Report of the WP3c meeting in Toulouse on 9 December 2003

Attendance: Yiwen Xu (MPI), Serge Planton, Vincent-Henri Peuch, Huebert Teyssedre (Meteo France),Sophie Valcke (CERFACS), Frans Alkemade, Peter van Velthoven (KNMI).

Reporter: Peter

Agenda:

Approach to coupling

Status of work

Interface definition

Where and how to do the interpolations of coupling fields

Compilation environment

Demonstration runs

Sample netcdf output for WP4

Action items

Approachto coupling

At the beginning of the meeting a note prepared by Guy Brasseur was presented, which was used to guide the discussions:
"The purpose of the meeting should be to establish a methodology nd concrete steps to link different CTM's (MOZART, MOCAGE, TM) with several AGCMs (ECHAM-5, ARPEGE) through the PRISM-system.

A first "coupling" should be MOZART-ECHAM5 . Subsequently "couplings" MOCAGE-ECHAM5 and TM3-ECHAM5 should be coupled, Only thereafter other combinations will be considered. It would be nice if some CTM could also be coupled to ARPEGE (MOZART or TM).
The idea is that models can be interchanged, but only tested interfaces and tested combinations can be considered for potential demonstration runs within PRISM.

Steps should be clearly identified so that different combinations can be tested within
a few months."

It was agreed by all that this is a reasonable way to proceed.

Furthermore it was agreed that:
Several coupled models (ECHAM-MOZART, ECHAM-MOCAGE, ECHAM-TM) should be able to run on the same computer platform. Highest priority is NEC (Hamburg), then Fujitsu (Meteo-France) and SGI (KNMI).
Yiwen Xu will try to run MOCAGE on the NEC supported by Vincent Peuch/Huebert Teyssedre.
Vincent Henri Peuch will attempt to run MOZART-ECHAM on the Fujitsu.

Status of work

MOZART-ECHAM5 coupling (on NEC) is ready, based on file-exchange.
MOCAGE-ECHAM5 coupling (on FUJITSU) is almost ready.
TM-ECHAM5 coupling: The TM meteorological pre-processor has been adapted for use of ECHAM output data (directly, i.e. without use of the afterburner). A few things remain to be done, such as the calculation of vertical velocity from the divergence. Also a few adaptations (such as height dependent emissions) are still necessary in TM itself to allow it to run with a different vertical model structure.

Interface definition

In order to make the coupling more flexible (general) and optimize the amount of work required to interface with PRISM it was decided to investigate the use of the coupler library (PSMILE) to exchange the data between GCMs and CTMs. This will allow us to easily swap from file exchange to use of a 2D-couple (OASIS3) or 3D-coupler (OASIS4). Another advantage is that the files will automatically be netcdf and PRISM CF-compliant.

The further adaptation of the CTMs to PRISM will be done in a 2-step procedure:

1. First an atmosphere toy model will be coupled to an atmospheric chemistry toy modeling order to demonstrate the way to implement this.
   The coupling will make use of the prism_put and prism_get subroutines and exemplify sequential running of the two models through the coupler.
   The coupling fields and coupling frequency are specified in a so-called namcouple file that describes the coupling configuration.
   Only one or a few fields will be exchanged between the two toy models, albeit in both directions. This will be done by Frans Alkemade (deadline 20 January).
2. The developed toy models will illustrate to the 3 modelling groups how to implement the prism_put an prism_get in the atmospheric chemistry models and how to use namcouple which defines the input and output fields of atmospheric and chemistry model. The next step will therefore be to make namcouple files and replace the file-write and file-reads by prism_put and prism_get statements in the atmospheric chemistry. models. (Please do this with a switch so that your original coupling remains possible and usable). The deadline for this is March 2003. Note that ECHAM-MOZART needs to be available for the full PRISM demonstration run after March 2003 according to the time schedule set up in the November 2003 project meeting in Brussels.

There was also a discussion about the possible time sequences of coupling. It was believed that sequential running in the way shown in the figure below is the best option. This requires swapping executables from the processor memories but that has not much consequences for the NEC. It was noted that a short time delay (1 or half an exchange time step) of the tracer concentration c may not be a problem for the radiation calculations in the GCM (unless we are dealing with chemical species that are strongly varying diurnally which seems not the case). The reason that this may occur is that some CTMs need information at the beginnning and the end of the time step (e.g. surface pressure for mass conservation).

Where to do the interpolations?

Interpolation is done by interpolators for the meteorological data (R1) and tracer data (R2), see figure below.

Ideally any PRISM GCM-CTM combination should ultimately be able to run

• without interpolation (i.e. identical grids)
• with reduced (or increased) spatial resolution (vertical and horizontal).

In case of reduced vertical resolution interpolation should preferably be done by merging model layers in the vertical, and by merging cells in the horizontal.

At present MOZART and MOCAGE can run on reduced gaussian grids. TM runs on a reduced equidistant grid in longitude-latitude, but has already implemented the option to merge layers or cells. All models can deal with arbitrary hybrid sigma-pressure coordinates, which will do for the moment because ECHAM and ARPEGE do so too.
The coupler and its library are unable to do the complex interpolations and transformations needed to couple CTMs and GCMs. OASIS4 will provide some more possibilities for 3-D coupling and calculations, but not all the complex interpolation methods applied in some CTM-GCM couplings. With OASIS4 part of the data exchange can certainly be handled more efficiently.
All transformations and interpolations of meteorological data that go from GCM to CTM are dealt with in the MOZART, MOCAGE and TM software, either in the same code or by a pre-processor code (see figure). This is probably the only feasible solution within the PRISM project period. As described above we will attempt to make this PRISM- and CF-compliant by making use of the PSMILE coupler library (note that this is a big effort!). MOCAGE and MOZART use a common program "R2" to transfer the zonal mean ozone distribution to ECHAM. TM is presently not capable to do so, but Yiwen Xu will make R2 available so that it can also be used in combination with TM. Ideally R1 and R2 should neither be part of the GCM- nor the CTM-executable. The R1* of TM will be decoupled from TM.
The code of the CTM (or CTM preprocessors) have strong dependencies on the parameters that describe the horizontal and vertical grid of the GCM. These parameters should therefore be available before construction of the compile script of the CTMs, through the SMIOC.

Compilation environment

This includes adapting to the directory structure and the writing of compile (.frm) and run-scripts (automatically generated makefiles), as well as code adaptation (a.o. to spatial resolution) by cpp-keys
We see clear advantages to adapting to the compilation environment but do not consider it a high priority, so it may be finished after June 2003. Peter will ask Stefanie Legutke if helping us will fit into her time schedule for the last half year of the project.

Demonstrations runs

a. Testing different combinations of GCMs and CTMs
We will run MOZART-ECHAM, MOCAGE-ECHAM, TM-ECHAM and MOZART-ARPEGE on the NEC. The latter 3 combinations have been developed in PRISM.
The models will be run for at least 3 spring months starting January 1 after a spinup period of several weeks and making use of PV regridding of CTM distributions of tracers with lifetimes longer than 1 week for the initialisation of the spin-up run. Meteo-France will make available their regridding software for this. Note that we need to run the GCM also for the spinup period and output its PV distribution at the beginning of the spinup-period e.g. on 1 December.
Vincent Henry Peuch will make a draft list of standard diagnostics, that will probably include ozone column and model climatologies of tracer concentrations at NOAA surface stations.

The tracer quantities to be transferred to the GCM are:

• Zonal mean ozone volume mixing ratio
• Water vapour volume mixing ratio profile (to be used in the GCM in the stratosphere, e.g. for volume mixing ratios smaller than 10 ppmv as in MOCAGE).
• Zonal mean methane volume mixing ratio? or Global mean methane volume mixing ratio?

Deadline is 4 March 2003 for MOZART-ECHAM and June 2003 for the other combinations, in view of the full run described below.

b. Full PRISM model run

This will be done with ECHAM-MOZART on the NEC in Hamburg, using file-exchange rather than MPI-messaging. Deadline for the run is June 2004.

Sample output for the visualisation workpackage

Yiwen Xu will make netcdf output files from Mozart available for development and testing of the visualisation software by WP4 on the Hamburg ftp-server (has been done).

Action items/Time planning