FloodnetPot: Estimating flood quantiles using peaks over threshold

In floodnetProject16/floodnetRfa: FloodNet Regional Frequency Analysis

Description

Return the flood quantiles, standard deviation, lower and upper bounds of the confidence interval based on a Peaks Over Threshold (POT) model.

Usage

FloodnetPot(
  x,
  u = NULL,
  area = 59874,
  period = c(2, 5, 10, 20, 50, 100),
  nsim = 2000,
  level = 0.95,
  tol.year = 346,
  out.model = FALSE,
  verbose = TRUE
)

Arguments

x	Hydrometric data. Dataset having three columns: site, date, value.
u	Threshold value provided by the user.
area	Drainage area. Use to determine the minimal separating time between flood events. The default value corresponds to 15 days.
period	Return period for which the flood quantiles are estimated.
nsim	Number of bootstrap samples used for inference.
level	Confidence level.
tol.year	Number of days necessary to consider a year complete. Otherwise the data is removed.
out.model	Logical. Should the model be output. This corresponds to the output of FitPot.
verbose	Logical. Should message and warnings be output.

Details

Estimation is carried out by maximum likelihood using the Generalized Pareto distribution. If not provided, a best candidate threshold is searched. It corresponds to the lowest threshold where the goodness-of-fit test of Anderson-Darling has a p-value greater than 0.25. In addition, the selected threshold is forced to have (in average) fewer than 2.5 peaks per year and at least 30 peaks. If no such threshold is found, the threshold with the maximum p-value is used instead. The declustering technique for extracting the flood peaks follows the recommendations of the Water Ressources council of the United States. In particular, flood peaks must be separated by at least 4 + log(A) days, where A is the drainage area of the basin in kilometers squared. Inference on the estimated flood quantiles is performed by parametric bootstraps.

If verbose == TRUE, a Mann-Kendall test and logistic regression are used to verify the presence of trends in the mean excess and the number of peaks per years. For the logistic regression, a t-test examines the significance of the slope parameter for a linear trend. If the data fails one of the tests at significance level 0.05, a warning is issued.

References

Durocher, M., Zadeh, S. M., Burn, D. H., & Ashkar, F. (2018). Comparison of automatic procedures for selecting flood peaks over threshold based on goodness-of-fit tests. Hydrological Processes, 0(0). https://doi.org/10.1002/hyp.13223

Durocher, M., Burn, D. H., & Ashkar, F. (2019). Comparison of estimation methods for a nonstationary index-flood model in flood frequency analysis using peaks over threshold. Water Resources Research. https://doi.org/10.1029/2019WR025305

See Also

FloodnetAmax, FloodnetPool.

Examples

 ## Performing the analysis
 x <- DemoData('daily')
 fit <- FloodnetPot(x, u = 1000, area = 14700)
 summary(fit)

floodnetProject16/floodnetRfa documentation built on Oct. 24, 2020, 9:19 a.m.