1. Implementation of a general tracer treatment
Anne Roches, C2SM
Oliver Fuhrer, MeteoSwiss
COSMO GM 2011, Roma September 2011

2. Goal Provide a consistent and flexible tracer treatment In order to
Make the introduction of a new tracer straightforward
Remove code redundancies
Improve readability and maintainability
Facilitate new developments
Ensure coherence
Suppress -> remove, avoid
Make the code easier to understand and manage -> improve readibility and maintainability 2

3. Applications “Conventional” tracer users:
Microphysics
Radionuclides
Aerosols
Chemical compounds
Pollens
Volcanic ashes
Water isotopes
Artificial tracers (indicator variables)
Pollens -> Pollen
ashes -> ash
- Radionuclides 3

4. What if… ?
I want to introduce a new tracer variable for the microphysics?
For example, the number concentration of rain drops for the 2-moment scheme?
remove “e.g.”
Let’s say the number concentraion… -> For example, the number concentration…
I would keep left-aligning the text (as on the previous slides!)
I would remove indent of second and following line of a paragraph… 4

5. Currently Introducing a new tracer (prognostic variable) is…
… a lot of work 5

6. 1st step
data_fields.f90
src_allocation.f90
src_setup_vartab.f90
qnr exists in the model, is = 0, and can be written in the output 6

7. 2nd step
organize_data.f90
src_relaxation.f90
qnr has IC and BC and a “correct treatment” of the initial and lateral boundaries 7

8. 3rd step
src_advection_rk.f90
lmorg.f90
qnr has an is advected and with appropriate boundary conditions 8

9. 4th step qnr undergoes artificial smoothing and turbulent mixing
hori_diffusion.f90
src_slow_tendencies_rk.f90
qnr undergoes artificial smoothing and turbulent mixing 9

10. Currently … and you still have “nothing” (e.g. microphysics)
Declare it in data_fields.f90
Allocate/deallocate it in src_allocation.f90
Add it to the table in src_setup_vartab.f90
Add it to the restart, IC and BC in organize_data.f90
Do the bound. relax. and Rayleigh damp. in src_relaxation.f90
Advect it in src_advection_rk.f90
Update the time step in lmorg.f90
Perform turbulent mixing in src_slow_tendencies_rk.f90
Perform horizontal diffusion in hori_diffusion.f90
Perform numerous boundary exchanges and clippings
… and you still have “nothing” (e.g. microphysics) 10

11. Further steps qnr used and modified in the microphysics src_gscp.f9011

12. Functionalities provided
Initialization
Memory management
I/O
IC and BC treatment
Advection
Turbulent mixing
Horizontal diffusion
Arbitrary meta-data 12

13. In the future
Introducing a new tracer (prognostic variable) looks like…
… an easy task 13

14. 1st step
organize_physics.f90 14

15. Further steps qnr used and modified in the microphysics src_gscp.f9015

16. Implementation 3 new modules: Modified modules: data_tracers.f90
src_tracers.f90
src_metadata.f90
Modified modules:
For advection (src_advection_rk.f90, src_leapfrog.f90, src_runge_kutta.f90)
For artificial diffusion (hori_diffusion.f90)
For turbulent mixing (src_slow_tendencies_rk.f90, slow_tendencies_rk.f90)
Main program (lmorg.f90)
For the restart, IC and BC treatment (organize_data.f90)
For BC treatment (src_relaxation.f90) 16

17. The tracer API User subroutines: trcr_new : definition of a new tracer
Provide its name, Grib n°/table, units, and optionally decide which operations it should undergo
Get optionally its index for the data, the boundaries and tendencies
trcr_get : access to a tracer
Provide the name/index of the tracer and optionally a time level
Get optionally its name, its data at the specified time level, its boundary data or tendencies, info about operations
trcr_get_ntrcr : access to the total number of tracers
trcr_errorstr : return of an error message
for each tracer 17

18. The tracer API Infrastructure subroutines:
trcr_init : initialization of a tracer structure
trcr_alloc : allocation of the memory for all tracers (data, boundaries, tendencies)
trcr_setup_vartab : I/O (mimic src_setup_vartab)
trcr_print : print of the list of tracers and associated metadata in sd out
trcr_cleanup : deallocation of the memory
1 x 18

19. Status First ß-version implemented and already used
Attempt to handle microphysics (qc, qv, qi, qr, qs, qg) using the tracer mechanism for RK core
Incoherence in their treatment
Different processes (e.g. surface processes)
Need of additional switches (=metadata)
Development of a metadata mechanism
Respect of the coding standards and contact with SCA 19

20. Outlook Finalize ß-version:
Resolve incoherence problems
Integrate the metadata mechanism
Prove correctness: fully (?) replace the microphysics
Performance analysis and improvements
Documentation
Undergo the procedure described in the COSMO standards:
Submit the idea to the SMC
Pass the tests (technical) and verification (scientific)
Nominate a responsible person
Submit the code and documents to the SMC and SCA
Implementation of additional processes 20

21. Acknowledgments: Pirmin Kaufmann, MeteoSwiss Ulrich Schättler, DWD Ulrich Blahak, DWD21

22. Thank you! 22