Upload
harry-whitehead
View
213
Download
0
Embed Size (px)
Citation preview
© Crown copyright 2007 Page 1
Current Activities and Plans in NWP
Clive Wilson, 19 September 2007
© Crown copyright 2007 Page 2
New Chief Executive from 17 September
Mark Hutchinson returning to Ministry of Defence after secondment to the Met Office
John Hirst, BA(Economics) No Atmospheric Science/Meteorology experience History:
“CEO of Premier Farnell from 1998-2005 Built an excellent leadership team redefined the business both operationally and financially repositioned it so as to grow market share and earnings. At ICI, during a period of high revenue growth, and undertook a
series of successful acquisition and merger negotiations” "I am very excited at the prospect of becoming Chief Executive of
an organisation at the forefront of one of the world's great issues. The Met Office is a truly world-class science organisation, with an excellent reputation, built on the talents of the staff who work for it. Working together, I am confident that we can harness our leading science base to enhance our impact and deliver even greater value to our public and commercial customers"
Probably means more emphasis on commercial activity and consultancy
© Crown copyright 2007 Page 4© Crown copyright 2007
CRM
SCM
OBS
UNIFIED MODEL
AQUAPLANET DYN CORE JULESIdealised
NWP &THORPEX
SEASONALDECADAL
CLIMATE
Km Scale
Global
Regional CLIMATEREGIONAL
NWPREGIONAL
Data Assimilation
Evaluation
Forcing
Evaluation
NWP configurations
© Crown copyright 2007 Page 7
2007 Operational Forecast Systems
UK 4km
Regional 12km+EPS
24member,24km
Global 40km+EPS 24member,90km
50 levels
38 levels
© Crown copyright 2007 Page 8
Model characteristics
N144 38LMOGREPS-G
N320 50LDeterministic
Grid points 288 (EW) x 217 (NS)
640 (EW) x 481 (NS)
Resolution: Mid-latEquator
~90 km 138 km
~ 40 km 63 km
Time-step 20 min 15 min
Same resolution to 15km
© Crown copyright 2007 Page 9
Limited area models
North Atlantic European (NAE)
UK 4 km 1.5km “On Demand”
Grid points/levels
600 x 360x38L 288x 360x38L 300 x 300x38L
Resolution 0.11o (12km) 0.036o (4km) .0135o(1.5km)
Time-step 5 min 100 s 50 s
Forecast runs/length
00,06,12,18 UTCT+48
03,09,15,21 UTCT+36 (+42 21UTC)
03 UTCT+18
Data assimilation
4DVAR+MOPS 3DVAR+MOPS Interpolated from UK4
Lateral boundaries
Global (T-6 run)hourly
NAE (T-3 run)30 min
UK4 (contemporary)
30 min
© Crown copyright 2007 Page 10
1
2
3
4
7
6
98
5
1.5km “On demand”
Global ModelSean Milton
© Crown copyright 2007 Page 12
Recent Global Model Upgrades
G37 Aug 2005 Implementation of soil moisture nudging scheme G38 Dec 2005 Resolution upgrade - 40km (N320) , 50 levelsG39 Mar 2006 Re-introduce AIRS and ERS-2 Scat. Replace Met-7 Satwinds with Met-8 satwinds. Tropical improvements - Physics upgrade (adaptive detrainment convection, improved surface
transfer in Marine BL)G40 Jun 2006 Improved soil moisture nudging over bare soil G41 Sep 2006 Satellite changes - SSMIS and GPS ROG42 Dec 2006 Increased use of ATOVS & Introduction of GPS RO.Model Spectral Gravity wave drag & Stratospheric stability package.Convective cloud decay.G43 Mar 2007 Introduce METOP data to replace NOAA15 data (early introduction 16th Jan) G44 May 2007 Introduce biogenic aerosols climatology bare soil albedo based on MODIS moisture flows into level below saturated or frozen soli rather than lost as run-off seasonal varying leaf area index (LAI) Assimilation of COSMIC GPS radio occultation data
© Crown copyright 2007 Page 13
N.Hemisphere T+24 RMSE – MSLPVerification vs. analysis
© Crown copyright 2007 Page 14
S.Hemisphere T+24 RMSE – MSLPVerification vs. analysis
© Crown copyright 2007 Page 16
Model Cycle G44 Operational 15 May 2007
© Crown copyright 2007
Problem – Poor Summer performance : warm bias over land, MSLP too low, NWP index poor vs other centres.
Global Model Physics Changes (also in HadGEM2) Change bare soil albedo – MODIS. Change runoff formulation in MOSES2 – deep soil layers wetter. Introduce biogenic aerosol – oxidation of hydrocarbon emissions from
plants. Introduce seasonally varying leaf area index (LAI). Two corrections to convective cloud
cloud phase was set according to the temperature at the top of the model rather than the top of the cloud. All convective cloud erroneously seen as ice cloud within the radiation scheme.
deep convective clouds are seen as liquid by the radiation scheme below the freezing level.
Satellite Observations – COSMIC (GPS radio occultation)
© Crown copyright 2007 Page 17© Crown copyright 2004
Model Cycle G44Improvements
in Model SurfaceAlbedo (vs. MODIS)
(Work in collaboration with U. Swansea as part of CLASSIC project)
© Crown copyright 2007 Page 18© Crown copyright 2004© Crown copyright 2007
Impact of Improving Model Surface Albedo Model cycle G44 – from 15 May 2007
UM too dark UM too bright
Model – MODIS (Annual) surface albedo Impact new surface albedo on 1.5m TempsT+48 Forecast
Prior to May 2007 physics changes the model surface albedo was too bright across the Middle East & SW Asia. This tends to make model too cold near surface as too much solar radiation is reflected back to space
Impacts on forecast temperatures – general warming of 1.5m temperatures throughout the Middle East & SW Asia– max values of order 3K over Iran and around Red Sea.
© Crown copyright 2007 Page 19
Screen T warm bias over land – Improvements with Cycle G44 Physics Package operational from May 2007JJA 2006 trials
Biases in screen T over NH land points : 1 Jun 2006 to 5 Aug 2006
Improved cycle G44 “Physics” reduces the summer warm bias
Summer warm bias not evident prior to soil nudging as “too moist”soil climatology kept surface moist and cool
© Crown copyright 2007 Page 20© Crown copyright 2004
Future Upgrades to Global Model
Nov 2007 Introduce IASI into data assimilation.
New OSTIA SST analysis. 2008
Enhance vertical resolution from 50L to 70L.
Improved soil properties & Van Genuchten soil hydrology.
PC2 cloud scheme.
Improved seasonally varying aerosol climatology (Mineral dust parametrization in Southern Asia Model)
Introduce snow analysis and improved snow parametrization (e.g. snow canopy) 2009
Enhance horizontal resolution – 40km to 25km
Turbulence based convection scheme – shallow, congestus, (deep?)
1 hour radiation timestep – currently 3 hours in global model. 2011
Enhance vertical resolution – 70L to 90L.
ENDGAME – Even Newer Dynamics
Limited Area modelsClive Wilson, Mike Bush, Jorge
Bornemann
© Crown copyright 2007 Page 22
Regional v Global
© Crown copyright 2007 Page 23
NAE recent developments
E10 13th December 2005 ATOVS & Satwind changes
E11 14th March 2006 Introduction of 4DVAR
E12 14th June 2006 Data assimilation upgrade
E13 26th September 2006 Schedule change (replace U.K Mesoscale model) Introduction of AIRS SST improvements Physics upgrade
E15 6 March 2007 Aerosol changes for visibility New balanced lbcs w-based convection instead or RH based
E16 15 May 2007 Reduced domain
E17 14 August 2007 Extra Perturbation Forecast model physics (convection)
© Crown copyright 2007 Page 24
NAE model cycle E11 – 14th Mar 2006
4D-VAR introduced 500hPa Height biases and RMS errors reduced.
© Crown copyright 2007 Page 25
NAE model cycle E11 – 14th Mar 2006
4D-VAR introduced 250hPa Wind biases and RMS Vector errors reduced.
© Crown copyright 2007 Page 26
RH-based CAPE closure
Until PS14 the NAE and Global models shared the same RH dependent CAPE closure to the convection scheme.
RH based closure was introduced to target grid point storms (GPS) by shortening the 30 minute CAPE timescale when RH>80%.
However the higher resolution of the NAE model compared to the Global model means that fronts are better resolved.
The higher associated RH along the fronts leads to CAPE timescales as small as 5-10 minutes even though there are no GPS along the front.
The effect of this short timescale is to produce noisy (both spatially and temporally) convective precipitation rates.
© Crown copyright 2007 Page 27
CAPE timescale as a function of RH
© Crown copyright 2007 Page 28
W-based CAPE closure impact on the NAE
The new vertical velocity based CAPE closure also targets GPS, but does not speed up convection where GPS are not occurring.
The CAPE closure is timescale is fixed at 30 minutes and does not reduce unless the vertical velocity exceeds 1m/s.
This produces smoother convective precipitation patterns and a redistribution between large scale and convective precipitation.
Case study testing has shown the w-CAPE scheme to improve precipitation and cloud scores
March 2007 U.K Index results show best ever ppn and cloud scores!
© Crown copyright 2007 Page 29
W-CAPE versus RH-CAPE – E15 6 March 2007
© Crown copyright 2007 Page 30
CAPE timescale as a function of W
© Crown copyright 2007 Page 31
W-based CAPE closure impact on the NAE
RH dependent CAPE (red) and w-CAPE (blue) 6hr ppn accumulation ETS scores averaged over 32 case studies
© Crown copyright 2007 Page 32
19th October 2006 case study at 15Z
© Crown copyright 2007 Page 33
Reduced domain to afford 70 levels
© Crown copyright 2007 Page 34
UK4 changes (2007)
U4.09 August 2007Physics upgrade (catch up with NAE)
Introduce radiation on slopes: including slope aspect and angle in direct solar radiationAerosol changes for visibilityNew balanced lbcs
Dec 2007Increase vertical resolution (70 levels)
Update Covstats (because of resolution change)
Planned:Introduce seasonal variability in Leaf Area Index
Revised soil moisture analysis (Van Genuchten soil hydraulics in the land surface scheme).
Split urban tile type into rooftops and street canyons (long term)
Review convection scheme (research ongoing)
© Crown copyright 2007 Page 35
70 levels for UK4
Lowest 4km
© Crown copyright 2007 Page 36
70 levels upgrade (Verification of 16 cases)
MSLP
Surface Temperature
© Crown copyright 2007 Page 37
70 levels upgrade (Verification of 16 cases)
Cloud fraction
10m wind
© Crown copyright 2007 Page 38
70 levels upgrade (Improved vertical structure ?)
© Crown copyright 2007 Page 39
70 levels upgrade (Subjective assessment)
38 L 70 L
Fog overdone in both forecasts
Plans
© Crown copyright 2007 Page 41
Global & NAE 70 Levels – March 2008
Lowest 4km
© Crown copyright 2007 Page 42
Supercomputer Procurement Plans
• Invitation to Tender Issued – Aug 2007
• Selected Supplier – May 2008
• System Installation – Q4 2008
• System Acceptance – Q1 2009
• Expectation of a 5 year contract with phased upgrades
• Initial System expected to provide >100TFlop peak with performance for applications ~6 relative to current NEC SX8
© Crown copyright 2007 Page 43
Met Office Production NWP Configuration Plans
2009 Global NWP 25km , 70 levels Global Ensemble, 50-60km , 70 levels to 15 days NAE ensemble 12km, 70L, 24M, 54hr UK NWP 1.5km , 70 levels, 36hr
2011 Global NWP 16-18km , 100 levels Global Ensemble, 100L, 50km to 15 days NAE ensemble 12km, 100L, 24M, 54hr UK NWP 1.5km ensemble (6 members+lagged) UK AQ 12km, 38L, 60tracers, 48hr f/c daily x2 DA; x1.4 physics/aerosol (relative to 5/07)
Even Newer Dynamics- EndgameNigel Wood
© Crown copyright 2007 Page 45
Aims of ENDGame
ENDGame: Even Newer Dynamics for Global atmospheric modelling of
the environment
Build on the foundations of the New Dynamics
Primary aim is improved robustness.
Secondary aims are: Improved accuracy
Retain greater non-linearity Reduce the need for significant off-centring of α
parameters Maintain/improve conservation
Inherent SL mass conservation Inherent water species + tracer conservation
While maintaining/improving efficiency
© Crown copyright 2007 Page 46
Improved accuracy
Iterative SISL scheme – Diamantakis et al (QJ 2007)
Following discrete normal mode analysis of Thuburn and Staniforth (MWR, 2004):
v (and v alone) to be held at poles (cf. present u, w and all scalars) improved energy properties
Coriolis terms based on mass flux variables improved Rossby mode propagation
© Crown copyright 2007 Page 47
Conservation
SLICE:Semi-Lagrangian Inherently Conserving and
Efficient advection scheme =
- SL finite-volume scheme
- Made efficient by applying a cascade approach
- i.e. a flow dependent dimensional splittingAll conserved variables advected via SLICEAll other variables advected using full 3D
interpolating scheme
Flux form for gradient operators
© Crown copyright 2007 Page 48
Switchable options
Retain deep-atmosphere, nonhydrostatic formulation
But introduce switches to investigate sensitivities:
Deep-atmosphere / shallow-atmosphere / Cartesian geometries
Hydrostatic / nonhydrostatic equationsSpherical / spheroidal coords Uniform / stretched grids Intrinsic mass & tracer conservation (on / off)
© Crown copyright 2007 Page 49
Where are we? And whither next
Now developing two 2D prototype models:Shallow-water model;Vertical slice model
Develop 3D prototype (2008)
Move to UM framework, couple to physics, DA (2009…)
Testing for operational implementation (2010/11?)
© Crown copyright 2007 Page 50
Questions & Answers
© Crown copyright 2007 Page 57
New Chief Executive from 17 September
Mark Hutchinson returning to Ministry of Defence after secondment to the Met Office
John Hirst, BA(Economics) No Atmospheric Science/Meteorology experience History:
“CEO of Premier Farnell from 1998-2005 Built an excellent leadership team redefined the business both operationally and financially repositioned it so as to grow market share and earnings. Formerly at ICI , CEO of the global Speciality Chemical Division during a period of high revenue growth, and undertook a series of
successful acquisition and merger negotiations” "I am very excited at the prospect of becoming Chief Executive of
an organisation at the forefront of one of the world's great issues. The Met Office is a truly world-class science organisation, with an excellent reputation, built on the talents of the staff who work for it. Working together, I am confident that we can harness our leading science base to enhance our impact and deliver even greater value to our public and commercial customers"
Probably means more emphasis on commercial activity and consultancy