Changes between Version 23 and Version 24 of GeneralImplementationConcept

Timestamp:

May 17, 2025, 10:09:37 AM (25 hours ago)

Author:

lnerger

Comment:

--

GeneralImplementationConcept

@@ -17 +17 @@
 [[PageOutline(2-3,Contents of this page)]]
 
-PDAF bases on the logical separation of the assimilation system in 3 parts. These are depicted in figure 1.
+Data assimilation requires 3 components. These are
+ * **Model**:[[BR]] The numerical model provides the initialization and integration of all model fields. It defines the dynamics of the system that is simulated.
+ * **Observations**:[[BR]] The observations of the system provide additional information. 
+ * **Assimilation method**:[[BR]] The assimilation method combines the model and observational information. 
+
+Figure 1 shows how these components are combined by PDAF.
 
 [[Image(//pics/da_structure.png)]]
-[[BR]]'''Figure 1:''' Logical structure of an assimilation system.
+[[BR]]'''Figure 1:''' Components required for data assimilation system. The data assimilation methods ('DA method') are provided by PDAF's core library. The handling of observations is supported by PDAF-OMI. 
 
-The parts of the assimilation system are
- * **Model**:[[BR]] The numerical model provides the initialization and integration of all model fields. It defines the dynamics of the system that is simulated.
- * **Observations**:[[BR]] The observations of the system provide additional information. 
- * **Assimilation method**:[[BR]] The assimialtion methods combine the model and observational information. 
 
 Generally, all three components are independent. In particular, the assimilation methods are implemented in the core part of PDAF. To combine the model and observational information one has to define the relation of the observations to the model fields (For example, model fields might be directly observed or the observed quantities are more complex functions of the model fields. In addition, the observations might be at different locations than the model grid points so that interpolation is required.) In addition one has to fill the state vector that is using in the DA methods with the model fields and write the analysis state vector back to the model fields. These relations and operations are implemented in separate routines that are supplied to the assimilation system by the user. These routines are called through a well-defined standard interface. To ease the implementation complexity, these user-defined routines can be implemented like routines of the model code. Thus, if a user has experience with the model, it should be rather easy to extend it by the routines required for the assimilation system.