| Home > Research > Ifsdocs > OBSERVATIONS > |
IFS documentation Front Page
Chapter 2. Observations: Types, variables and error statistics
Chapter 3. CMA creation (MAKEMA)
Chapter 4. The FEEBACK task
Chapter 5. The TOOLS task
Chapter 6. Central-memory array (CMA) structure/format
Chapter 7. BUFR feedback data structure/format
Chapter 8. SIMULATED-observations data structure/format
Chapter 9. NAMELISTS
Chapter 10. Processing of scatterometer data
REFERENCES
Previous Section
1.5 Invoking, initializing and controlling the OBSPROC
OBSPROC is invoked via program AAOBPPRO which in turn immediately calls the main controlling subroutine CNTOBSPR (Fig. 1.2 ). It is in this subroutine that it is decided what to do next. First, from the environment variables it is found whether it is a parallel or a single task run. This is done by calling UTIL_IGETENV subroutine. Then, several logical switches are preset and the basic I/O units defined. These presets are mainly to prepare things for a possible parallel run. Also, an initial setup of the OBSORT is carried out by calling the OBSORT subroutine SETUP_OBSORT.
The next step is to find out which task is to be carried out. As mentioned in Section 1.1, OBSPROC is a namelist driven module. Thus, the NAMELIST handling is initialized (ININAM), first, and the top level namelist NAMRUN is read in (READNL) next. The NAMRUN namelist contains, at the moment, 16 logical switches (for details see Chapter 9 `NAMELISTS' and Section 9.1). By setting one or more of those switches to .TRUE. an appropriate task/mode will be activated. Not all the NAMRUN namelist parameters are relevant for each task/mode. LMKCMA, LFEEDBAC and LTOOLS are the MAKECMA, the FEEDBACK and the TOOLS tasks switches, respectively. LMPP is the mode switch. All the other switches are to help in choosing how certain aspects of running the task are dealt with.
Once it is established what task and in which mode to run it, CNTOBSPR branches itself off into a section dedicated to that task. This is shown in Fig. 1.3 . What is actually done in a task section and underneath will be discussed in chapters dedicated to the relevant OBSPROC's tasks. However, regardless of which task section it ended up in, they all straight away will perform a preliminary initialization of various parameters and additional switches. This initialization consists of defining:
| • I/O units (SUIOD), |
| • various numerical constants (SUNUMC), |
| • various common parameters (SETCOM), and |
| • run settings (DEFRUN). |
In the context of operational running, the MAKECMA task is started by having:
| • LMKCMA=.T.,LCMASORT=.T., LMPP=.T., LCOLLECT=.F. |
and all the rest set to .FALSE.. This particular choice of LMKCMA and LMPP setting is obviously to run MAKECMA task in a parallel mode. The choice of setting for LCMASORT switch means that sorting of both the BUFR and the CMA data is activated. The aim of this sorting is to achieve a reasonable data distribution across processing elements (PEs) with a view to getting a good load balance. On the other hand, the choice of setting for LCOLLECT switch means that the number of output files is equal to the number of PEs used, alternatively they would be collected into one output file.
When it comes to running the FEEDBACK task in the operational context switches are set as follows:
| • LFEEDBAC=.T., LBFDBACK=.T., LMPP=.T.,LCOLLECT=.T., LMATCHUP=.T. |
and with all the rest set to .FALSE.. Regarding the LMATCHUP switch there will be more said about it later in the chapter dedicated to the FEEDBACK task.
Various settings of switches for the TOOLS task will be explained in the chapter dedicated to this task.
Next Section
Previous Section
 |
16.04.2002 |
 |
© ECMWF |
