How to set up a Sobek RURAL Model in FEWS

How to set up a Sobek-RURAL Model in FEWS

Needed:

• installation of SOBEK (incl license) in the root of the same partition of your hard disk (e.g. c: or d: ) as where FEWS is run.
• SOBEK project (*.lit folder) in /Modules/SOBEK/[application name]/ *.lit
• SOBEK adapter: /Modules/SOBEK/[application name]/ SOBEK_Adapter.xml
• general adapter file: Config/ModuleConfigFiles/SOBEK_Forecast and SOBEK_Historical

Settings inside the SOBEK case

Simulation period and calculation time step
The simulation period needs to be triggered by the meteorological data. When running SOBEK from the User Interface there is a meteo-event file used. This file is called *.bui and is located by default in the SOBEK folder \SOBEK211\FIXED. The meteo event has a time step which will influence the calculation time step in the case. It is recommended to make this time step in the meteo event file larger than the desired calculation time step, to avoid SOBEK resetting your time step.

When running SOBEK under FEWS there is also a meteo event file needed. This file (e.g. Rainfall.xml) is exported by the general adapter-file. So even if there is no rainfall needed for running the SOBEK model, it needs to be exported for determining the meteo event and thus the simulation period.

Model state files
FEWS propagates the model state of the SOBEK model with the historical runs. Every next run is started from the last known model state, a warm state. If the model state is disregarded this means a cold state start is made. In both cases the model state is read from so called 'restart-files'. The SOBEK case needs to be instructed to write and read the same restart-files in the settings under the tab 'initial data'. You will receive a warning that the restart files have the same name and will be overwritten.
In the project folder (xxxx.lit) the restart files 1DFLOW.RDA, 1DFLOW.RDF and 1DFLOW.RDS can be found in the case, the folder starts with a number. In the folder xxxx.lit/FIXED the 2D restart files 2DFLOW.FDS and 2DFLOW.FLS can be found. These restart files all need to be copied into the ColdStateFiles/SOBEK_Historical default.xml.

Water Quality calculation
In case of using any WAQ module, ensure the option for using the previous flow results is switched off in the SETTINGS task.

Copying the Project folder
Before copying the project folder from the SOBEK program to the FEWS Modules directory, you can reduce the size of the complete case directory by removing the original results

1. run the simulation task
2. cancel the simulation once started
3. press [Show permanent] button on error message box
4. delete all next files in work directory of the case
   1. out;.his;.map;.stu;.pst;.hia;.log;.prn;.hia;.inc;.asc;.mon;.lst
   2. 2DAREA.DAT
   3. 2DFLOW.DAT
   4. 2DSOBEK.CHZ
   5. 2DVELOC.DAT
   6. 2DVOLUME.DAT
   7. 2DPOINT.DAT
5. remove rubish data from the project directory
   1. boundary subdir
   2. data subdir
   3. initial S subdir
   4. Network subdir
   5. Many files in fixed subdir
   6. Many files in Bathymetry subdir

The case must be saved before copying with at least the [Schematisation] task green.

Schematisation
Note that the boundaries, laterals and controllers which data is being updated by FEWS should already been defined as time tables, both for Flow and WAQ modules!**

Output
To make sure the amount of output produced by SOBEK is no more than needed, it is important to uncheck all unnecessary output creation. If only water depth is needed in 2D then only have this option checked.

Settings inside the SOBEK module directory

Inside the module directory SOBEK needs to have a /bin folder with SBKbatch.exe and SBKbatch.ini . Inside the other directory the folders Inputs, Logs, project folder copied from SOBEK, Output and States are needed. In the root of this directory the SOBEK_Adapter.xml is found.

SOBEK_Adapter.xml

In this file the adapter for SOBEK is configured.

General

The file describes in the 'general' section which project to use and which case to use inside the project.

NB: The elements parsenMinimumTimeSeriesLength, parsenSourceFiles are only required in case useParsenUpdateMode is used. So use either all three, or none of them. In case you have useParsenUpdateMode = TRUE, the PARSEN (SOBEK 1D Flow pre-processor) will not completely generate the NEFIS files sobek.mda and sobek.mdf, but will only update the time series in the file. This improves the performance for large networks.

postProcessingActivities

In the 'postProcessingActivities' the 2D map outputs from SOBEK are moved Output folder and the restart files written to the States folder. Please note that whether they are read by FEWS depends on the generalAdapter file.

parameters

In 'parameters' it is described which timeSeries need to be translated to which SOBEK files and how they should be interpreted.
Notice that updating of time series in SOBEK files should be defined as follows, where the description is very important:

• comment: has no meaning, can be any string just for your reference
• description: Description of sub parameter, previously used as Sobek object ID
• sobekObjectId,  like laternal node ID, boundary node ID, controller ID
  If not defined the description element is used (as it was in the older days)
• locationID : is the location ID as used in the PI XML time series file
• parameterID : is the parameter ID as used in the PI XML time series file
• type: type of data or time series. See the table below to select the type of parameter to select.

Please note that the rainfall, evaporation and wind time series should be all in one separate single XML file, so one for all rainfall series, one for all evaporation series and one for all wind series (speed and direction both in one file).
Rainfall series are always converted independent on the parameterID, where all locationIDs are converted to meteo stations. Evaporation works the same. Regarding wind series, the adapter looks for the following time series:

• there should be at least two time series, one containing the wind speed and one with the wind direction
• if a time series parameter ID contains the string "speed" (not case sensitive), this series is used as a wind speed series
• if a time series parameter ID contains the string "dir" (not case sensitive), this series is used as a wind direction series
• the location IDs are used to identify different wind fields
• the first defined wind field (locationID) is always used as global wind field.
• in case more than one different wind fields (location IDs) all series are also written as additional wind fields, next to the global wind field.
• The resulting wind file is always converted to the corresponding WQ Wind file (*.qwt), but only containing the global wind field.

parameter types

Flow_dam and Flow_dambreak should be triggered by time series with non-missing values. The flow_dam should have timeseries where the location id is equal to the reachsegment id. The adapter creates the weir/dam at this segment. The time series values represent the height of the weir. This is automatically added to the bed level to get a correct crest level of the weir.

The dambreaks can be created at all node types that are calculation points (fixed and regular calculation points, connection nodes and linkage nodes). The time series that trigger a dambreak should contain by default missing values, but a -1 for a dambreak at the left side, a +1 for a dambreak at the right side. The adapter creates a new reach to represents the dambreak. At this reach a weir structure is placed, with a crest level that will be lowered at the start of the dambreak.

Notice that using dams or dam breaks will make that the restart files do not fit anymore with the network and will be neglected.

The flow_lateral_dat_file has some special features. This routine reads the network.cn file to know the type of the laterals (FLNO, FLBR, FLDI) and then creates a complete new lateral.dat file, based on the series PI-XML file. In case a time series is not in the XML file, the original definition is re-used. However, this method is not very efficient so this whole flow_lateral_dat_file routine is most efficient in case you convert ALL the laterals. The locationID is used as the required lateral id. The parameterId is not used, although you can use it to parse the interpolation option to the definition. Therefore include the string "block" in to the parameterId, so the routine knows that the block interpolation option should be used. Default interpolation option is linear. It is of course (in operational systems) not possible to define a periodicity.

Jobs

In the 'jobs' section nothing needs to be changed.
It should be:

<jobs>
  <job>
    <jobName>Sobek historical run</jobName>
    <storeDir/>
    <parameter>
      <classIndex>1</classIndex>
    </parameter>
  </job>
</jobs>

Output

In the 'output' section the output locations need to be identified. Moreover the SOBEK_Adapter needs to know in which file to read the output and to which PI-file to write the output so that the generalAdapter can read it.

mapStack

In the 'mapStack' part the 2D maps are put together in a so called map stack. The generalAdapter will import these map stacks as a grid timeSeries.

Example:

<mapStack>
    <fileName>output\dm1dmapstack.xml</fileName>
    <fileMask>dm1d????.asc</fileMask>
    <geoDatum>Rijks Driehoekstelsel</geoDatum>
    <locationId>map</locationId>
    <parameterId>InundationDepth</parameterId>
    <timeStep>unit="hour" multiplier="1"</timeStep>
    <module>2dflow</module>
    <mapStackActive>true</mapStackActive>
  </mapStack>

States

The last section in the SOBEK_Adapter.xml file is 'states'. This describes the read locations and writelocations of the model states. This is similar to other external modules.
RR states: * rrr
1D states: *.rda, *.rdf
2D states: *.fls
WQ states: *.res

Example:

<states>
  <stateXmlOutputFile>States\Outputstates.xml</stateXmlOutputFile>
  <state>
    <readLocation>States\Input.rda</readLocation>
    <writeLocation>States\Output.rda</writeLocation>
  </state>
  <state>
    <readLocation>States\Input.rdf</readLocation>
    <writeLocation>States\Output.rdf</writeLocation>
  </state>
  <state>
    <readLocation>States\Input.fls</readLocation>
    <writeLocation>States\Output.fls</writeLocation>
  </state>
  <state>
    <readLocation>States\1Dlevels-in.xyz</readLocation>
    <writeLocation>States\1Dlevels-out.xyz</writeLocation>
  </state>
  <state>
    <readLocation>States\Input.rrr</readLocation>
    <writeLocation>States\Output.rrr</writeLocation>
  </state>
</states>

Longitudinal profiles

GeneralAdapter

The generalAdapter is similar to other generalAdapter files. For the exact syntax inside the executeActivities part of the file always copy an existing SOBEK generalAdapter. A special point of attention is that there should always be a meteo time series exported to SOBEK. This file will determine the start and end time of the run.

Please make sure the GA run is configured to set the missing value definition in the XML files it exports to -999.99 and not the default NaN. This should be set in general section, please see the example below. If NaN is used as missing value definition the SOBEK adapter will complain about missing value when there is 0.0 precipitation.

<general>
  <rootDir>$REGION_HOME$/Modules/Sobek/Singapore</rootDir>
  <workDir>%ROOT_DIR%</workDir>
  <exportDir>%ROOT_DIR%/Input</exportDir>
  <exportIdMap>IdExportSobek</exportIdMap>
  <importDir>%ROOT_DIR%/Output</importDir>
  <importIdMap>IdImportSobekForecast</importIdMap>
  <dumpFileDir>$GA_DUMPFILEDIR$</dumpFileDir>
  <dumpDir>%ROOT_DIR%</dumpDir>
  <diagnosticFile>%ROOT_DIR%/Logs/diagnostics.xml</diagnosticFile>
  <missVal>-999.99</missVal>
  <convertDatum>true</convertDatum>
  <timeZone>
    <timeZoneOffset>+08:00</timeZoneOffset>
  </timeZone>
</general>

Supported input parameters

The SOBEK adapter supports the following input parameters:

With respect to the last three parameters: these are used as timeseries for a controller that supports timeseries. The location ID should be the name of the controller and the parameter ID one of the three names given in the table. Only controllers are currently supported by the adapter.

How to run a model in Ensemble mode?

The default method for running in ensemble mode is that all runs are performed sequentially. There are two alternatives to running the models simulatiously to reduce the overall time:

1. run the model in a directory per ensemble member with default settings (SbkBatch uses the Case Management)
2. run the model in a temporary directory without using the Sobek Case Management.

The first option is the preferred one as it guarantees that all Case Management is correct. However, it requires quite a lot of configuration as all the possible workingdirectories have to be pre-configured. Instead of one, for all the ensemble members a working directory should be prepared. In the General Adapter you should link to these directories by using
<rootDir>$REGION_HOME$/Modules/Sobek/Singapore_%ENSEMBLE_MEMBER_INDEX%</rootDir>.
See also 19 Parallel running of ensemble loops.

The second option is more advanced but faster and especially useful for small models where the CMT activities are a significant part of the simulation time. Next to that the configuration is much easier as only one temp directory has to be specified.
In the General Adapter you specify only <rootDir>%TEMP_DIR%</rootDir> where the TEMP_DIR will be defined as the system TEMP_DIR (or defined in the global.properties) and an added unique subdirectory. In case your system TEMP_DIR variable is "c:\windows\Temp" the General Adapter will use as TEMP_DIR in each particular instance of the General Adapter (i.e. per ensemble member) TEMP_DIR = "C:\windows\temp\<FSS><ModuleInstance><ENSEMBLE_MEMBER_INDEX>_<threadId>."
In this configuration it is quite useful to have no CMT activities anymore. Therefore define in the SBKBATCH.INI the key [General]RunFromTempDirWithoutCMT=-1. Then you need to configure:

• a CMTWORK and WORK directory where all the files are available already
• FNM files in the CMTWORK that has only relative paths or links to files that are not specific to the actual run (like language files)
• use as [General]SimulateCall and [General]SimulateCallRtnFile directly the required program or a batch file that runs multiple programs. For example if you need to run only the RR module, you define:

  [General}
  SimulateCall=\Sobek212\Programs\3b\sobek_3b.exe sobek_3b.fnm sobek_3b.rtn
  SimulateCallRtnFile=sobek_3b.rtn

• save the whole setup in a moduleDataSet that will be exported before starting the simulation

Notice that this functionality is available only since SbkBatch.exe version 2.1.0.112

Problems/Solutions

SbkBatch seems to hang at startup and use a lot of cpu power

This may happen if sbkbatch cannot write it's diagnostics file. Check if the directory is present and if it is readable by the adapter. The location of the diagnostics file is set in the general section of the xml configuration file:

<fewsDiagnosticsFile>Logs\Diagnostics.xml</fewsDiagnosticsFile>

This problem has been solved since July 2008.