In martindurocher/floodStat: FloodNet Regional Frequency Analysis

UNFINISHED

Introduction

In this document I will show how to use floodnetRfa to perform flood frequency analysis at ungauged sites. For this analysis, the flood quantiles of unknown site is predicted based on a set of known descriptors. Information about various sites located in Canada are available in the dataset canFlood. Here four descriptors are considered: Drainage area, percentage of water bodies, stream density and mean annual precipitation. The descriptors are transformed to logarithm scale and standardized in order to obtain less skew distributions. The coordinates used in this study were obtained by multidimensional scaling to have a set of Cartesian coordinates that preserve geographical distances.

library(floodnetRfa)
data(canFlood)
attach(canFlood)

xdf <- data.frame(q100   = 0,
              area   = scale(log(BASIN_AREA)),
              wb     = scale(log(WB_AREA_PC)),
              stream = scale(log(STREAM_DEN)),
              map    = scale(log(PPTAVG_BAS)),
              lon    = ELON,
              lat    = ELAT)

The quantile regression technique consists to predict at ungauged site the flood quantiles obtained from at-site flood frequency analysis. The L-moments of each site in canFlood are provided and used to evaluate the flood quantile of 100 years return period assuming a generalized extreme value distribution.

library(lmomco)
lmm <- split(as.matrix(canFlood[,12:14]), 1:nrow(canFlood))
lmm <- lapply(lmm, vec2lmom, lscale = FALSE)
para <- lapply(lmm, lmom2par, 'gev')
xdf$q100  <- log(sapply(para, function(z) par2qua(.99,z)))