FloodnetRoi: Prediction of flood quantiles at ungauged sites using Region...
In floodnetProject16/floodnetRfa: FloodNet Regional Frequency Analysis

Description Usage Arguments Details References Examples

Return the flood quantiles, root mean square error, relative bias, lower and upper bound of the 95 local regression and kriging technique.

FloodnetRoi(
  x,
  sites,
  target,
  sites.coord = NULL,
  target.coord = NULL,
  size = 20,
  period = 100,
  distr = "gev",
  corr = 0,
  nsim = 0,
  out.model = FALSE,
  verbose = TRUE
)

`x`	Dataset containing the hydrometric data.
`sites`	Descriptors of gauged sites. A data.frame where the first column is the name of the basin.
`target`	Descriptors of the target sites. A data.frame where the first column is the name of the basin.
`sites.coord`	Coordinates of the gauged sites.
`target.coord`	Coordinates of the target sites.
`size`	Size of the region of influence.
`period`	Return period to predict. Must be a single value.
`distr`	Distribution of the gauged sites. Can a common distribution or a vector of the individual distributions.
`corr`	Correlation matrix or coefficient that represent the intersite correlation. If no specified the average coefficient of correlation is used between for all pairs.
`nsim`	Number of bootstrap samples.
`out.model`	Logical. Should the model be returned.
`verbose`	Should progress and message be display during the procedure.

The function uses a quantile regression technique (QRT) to predict flood quantiles of a given return period at ungauged sites. The function first perform at-site flood frequency analysis of the annual maximum discharges extracted from the HYDAT database (or provided data). At-site estimates of the flood quantile are evaluated by the L-moments method. The resulting flood quantiles are fed to the QRT model to predict the flood quantiles at ungauged sites according to its basin characteristics. The QRT model uses a local regression method to evaluate flood quantiles at specific target locations. If coordinates are provided, simple kriging is additionally used to improve the prediction by extracting further information from the spatially correlated residuals.

If the size of the region of influence (ROI) is not provided, a value is automatically selected. If the dataset include fewer than 30 sites, all sites will beincluded by default, otherwise ROI sizes of 20 and more are tried by steps of 5. The one associated with the lowest mean absolute prediction error (based on leave-one-out cross-validation) is finally selected

Please note that the distance between sites is the Euclidean distance and that proper projection of the coordinates is required for input.

A mix of parametric bootstraps for the at-site distributions and residual bootstraps for the QRT model is used to evaluate model uncertainty.

Durocher, M., Burn, D. H., & Mostofi Zadeh, S. (2018). A nationwide regional flood frequency analysis at ungauged sites using ROI/GLS with copulas and super regions. Journal of Hydrology, 567, 191–202. https://doi.org/10.1016/j.jhydrol.2018.10.011

Durocher, M., Burn, D. H., Zadeh, S. M., & Ashkar, F. (2019). Estimating flood quantiles at ungauged sites using nonparametric regression methods with spatial components. Hydrological Sciences Journal, 64(9), 1056–1070. https://doi.org/10.1080/02626667.2019.1620952

## Not run: 
	## Path the HYDAT database
 db <- DB_HYDAT

## Extract catchment descriptors
xd <- with(descriptors,
 data.frame(
 	site = station,
   area = log(area),
   map  = log(map_ws),
   wb   = log(.01 + wb),
   stream = log(.01 + stream),
 	elev = elev_ws,
 	slope = log(.01 + slope)))

## Put the target site apart
target.id <- (xd$site == '01AF009')

 target <- xd[target.id,]
 xd <- xd[-target.id,]

 ## Fit the model
 FloodnetRoi(target = target, sites = xd, db = db,
					period = 100, size = 30, nsim = 30)


## End(Not run)