...
This display provides an overview of all threshold crossings in the observed and forecast time series. By matching observed and forecast thresholds, various skill scores can be computed.
Background
The number The first step is to set up a contingency table. Given criteria on time intervals to consider when matching thresholdsthreshold crossing events, the values of a , b , c and d in the table below can be filled in (where e.g. a is the number of matched observed and forecast thresholdsthreshold crossing events).
Once the contingency table has been filled, different skill scores can be established:
False Alarm Rate
Probability of Detection
Critical Success Index
Critical Reliability (checks if a forecast is available at least some time (but not too late) before the observed threshold crossing)
The First Forecast of Threshold (FFOT) is determined as the average time between the T~0~ of the forecast run in which a threshold crossing was detected and the time of the observed threshold crossing (i.e. the average lead time of the category a thresholds threshold crossing events in the contingency table).
The Bias of paired thresholds threshold crossings is the average time between paired observed and forecast events.
Setting Criteria for Analyses
| Wiki Markup |
|---|
To specificspecify what matching threshold crossings mean, a number of criteria can be used. When clicking the \[Change Criteria\] button the following display appears... |
The criteria in the display have the following meaning...
- Min/Max T~0~ difference . These criteria define in what time interval the T0 of a forecast should fall that has predicted a threshold crossing in order for this threshold crossing event to be included in the analysis.
...
In the Matching Events tab different background colours are used to indicate in which sector of the Contingency Table events fall...
- green: observed and forecast threshold crossing event match (sector a of contingency table)
- yellow: a threshold crossing has been observed but not forecasted (sector c)
- orange: a threshold crossing has been forecasted but not observed (sector b)
...
In general, skill scores are determined for all forecast locations being at a gauging station and having level thresholds. Typically these are determined separately, but the structure of the display allows skill scores to be established for several modelsdifferent models. This way the skill scores of the different models can be compared.
The performance indicators that are computed on the basis of the selection in this tab are...:
- probability of detection
- false alarm rate
- critical success factor
- first forecast of threshold
- bias of paired thresholds
...
In the Forecast Available for Events tab different background colours are used to indicate in which sector of the Contingency Table events fall...
- green: a forecast is available for an observed threshold crossing event (sector e of contingency table)
- yellow: no forecast available for an observed threshold crossing event (sector f)
...
Archiving Events
| Wiki Markup |
|---|
As the thresholdsthreshold crossing events are stored in the operational database for only some days (according to the length of the Rolling Barrel), it's possible to manually export the list of thresholdsthreshold crossing events by pressing the \[*Export* \] button in the skill scores display. This list can later be imported in a Stand Alone system by pressing the \[*Import* \] button. In this way, longer records can be established for analysis in an off-line modus. |
| Wiki Markup |
|---|
Instead of exporting the whole list of threshold thresholdscrossing events, as described above, it's also possible to select some of them and export only thisthe selected ones by using the \[*Save* \] button. |
...