Configure topology of an IFD environment
Where to Use?
Mandatory for an IFD installation
Why to Use?
The topology.xml is necessary to be able to use panels like the topology panel and the forecast panel
Topology panel is used to define the topology of an IFD environment. Also the behaviour of the forecast panel which is used to start
The concept of the IFD (interactive forecast displays) was introduced in 2010. A central role in the configuration of the IFD is the topology.xml. This part of the wiki explains the configuration details of the topology.xml.
First the background of the topology.xml wil be explained in the overview section. In the following sections the configuration options will be explained in detail.
The topology.xml is a mandatory configuration file when you are setting up an IFD-environment. This configuration file is used to configure the topology of a FEWS-system.
The topology is defined by individual nodes and their connectivity. The topology can be viewed in the topology panel, which shows a block diagram of the topology. An example is shown below.
or in the forecast panel, which shows a tree view of the topology.
The behaviour of the forecast panel can also be configured in the topology-file. This will be explained in more detail in the next sections.
The topology.xml plays a central role in configuring an IFD-environment since it is used to configure the forecast panel which is the central panel in an IFD-environment.
The topology.xml was originally designed to configure the topology of a Delft-FEWS system but there are a lot of Delft-FEWS systems in which the topology.xml is used to configure the work process of the forecaster instead.
An example is given below.
Running workflows from the topology tree
The topology tree can be configured in the following way. The example below shows a simple topology tree with two leaf nodes A and B.
Both of them belong the group node C.
It is possible to run workflows from the topology panel. In the example below some of the node(s) have a workflow configured.
By default all the workflows started from a leaf node are started in IFD mode. The results of an IFD-run are only temporary available in FEWS (until the next restart) and are only available and visible at the operator client which started the run.
IFD runs are typically used to evaluate the effect of modifiers on the results of a forecast.
A workflow which is started from a group node will by default be started as a server run. A server run will run at a FSS when it is started from an operator client. When it is started from a stand alone it will run locally but the results of the run will be available after a restart of the system.
It is possible to change these default settings by using the element localRun.
In the example below the runs from leaf node B will run in server mode and the runs started from the group node C will run in IFD mode.
It is possible to connect nodes in the topology. In the example below the node STKA2 is connected to two previous nodes (SBFQ2 and ISKQ2).
The way nodes are connected plays an important role when running workflows in IFD mode.
The example below shows the situation before the forecaster starts an IFD run for the node STKA2. The topology tree shows that this node has not run yet (icon is a black dot) and that it is connected to two previous nodes SBFQ2 and ISKQ2.
When the forecaster starts an IFD run for node STKA2 Delft-FEWS will detect that node STKA2 is connected to two other nodes which have to run before the run for STKA2 can be started.
One of these nodes (SBFQ2) also has a previous node which has to run prior to running this node. Before running node STKA2 the IFD system will make sure that first all the necessary previous nodes are run in the correct order before node STKA2 will be run. The example below shows the forecast tree after the run for STKA2 was finished. It shows that not only an IFD run was started for STKA2 but also for all its previous nodes.
By default the workflow on a leaf node is started automatically after selecting the node. This can be disabled by setting the option <enableAutoRun> to false.
The option will apply to all nodes.
For IFD runs the connected previous nodes are always started before the IFD run for the selected node can be run.
For server runs the connected previous nodes are by default not taken into account. If the option <enableRunUpstreamServerNodes> is set to true the previous nodes will also be started prior to running the workflow of the selected node.
The option <checkStatusPreviousServerRun> can be used to force that a server run can only be started if the status of all of its previous nodes is fully successful.
Previous nodes which are connected to a node but belong to a different group in the topology tree are ignored by default when starting previous nodes for a selected node. By setting the option <enableCrossGroupNodeReferencing> to true previous nodes in a different group will also be taken into account.
It is possible to define a secondary workflow for a node. This secondary workflow can be run by clicking the buttonin the button bar.
Only the primary workflow determines which icon is shown in the forecast tree.
Below is an example of a node with a secondary workflow.
Run options panel
It is possible to manually select the task properties which will be used in the runs started from the IFD. This can be done in the panel below the forecast tree. The picture belows shows an example of this panel.
The options which are shown in this panel are configurable. Each node or group of nodes in the topology can have different options.
More advanced options are available in the run options panel. This panel is started when the forecaster clicks on the edit run options button.
The "edit run options" button can be disabled by setting the option <disableAdvancedButton> to true. The "edit run options" button can be disabled for a single node or a group node and its children.
In the section below several typical configuration examples are given which will be explained in detail.
Warm state selection
In a lot of systems, for example NWS systems, only the state selection is configured. The time zero will in this case be equal to the system time.
The forecast length will be determined by the forecast length estimator.
Below a screenshot of the system prior to running a node. The state selection is set to -10 days. The forecast length is not set and the time zero is equal to the system time.
After the run the forecast length which was determined by the forecast length estimator in the run is shown in the panel.
The forecaster can adjust the forecast length by using the time chooser.
It is possible to disable the possiblity to edit the forecast length by using the option <useForecastLengthFromInteractiveForecastDisplay>.
An configuration example is shown below.
When the system time changes the selected state will also change so the relative time compared to time zero will stay the same.
Warm state selection period
The option warm state selection period can be used to set the state selection to a period (relative to the system time) in which the system will search for a warm state.
An example of a node which is configured with a warm state search period is given below.
A configuration example is given below.
Cold state selection
It is also possible to set the default state selection for a node or a group of nodes to a cold state selection. This is used, for example, in FEWS system which are used for calibration purposes.
An configuration example is given below.
No initial state
The option <noInitialState> can be used to disable the state selection for a node so that the state will be selected by the workflow.
An example of a node which has this option enabled is shown below.
A configuration example is given below.
Relative period panel
When the forecaster wants to run a workflow for a certain period of time then the relative period panel is a good option to use. The picture belows shows an example of the relative period panel.
This option is used, for example, in the HERMES-system. Below a configuration example:
When the system time is changed then the start time and end time will also change so that the relative period will stay the same for a topology node.
It is possible to force the initial start time and end time to match a certain time step. This can be done by using the options <initialStartTimeCardinalTimeStep> and <initialEndTimeCardinalTimeStep>.
To force user edits to the start and/or end time to a certain time step the options <cardinalTimeStepStartTime> and <cardinalTimeStepEndTime> can be used.
If the option <initialStartTimeCardinalTimeStep> is used then the start time will not change when the system time changes. The same applies for the end time and the option <initialEndTimeCardinalTimeStep>.
A configuration example is shown below.
Relative period panel with only a start time
It is possible to a create a relative period panel in which the end time is hidden.
Below an example of such a panel.
This panel can be configured in the following way.
Cold state from current run
To set the state selection at a node to the cold state selection which was used in another run the option <coldStateFromCurrentRun> can be used.
Below a configuration example
In the example above the cold state which was used in the current run from workflow "Tweed URBS IFD forecast" will be used.
It is possible to define a default state selection for the cases that a current run is not available yet. In the example above this default is set to a cold state selection of 2 days before time zero.
If time zero changes the cold state selection should also change to keep the relative period the same.
To set the forecast length at a fixed length the option <forecastLength> can be used.
It is possible to force the end time of the forecast (time zero plus forecast length) to match a certain time step by using the option <initialForecastLengthCardinalTimeStep>.
The end time of a node is only forced to match this time step when the user selects the node for the first time.
To force the end time of a node to match a certain time step when the forecaster tries the edit the default forecast length the option <cardinalTimeStepForecastLength> can be used.
By default the end time of a node changes when the system time changes so that the forecast length stays the same.
However if for a node the option <initialForecastLengthCardinalTimeStep> is configured then the end time of the node stays the same if the time zero changes.
A configuration example of a node for which a forecast length, an <initialForecastLengthCardinalTimeStep> and a <cardinalTimeStepForecastLength> is configured is shown below.
Below a screenshot of this node after it is selected for the first time is shown.
The end time of the run (forecast length) was first set to time zero plus 24 hours. Because this time wasn't a valid time for the configured
<initialForecastLengthCardinalTimeStep> the end time was set to first valid time prior to the end time.
Below a screenshot of the same node after the sytem time was changed. The time zero is changed but the end time of the node is still fixed at the same time.
By default the time zero of a node is equal to the system time and not editable.
It is possible to make the time zero editable and to set it to a fixed date/time.
The code snippet below shows a configuration example.
To make the time zero editable and/or to shfit the time zero compared to the system time the option <timeZeroShift> can be used.
If the shift is set to 0 then the time zero will be editable but equal to the system time.
Custom info label
It is possible to define a custom info label which be shown (if configured) below the run options.
It is possible to use location attribute in the custom info label. if a value is modified it will be colored blue.
Below a configuration example.
It a date/time attribute is not modified the default value will be shown. It is also possible to configure an overruling value if the date/time attribute is unmodfied.
This can be done by using the overrulingUnmodifiedDateTimeAttributeValue. The value can be configured by defining an offset and a time step.
The overruling value will be determined by adding the configured offset to the time zero of the node and setting the determined time to the first valid time of the defined time step.
You can configure the text and background color(selected and non-selected). These are optional fields, you can configure all, none, or some. These properties can be configured on individual nodes and group nodes. Feature available since 2019.02.
Approve forecast automatically
When the option enableAutoApprove is not used the IFD will use the logic described below.
When a run is started from the IFD it is by default automically approved. Only when the configurator defines in the workflow descriptors that the workflow should not be automically approved the forecast is not automically approved.
When the config option enableAutoApprove is used a checkbox will appear on the run options panel. With this checkbox the forecaster can determine if the forecast should be approved automically after the run.
The value of the tag enableAutoApprove will determine if the checkbox is initially selected or not.
It is possible to split the configuration of the topology in separate configuration files.
The main configuration file is always the topology.xml. The other (supporting) configuration files are the topology group files. The topology group files should be placed in the RegionConfigfiles folder. The name of a topology group file should start with TopologyGroup.
An example of a topology group file is shown below.
The defined topology groups can be used in the topology.xml by using the element <groupId>. An example is shown below.
Interaction with other displays
The forecast panel interacts with the other panels in Delft-FEWS in a lot of ways.
It is possible to define which panel or tool window(s) should be made visible after selecting a node.
This can be done using the elements <mainPanel> and <toolWindow>.
An example is shown below.
After selecting a node in the topology a plot connected to this node will be automatically displayed in the plot window.
The element <displayGroupId> can be used to connect a topology node to a display group. After selecting a topology node Delft-FEWS will first try to find a display which has the same location and parameter as currently displayed in the connected display group. If such a display cannot be found the first display of the topology group will be shown.
This behaviour can be disabled by using the element <selectFirstPlotOnSelectionChange>. If this option is set to true the first plot will always be selected.
The plot overview panel shows an overview of all the displays in the display group.
To be added by Onno...timeSeriesButtonPanelId
To automatically select a zoom extent in the map after selecting a node, the element <mapExtentId> can be used.
To automatically select a filter after selecting a node the element <filterId> can be used. By default the parameters in the filters will not be selected automatically. If the parameters selected by the user still are present in the new filter, they will remain selected. To enable that all the parameters will also be selected automatically the option <enableAutoSelectParameters> should be set to true. To enable that no parameters will be selected, i.e., the parameter selection will be automatically cleared, the option <enableAutoClearParameters> should be set to true (available since 2018.02).
The element <forecasterHelperDirectories> is a configuration option which applies to all nodes. It can be used to define in which directories the forecaster helper should look for files. See for more info: 23 Interactive Forecasting Displays#23InteractiveForecastingDisplays-Forecasterhelp(akaDocumentViewer)
The element <enableSelectNodesFromMap> is a configuration option which applies to all nodes. With this option set to true, the user can select nodes in the topology tree by selecting locations on the map. For this to work, you need to also configure at least one <locationId> and a <filterId> at each node which you would like to be able to select from a map. If the user first selects the configured filter in the Data Viewer, and then next selects (one of) the location(s) on the map, the corresponding node in the topology will be opened and the IFD will get the focus. It is possible to configure one or more locationIds for a node by using the element <locationId>.
An <areaId> can be configured for a topology node. This configuration option is used in combination with the archive panel. After selecting a node with an area id configured the area selection in the archive panel will be set to the area id of the selected node.
By default, the <ForecasterNotesDisplay> shows the notes for the the node (and its parent) that is selected by the user in the Topology GUI. This list of forecast notes can be extended with the notes that are created for the node with option <alwaysVisibleInForecasterNotes> is set to true.
The option <url> can be used to configure an URL for each node. After selecting the node the URL will be displayed automatically in the web browser display.
The icons of the forecast tree are explained in detail in this chapter 23 Interactive Forecasting Displays#23InteractiveForecastingDisplays-Forecasttree
The icons in the topology tree can be used to view the status of node. The icon show if a workflow has run for that node and what status of that run is.
A green or blue icon indicate that the workflow was succesfull, The icons also indicate if the time zero which was used in the last run of a node is the same as the time zero which is currently selected for a node.
If the time zero is different then the icon will indicate that the available run is invalid. In systems which have a small cardinal time step this behaviour can be problematic because runs are marked as being invalid too quickly.
The element <graceTime> can be used to introduce a grace time for the time zero was used in a run. If the difference between the time zero for a node and the time zero which was used in the last run is less then the grace time then the run will still be marked as valid.
If the status of a topology node is determined by a server run and its parent node has a more recent server run then the node will inherit the status from its parent.
The option is only available at group nodes. By default the modifiers panel doesn't show modifiers at group nodes. This choice was made because of performance reasons.
In addition it is not common that modifiers are made at group nodes. In the cases that it is required that modifiers can be shown (and edited) at a group node the default behavour can be
changed. This option is only available at group nodes.
The modifiers panel consists of two parts. An overview panel which shows all the modifiers which are available for the selected node and a detail panel which shows the details of the modifier which is selected in the overview panel or the new modifier which the user is creating.
The overview panel uses the half of the space of the modifiers panel. By hiding this panel a lot of space is available for the detail panel. This is a useful feature when users only create modifiers at a node and want to use all the available space for the detail panel.
This option can be used to disable the possibility to add, edit or delete modifiers at a node.
If a default modifier id is defined at a node then the first available modifier of the defined type will be selected in the modifiers panel. If there is no modifier of this type is available then a new modifier template is created in the modifiers detail panel. An example is shown below.
In the example above a node "Run all Models" is shown. This node has a defaultModifierId defined and set to URBS.
After selecting the node "Run all models" a modifiers template for the modifiers type "URBS" will be created. The user can edit this template and create a new modifier by pressing apply.
Below an another example:
In this example the user has selected the same node. The difference with the previous example is that there is already a modifier of type URBS available.
Because this modifier is available it will be selected instead of creating a new modifiers template.
If this option is configured only the defined default modifier is editable.
This option defines the group of modifiers which is visible and editable at a node. It is possible to define multiple visible modifier groups for a node. By default all modifiers are visible and editable.
The modifier group should be defined in the modifierTypes.xml.
There are several configurations options available with regard to permissions.
This permission controls who can view this node
Permission to run the workflow for a node locally
Permission to run the workflow for a node at the server
Since 201801. Permission to run the secondary workflow
Details about the buttons can be found here: 23 Interactive Forecasting Displays#23InteractiveForecastingDisplays-Buttonbar
At the top of the forecast panel a group of buttons is shown.
The first Delft-FEWS applications which used the forecast panel had a button panel which had 6 buttons.
An example is given below.
The reduce the number of buttons which were shown in this panel a new design for this button panel was made.
This design offered the same functionality but had less buttons.
An example is given below.
By default the button panel which is showed above is available.
To switch to the button panel with more buttons the element <enableOriginalButtons> can be used.
An config example is shown below.
The element <saveLocalRunEnabled> can be used to allow the user to promote a local IFD run to a permanent server run.
If this element is set to true for the selected node and an IFD run is available for the selected node the save-button in the button bar will be enabled.
The element <showRunApprovedForecastButton> can be used to enable the user to start server runs from nodes for which the option <localRun> is set to true.
This means that users can start IFD-runs from that node but also server runs.
The element <showMacroButton> can be used to enable a button in the Run options dialog which allows the user to start a run with pre-defined task properties.