However - worse T2m than APS0
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
Good: Mossvale station – APS1 better
OBS APS1 APS0
MOSS VALE AWS ,-34.5253, 150.4217 (Lat,Lon)
Bad: Bankstown station – APS0 better
OBS APS1 APS0
BANKSTOWN ,-33.9181, 150.9864 (Lat,Lon)
APS1 Urban Fractions are realistic
Almost
No
Urban
Areas in
APS0
APS1
0.01
0.1
APS0
0.01
0.1
APS0 Urban fractions are not ….
Almost
No
Urban
Areas in
APS0
APS1
0.01
0.1
APS0
0.01
0.1
ACCESS-C (SY) - impact of Urban/Tree
APS0
T2m Forecasts
RMS Difference
from observed (K)
90 Stations
2.1
APS1
Control
APS1
APS1
Revised Urban
Parameters
Revised Urban
Parameters
Tree Height=10m
2.4
2.2
1.9
APS0 looks good because it has almost no urban tile
fractions and shorter trees!
Solar – Ngamai SC upgrade
• Solar decommissioned Dec 2013
• Ngamai declared operational Dec 2013
• 576 compute nodes, 6912 compute cores (Intel Sandy Bridge)
• Opn-Capacity: 12.5 Tflops (0.25 * 50) -> 52 Tflops (0.5 * 104)
• Example:
• APS1 ACCESS-G wall-clock: 95 min -> 59 min
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
APS1 ACCESS-TC
• Usual component/level upgrade (UM
7.5 / L70)
• Had tried this previously with no
success
“APS10”, APS1
• Upgraded forecast-error covariances
(over full TC domain), vortexspecification
• Helped with previous 4DVar
excursions
• Improvement in long-period track, also
removes last “APS0” elements from
suite
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
Towards APS2: ACCESS-G
• N512 (25km), L70, UM 8.2 early trial system running for months now
• Obs-set not yet complete – require upgrade of current OPS
• Still to do: SSMIS (from F17/18), CrIS/ATMS (Suomi-NPP), OSCAT,
Windsat
NH
WGNE-29, Melbourne, 10-13 March 2014
Aust
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
Towards APS2: Plans
• Regional (R) – technical upgrade only
• Flag possible retirement in APS3+ ?????
• City (C) – 2.2km convection-permitting, still FC-only
• Ensembles – finally get the N216 global ensemble to operations
• “ACCESS-X” – on-demand system(s)
• Have used early-prototype – eg., to setup research-only 1.5km Darwin
Towards POAMA3
POAMA3/ACCESS
•
To supersede POAMA-2P and 2M
•
Based on ACCESS1.0 coupled model
•
Full coupled initialisation (PECDAS)
•
Operational ~2015 (??)
Achievements
•
Trial POAMA3/ACCESS system implemented on raijin
•
Extensive hindcast set is being performed (1980 – present)
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
ACCESS “ESM1”
Annual mean net ocean (red)
and land (black) carbon flux to
atmosphere. Aim is zero flux
under pre-industrial conditions.
• 14/15 Emissions-driven simulations
• Interactive atmospheric CO2
• Pre-industrial, historical, RCP
• With and without land-use change
Global flux (PgC/y)
• ACCESS ESM1 = ACCESS1.4 + carbon cycle + CABLE 2.1 + OASIS-MCT
• Coupling successfully achieved, carbon exchanged between land, ocean
and atmosphere
• Interactive or prescribed atmospheric CO2
• Currently running prescribed CO2 pre-industrial case
• Assessing stability of carbon fluxes
• Historical and 2 RCP cases to follow.
Year
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
ACCESS-CM2 development
Motivation for upgraded model
•
Anticipate improvements in realism due to improved model physics and resolution
•
World-class Australian contribution to CMIP6 and tool for climate science
•
Upgrade required for effective national and international partnerships
Plan – ACCESS-CM2
•
Enhanced vertical atmospheric resolution – 85 levels
•
•
•
Horizontal atmospheric resolution
•
Lower resolution “N96” (as current) version
•
Higher resolution “N216” (~0.55°lat; ~0.8°lon) version
Horizontal oceanic resolution – compare 1, 0.25 degree
•
•
Based on “GA6” atmospheric code
Development of high-res versions depends on efficiency on raijin
Include CABLE2 using the JULES frameworkThe Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
22
Achievements - ACCESS-CM2 development
New atmospheric component GA6.0 implemented at NCI
• At standard (N96) resolution so far – higher res (N216) will follow
• 25-year test simulation is underway – 8 years done so far
• Work is underway to couple to the other model components
There has been good progress in coupling CABLE to the atmospheric code via
JULES.
A collaboration has been established with the CoECSS in ocean model
development
• Global ocean model (0.25° lat/lon) implemented using the MOM5 code
• Potential to adopt as ocean component for ACCESS-CM2
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
Gary Dietachmayer
ESM/AMP RGL (acting)
Phone: 03 9669 4749
Email: [email protected]
Web: www.cawcr.gov.au
Thank you
www.cawcr.gov.au
Thank you
WGNE-29, Melbourne, 10-13 March 2014
The Centre for Australian Weather and Climate Research
A partnership between CSIRO and the Bureau of Meteorology
From Best et al (2006)
• “model runs revealed MOSES2.2 to be sensitive to
changes in the canopy heat storage”
• “Model performance was optimum with heat capacity
values smaller than those generally considered”
• The Urban tile canopy heat capacity parameter is “not
well defined and cannot be measured”
• “we conclude that the basic canopy scheme in MOSES 2.2
does not adequately capture the physical processes of the
urban atmosphere to accurately represent an urban area.”
Jules Urban Types
from Porson et al 2010
Thick Roof & Canyon
Thin Roof & Canyon
Watts/m2
Urban
Canopy
Thin Roof & Canyon with Radiative
Exchange
All fluxes
57
60
38
29
Net Radn
35
46
37
31
Sensible Heat
48
43
30
17
RMSE
Default
Scheme
Urban Canopy
Experimental Schemes
Roof
Roof
Canyon
Canyon
Fixed Tb
Soil
Soil
Comparison against observations from Mexico city
Soil
Conclusions
1) Many observing stations are in Urban areas
1) Most Australians live in the Big cities!
2) Properties of the Urban tile and Tree heights are
important
3) Models make simplistic assumptions about tree heights
4) Need for an accurate spatial map of Tree heights
5) Need for Improved Urban and Lake Tile Models for
ACCESS
6) We have performed NWP tests with revised Urban
parameters and tree heights that show significant
improvements in forecasts