Coupling TM5 to IFS within the GEMS-framework

1. Coupling TM5 to IFSwithin the GEMS-framework Frans Alkemade Henk Eskes (credits to Arjo Segers, Bram Bregman, Maarten Krol) GEMS Kick-off Meeting, Hamburg, 04-07-2005

2. Overview • Description of TM5 • Adapting TM5 for use in a coupled system with IFS, based on our PRISM experience • The route forward & further forward. • Some (rather trivial) conclusions.

3. TM5 – Tracer Model version 5 • Global offline CTM • History: • TM2 (Martin Heimann, MPI-Meteorology), • TM3 (IMAU, KNMI, CWI) : + hybride sigma-pressure levels • TM5 (IMAU, KNMI, SRON, Univ. Krete, JRC, NOAA) + parallel + zooming

4. Physics in TM5 • Advection: slopes [Russel and Lerner, 1987] or second moments [Prather, 1989] • Convection: [Tiedke, 1987] • Turbulent mixing: [Holtslag and Boville, 1993, Vogelesang and Holtslag, 1996] • Chemistry schemes: • Troposhere, based on CBM-IV (25 layers, 26 transp.tracers) • Stratosphere-troposphere (33 layers, 44 transp.tracers) • Parameterized ozone • Tracers: radon, SF6, methane, CO2, ..

5. Horizontal zooming • Arbitrary zoom regions • Two-way nesting • Typical runs with 3 level nesting

6. Parallelisation • Advection, diffusion, etc: parallel over tracers • Chemistry: parallel over layers --> limitation to 16-32 processors for reasonable speedup --> typical experiments run on 9 processors

7. “Classical” TM needs Meteo preprocessing TM5 CTM IFS TMPP MARS TM5

8. TMPP (TM PreProcessor): • Input from MARS: • Log.surf.pressure, vorticity, divergence • Temperature, humidity • Cloud fields • Surface fields • Horizontal Interpolation: spectral/gg -> lat/lon • Vertical collection of levels • Output: • Massfluxes through gridbox boundaries • Gridded temperature, humidity, etc

9. To couple TM5 to IFS, the ‘simple’ GEMS-adaptation would – in theory - look like this: IFS  OASIS4  TMPP  OASIS4  TM5

10. However, within the PRISM project TM5 has already been adapted to a coupled system with ECHAM: • Interpolation routines from TMPP were copied into TM5 • The “PRISM” version of TM5 is now able to: • Read meteo fields from disk … • … or from memory arrays (PRISM interface) • Interpolate from spectral/gg to lat/lon • Meteo input is 'embarrassingly' parallel: each processor a different field …

11. With the essentials of TMPP fitted inside TM5, it would be possible to use the following configuration: IFS  OASIS4  TM5 But another trick borrowed from our PRISM approach would be to compile TM5 into a library, which is called from a small separate program, “chem” (adapted from a PRISM toy model), that would handle all field exchanges with IFS. Further TM5-development can then take place completely separated and independent from GEMS-issues: IFS  OASIS4  chem {TM5-library}

12. The route forward: • We have started with the stable tropospheric version (CBM4 + aerosols) presently running at ECMWF • The interpolation routines from TMPP are being build into this TM5, yielding TM5+ • The explicit calculation of the chemical production- and loss terms for O3, CO, NO2, SO2, and HCHO (GRG; Global Reactive Gases) will be added to the model

13. The route further forward: • In the “work plan” proposal (dd 03/06/2005) step 2 involves the coupling between a toy-CTM model and IFS. • Presumably all relevant fields / arrays will be transferred to and from this toy model via OASIS4. • TM5+ can than be compiled into a library, and called from this toy-model.

14. The “parallel development” route: • In the mean time other people in the TM-community will continue working on the stratospheric version and on further parallelisation. • When a new version becomes available it can be compiled into a library that simply replaces the earlier library.

15. Conclusions • We have started. • But some work still has to be done. • The work-plan seems to be a good way forward.