CTM IFS interfaces GEMS GRG

1
CTM - IFS interfacesGEMS- GRG

• Review of meeting in January and more recent
  thoughts
• Johannes Flemming

2
Background

• Assimilating remotely sensed observation about
  atmospheric composition
• GEMS production is hosted at ECMWF
• Integrated Forecast System (IFS) provides
  background
• Transport, emissions, deposition and chemical
  conversion
• GRG-specific problem
• Chemical mechanism with gt 40 species can hardly
  be incorporated in IFS

3
GRG - approach

• Include observable species in IFS (transport and
  assimilation)
• O3, NO2, SO2, CO and HCHO can be observed from
  space (?)
• Assimilate species in IFS with ECMWF 4D-VAR
• Simulated source and sinks by coupling Chemical
  Transport Models (CTM)
• CTM MOZART (MPI-H), TM5 (KNMI), Mocage
  (meteo-france)

4
What do we have to consider?

• Fields to be exchanged
• Choice of coupling method
• Choice of model grid
• Frequency and timing of exchange in forecast and
  assimilation runs
• Common computing environment for IFS and CTMs
• Harmonised interfaces

5
Coupling method

• Coupler OASIS4
• Is being developed in PRISM (work in progress)
• Can exchange 3D fields between different grids
• Mass conservation atmospheric balances (?)
• Support of reduced Gaussian grid (?)
• Spectral and grid point presentation
• Coupling on the level of the MPI processes
• CTM should be MPI parallel
• File output possible
• Coding interfaces and control by XML scripts
• GEMS needs will be considered in development

6
CTM IFS

• Horizontal resolution
• IFS reduced Gaussian grid, T159
• CTM Regular or Gaussian grid, T63/ T42
• Vertical resolution
• IFS L90 Levels in sigma hybrid coordinate
• CTM L20-30 Levels in sigma hybrid coordinate
  (same coefficients)
• Orography surface pressure
• CTMs should use an orography consistent with the
  spectral truncation of the IFS output
• Advection scheme, chemical mechanism
• MPI parallelisation (?)

7
IFS fields to CTM

• Wind components mass fluxes
• Temperature
• Humidity
• Clouds
• Convective mass fluxes
• Profiles of precipitation
• Vertical diffusion coefficient
• IFS concentrations (data assimilation mode)
• Surface pressure
• ?

8
CTM fields to IFS

• Production P
• Loss rate L
• LC and PC due to chemistry (3D)
• PE due to emissions (2D)
• LD due to deposition (2D/3D cloud)
• Not standard output
• LC , PC ,LD and PE are not independent
• Totals only (?)
• Concentrations (Initial conditions) (?)

9
Dislocation problem
10
Dislocation problem

• IFS concentration fields differ from CTM fields
  because of different transport and data
  assimilation
• CTMs PL is not consistent with IFS
  concentration fields
• IFS chemical analysis is not chemically
  consistent with CTM fields (difficult to merge)
• Is this an issue ?

11
Forecast mode

• 2-way coupling requires synchronous run
• Exchanged met-data and PL have to be assumed
  constant till next coupling
• PL IC required
• How to consistently merge IFS chemical analysis
  in CTM IC
• 1-way coupling (CTM gets met-data)
• would allow lagged coupling (temporal
  interpolation of met-data)

12
Coupling in Forecast mode
13
ECMWF 4D-VAR Data assimilation T,u,v,q,O3
O3 transport chemistry
O3 advection only
O3 transport chemistry
14
Coupling in Data assimilation mode

• Tangent linear and adjoint of coupled system
  (IFS-OASIS4-CTM) would be needed (not feasible)
• Problems
• Including PL chemistry in inner loops (Do we
  need it? What are the consequences?)
• How to consistently transfer assimilated
  concentration to CTM IC
• Impact of analysed met data on CTM (PL)
• 1. Option
• No chemistry in 4D-VAR inner loops (as currently
  for Ozone)
• 2. Option
• Fixed L and P terms from CTM forecast
• no impact on CTM except for met data in out loop
• Does coupler work in 4DVAR (?)

15
CTM-implementation

• GEMS partners access to ECMWF HPCF
• Member state or special project access
• Source code management for CTM and Coupler
• Perforce projects for CTMS and OASIS (license?)
• Build and scheduling system for CTMs and Coupler
• ECMWF build-system for the CTMs
• Code sharing CTM and IFS
• no restriction to the CTM code (?)
• CTM parallelisation
• The CTMs need to be MPI-parallel for efficient
  coupling
• OASIS MPI1 mode only on IBM
• Model have to run with communicator provided by
  coupler
• Seems to be no problem

16
Status and (short-term) Plans at ECMWF

• IFS code
• CY29r2 based version with 5 GRG prognostic and
  Aerosol
• Identify best position for OASIS4 interfaces (met
  and chemistry)
• Routine for global budget calculation
• OASIS4 coupler
• Playing with code and toy models (local linux
  environment)
• No support of reduced Gaussian grid
• MPI2 mode (spawning) not possible
• Graphical Interface ready for editing of XML
  configuration files
• Build CTM toy model and couple it to IFS
• CTM Implementation
• Meeting on technical issues
• TM5 and Mozart run at ECMWF HPCF in special
  project status
• Implement CTMs in GEMS environment as soon as
  code is available

17
Issues CTM-IFS coupling I

• Technical implementation of coupler
• How long will it take to get OASIS4 running with
  IFS and CTM
• CTM efficiency, MPI communication, Coupler
  efficiency
• Data to be exchanged
• List complete
• Grids support (vertical and horizontally)
• Spectral or Grid point
• Different resolution and orography
• Mass conservative interpolation
• Coupling frequency
• High temporal coupling frequency reduces problems
  (requires efficient coupler)

18
Issues CTM-IFS coupling II

• Constant LP and met-data assumption
• Careful analysis of temporal scales of P and L
• Coupling in data assimilation mode
• chemistry modelling in inner loops
• Analysed met-data for CTM
• Different resolution and orography
• Different concentration patterns in CTM and IFS
  (Dislocation problem)
• PL fields dislocated
• Chemically consistent CTM analysis
• More chemistry in IFS (?)
• Using more balanced tracers (NOx, Ox) in IFS

19
END

• Thank You!