| Home > Research > Ifsdocs > ASSIMILATION > |
DATA ASSIMILATION
IFS documentation Front Page
Table of contents
CHAPTER 1 Incremental
formulation of 3D/4D variational assimilation-an overview
CHAPTER 2 3D variational assimilation
CHAPTER 3 4D variational assimilation
CHAPTER 4 Background term
CHAPTER 5 Conventional observational
constraints
CHAPTER 6 Satellite observational
constraints
CHAPTER 7 Background, analysis
and forecast errors
CHAPTER 8 Gravity-wave control
CHAPTER 9 Data partitioning (OBSORT)
CHAPTER 10 Observation screening
CHAPTER 11 Analysis of snow
CHAPTER 12 Land surface analysis
CHAPTER 13 SST and sea-ice analysis
CHAPTER 14 Reduced-rank Kalman filter
Previous Section
3.3 Inner and outer loops: practical implementation
Similarly to 3D-Var, job steps are carried out with different configurations of the IFS:
| (i) The first trajectory run (which includes screening) - conf=2, LSCREEN=.T. |
| (ii) The background error minimization, conf=131, LAVCGL=.T. |
| (iii) The main minimization, conf=131 |
| (iv) The update of the trajectory , conf=1, LOBS=.T. |
Steps (iii) and (iv) are performed
times where 
is the number of outer loops or, equivalently,
of updates of the trajectory.The first trajectory run (i), the background-error minimization (ii) and the first main minimization use the same input files as described for 3D-Var in Subsection 1.3.1 of Chapter 1 `Incremental formulation of 3D/4D variational assimilation-an overview' , the only difference being that the background field is a 3-hour forecast from the previous analysis at synoptic time, compared with a 6-hour forecast in 3D-Var.
The ouput of the minimization steps are the files MXVAxx000+000000, MXVAxx999+000000 (as in 3D-Var), trajxx+0000000 and VATRH. xx is an integer varying from 0 for the first minimization to (n-1) for the last minimization, where n is the number of updates of the trajectory. VATRH contains useful information for a warm restart of m1qn3 (including the diagonal of the Hessian). trajxx+0000000 contains the control variable at the end of the minimization. The file trajxx+0000000 is written out in SAVMINI called at the end of CVA1. This file will be an input to the next minimization in addition to the background file used as in the first minimization. It is read in in GETMINI called from CVA1. The file VATRH is written out in SAVMIN, and read in SUHESS, both called by CVA1.
The input of the second trajectory is the same as in 3D-Var. The output is an analysis at the initial time of the trajectory (type = 4v, step = 0) written out on the FDB. It contains the current estimate of the flow at initial time. Another output are the updated observation files, as in 3D-Var. The 4v fields are used in the following trajectory, replacing the background in the input files ICMSHxxxxINIT, ICMGGxxxxINIT and ICMGGxxxxINIUA (where xxxx is the `expver' identifier of MARS). Additional inputs are low resolution files (MXVA...) created during the previous minimization interpolated to high resolution as in 3D-Var. This data flow is represented in the diagram below.
In summary, the first two trajectories use the background as an input, and the following ones use the 4v fields created during the previous trajectory as reference files. All the trajectories except for the very first one add increments computed from the low-resolution files produced by the previous minimization, interpolated to high resolution. The first minimization uses only the background field, the following ones also use the control variable from the end of the previous minimization and some information for a warm restart of the minimization package.
The number of updates of the trajectory starting from 0 at the first minimization is carried inside the ODB files.
Next Section
Previous Section
 |
20.03.2002 |
 |
© ECMWF |
