• The purpose of this presentation is to guide the user throughout the steps of using the Graphical User Interface (GUI) of the R package floodnetRfa.

• The GUI allows performing flood frequency analysis of hydrometric stations using local and regional methods.

• Utilities for predicting flood quantiles at ungauged basins are only available from the R terminal.

• The GUI runs inside a web browser.
• It is normally launched from the R terminal.
• A script can be used to launch the GUI from the system terminal.

library(floodnetRfa)
FloodnetApp()

# Create a shortcut on the desktop
FloodnetApp(shortcut = "C:/Users/JohnDoe/Desktop/FloodnetApp.cmd")

• For the rest of the presentation, we will discuss the following points.
  • Guideline summary
  • Input data
  • Configuring models
  • Managing models
  • Exporting results

• The principal method in flood frequency analysis consists to fit a distribution on the sample of annual maximum streamflow.

• Estimation is performed by the method of linear moments (L-moments).

• Dispersion and confidence intervals are evaluated using parametric bootstrap.

• A second approach is to fit a distribution of streamflow peaks using the Generalized Pareto distribution (GPA).

• Peaks are exceedances extracted from a declustering method based on a threshold and the basin drainage area

• Estimation is done using the method of maximum likelihood.

• Dispersion and confidence intervals are evaluated using parametric bootstrap.

• RFA used the index-flood model, which assumes that all distributions inside a homogenous region are proportional.

• The sample mean is used as a scaling factor.

• Estimation is done by averaging the L-moments of the scaled sample.

• Dispersion and confidence intervals are evaluated using parametric bootstrap.

• An homogenous region is neighborhoods (pooling group) formed around a target site.

• It used a similarity measure based on the timing and regularity of the annual maximum discharge.

• The homogeneity is assessed by a heterogeneity measure based on the Linear Coefficient of Variation (LCV).

• Heterogeneous sites are sequentially removed from an initial pooling group.

For more details on the methodologies, please consult the Floodnet guideline or the documentation of the following R functions.

• FloodnetAmax
• FloodnetPot
• FloodnetPool

• This image present an overlook of the model screen that is used for configuring and managing models.

• The GUI may need two sources of data.
  • Hydrometric data.
  • Station data.
• Hydrometric data
  • Either a CSV file or an HYDAT database (sqlite3)
  • The CSV file must have 3 columns: Station, Date, Value.
  • Station is a unique identifier (alphanumeric).
  • The date must be in a valid format, e.g. '2020-07-15'
  • Depending on the analysis, the values can be annual maxima or daily discharges.

• Station data
  • Contains meta information about the stations.
  • Must be a CSV file.
  • Each row represents a station with multiple attributes in column.
• List of attributes
  • Station ID. Same as in hydrometric data
  • Drainage area in \(km^2\).
  • Mean annual precipitation (MAP) in \(mm\).
  • Super Region.
  • Threshold.

• Use the two buttons to select the input data.

• An explorer window appears.

• After a file is selected the name appears below.

The model configuration panel is used to configure the desired model. Once configured, press the Fit button.

• Model ID: An identifier of the model. It must be unique and meaningful.

• Method: Method used to perform frequency analysis (e.g. AMAX, POT, RFA).

• Target Site: Station ID of the site of interest.

• Return period: List of return periods used to evaluate the flood quantiles. Must be separated by a comma.

• For AMAX, a distribution is selected among four choices:
  • Generalized Extreme Values(GEV)
  • Generalized Logistic (GLO)
  • Generalized Normal (GNO)
  • Pearson type III (PE3)
• The best distribution can also be selected automatically using the AIC.

• Select the distribution using the drop-down list

• For POT, the distribution is known but the threshold and the target drainage area must be passed.

• The automatic procedure selects the lowest threshold that respects the modeling assumptions.

• For RFA, the user must select a Super Region that identifies a class of stations similar to the target site.

• Pooling groups include only stations inside the target super region.

• The super region is associated with a column of the station data file.

• RFA can be applied to both AMAX and POT approaches.

• RFA POT requires that the threshold are known for all sites in the pooling group.

• A panel of options is available to control the fitting process.

• Confidence Level: Probability coverage of the confidence intervals.
• Simulations: Size of the bootstrap samples.
• Heterogeneity: Level of heterogeneity measure above which a pooling group is considered as homogenous.
• Pooling group: Initial size of the pooling group.
• Intersite Correlation: Coefficient of correlation used to simulate a bootstrap sample with RFA.
• Graphical Theme: Style of the output graphics.

• After pressing the Fit button, the graphics of the return levels (flood quantiles) versus the return period is produced for assessing the quality of the fit.

• The flood quantiles are presented in the form of a table.

The standard error (se) and the confidence intervals (lower, upper) are also available.

After pressing the Fit button, the fitted model is added to a list, where each entry is identified by its Model ID.

• One or more models can be selected by clicking the list.
• Selected models can be removed from the list by using the Remove button.
• Additional results can be displayed by clicking on the Show button.

• The list of fitted models can be saved using the Save buttons.
• A file explorer will open and allow the user to select the file name and location.
• The saved models are in an R format with extension .Rdata.
• The Open button load previously saved models.
• The Reset button empties the current list of models.

• Here is an overview of the result screen displayed by the Show button.

• If multiple models were selected, the display panel allows switching between them.

• The <-Back button allows going back the the model screen

• Some graphics are specific to a model, some include mutiple models.

• For example, the comparison of the confidence intervals allows visualizing the difference the estimated flood quantiles and it variability.

• Some graphics provides information on the selected pooling groups.

• For example, the L-ratio diagram shows the dispersion of 3rd and 4th L-moment among the pooling group.

• The Export button saved the element of the screen in a PDF.
• The button will open a file explorer that allows to select the file name and location.

• Specific elements of the result screen can be selected.
• The flood quantiles and the parameters can also be exported in CSV.