In PhilippBuehler/FloodR: Separation and typing of flood events and typewise staistical return periods (TMPS)

FloodR

Description

This package provides tools for:

Separation of flood events from discharge timeseries
Spatial interpolation of precipitation
Separation of storm events from precipitation timeseries
Flood event typology
Typing of rain events by TQ-Value
Typing of snow events by clustering
TMPS: Floodtype-based Mixture Model of Partial Duration Series
typewise statistical estimate of return periods
typewise statistical estimate of quantiles

Installation and usage

devtools::install_github(repo = "PhilippBuehler/FloodR")

library("FloodR")

How to cite

Buehler, P., & Fischer, S. (2022). FloodR - type-based flood sttaistics. DOI: 10.5281/zenodo.10717767

Separation of flood events

For the separation of flood events from a discharge timeseries, we need the daily discharge.

# Create a data.frame with continuous daily discharge
dailyMQ <- data.frame(Date=seq(from=as.Date("01.01.2000", format="%d.%m.%Y"),
  to=as.Date("01.01.2004", format="%d.%m.%Y"), by="days"),
  discharge=rbeta(1462,2,20)*100)

#Run the separation
Flood_events <- eventsep(dailyMQ)
# The Separation might still contain overlaid flood events which need to be corrected
head(Flood_events)

Correction of flood events

Run the Web separation on the dummy Catchment As input to the function a minimum of discharge data is used. The Flood event tables can be opened from within the User-Interface

Run_WebFlood()

Separation of precipitation

After the flood events are corrected (if needed), the precipitation belonging to flood event need to be estimated For the both functions, a daily precipitation timeseries is needed, as well as the parameter indT, which has the position indices of begin, end and peak of the FLOOD event as vector.

# create a sample dataset
dailyprec <- data.frame(Date=seq(from=as.Date("01.01.2000", format="%d.%m.%Y"),
  to=as.Date("30.04.2000", format="%d.%m.%Y"), by="days"),
  discharge=rbeta(121,2,20)*100)

# Create indices of beginning, end and peak of the flood event
indT <- c(15, 30, 14+which.max(dailyprec[15:30,2])) 

# Run the separation for both methods
Date1 <- PreconeCP(dailyprec, indT = indT)
Date2 <- PrectwoCP(dailyprec, indT = indT)
print(c("Method1" = Date1, "Method2" = Date2))

Typing of flood events

For the typing of the flood event, multiple characteristics for each flood event must be calculated before:

Sum_SM: Sum of snowmelt during the floodevent in mm
Sum_N: Sum of precipitation during the floodevent in mm (For example from the separation of precipitation)
dir_Volume: Direct volume of the flood event (Volume minus baseflow) in Mio. m³/s (This is calculated in the floodevent separation as well as in the output after running the Run_WebFlood)
HQ_dir: Direct peak (instantaneous flood peak minus baseflow) in m³/s (This is calculated in the floodevent separation as well as in the output after running the Run_WebFlood)
PSI_SM: Runoff coefficient of the flood event WITH snowmelt + precipitation

# Open the sample flood event data
data("Sample_Flood_events")
head(Sample_Flood_events)

# Run the event typing
Floods_typed <- Flood_typology(Floods = Sample_Flood_events, n_G = 3, Type_3_min_samplesize = 10)

table(Floods_typed$Type)

# Plot the event typing
Floods_Rain <- Floods_typed[Floods_typed$Type %in% c("R1", "R2", "R3"),]

suppressMessages(library(ggplot2))
ggplot(Floods_Rain)+
  geom_point(aes(x=dir_Volume, y=HQ_dir, fill=Type), colour="black", shape=21, size=2)+
  scale_fill_manual(values = c("R1"="#D7191C", "R2"="#FDAE61", "R3"= "#1A9641"))

Typewise and combined return period estimation with the TMPS model

For the TMPS model, multiple characteristics for each flood event must be present in the input Floods_typed:

Peak_date: Date of the peak of the flood event
HQ: Direct peak in m³/s
Type: Flood type of the flood event, preferably as factor

Also, a daily discharge timeseries with a "Date" and "Discharge" column is needed

# Open the discharge data
data("Discharge")
head(Discharge)


Floods_typed <- Floods_typed[, c("Peak_date","HQ", "Type")]

TMPS_quantiles <- qTMPS(p = c(2, 5, 10, 20, 25, 50, 100, 200, 500, 1000), 
  Flood_events = Floods_typed, Daily_discharge = Discharge, 
  return_TMPS = c("TMPS", "R1", "R2", "R3", "S1", "S2"))

TMPS_quantiles <- as.data.frame(round(TMPS_quantiles,1))
TMPS_quantiles[TMPS_quantiles < 0] <- NA
TMPS_quantiles

# Calculate the AMS with TWO SIMPLIFICATIONS FOR EASYNESS : 
# 1) no hydrological years 
# 2) asummption instantaneous peaks == annual max of daily discharge * 1.1

AMS <- aggregate((Discharge$Discharge*1.1), list(as.numeric(format(Discharge$Date,"%Y"))), max, na.rm=TRUE)$x


# Plot the Results and compare it to the AMS
suppressMessages(library(fExtremes))
suppressMessages(library(reshape2))
suppressMessages(library(ggplot2))

AMS_params <- gevFit(AMS, type="pwm")@fit$par.ests  
AMS <- qgev(1-1/c(2,5,10,20,25,50,100,200,500,1000), mu=AMS_params[2], xi=AMS_params[1], beta=AMS_params[3])

Results <- rbind(TMPS_quantiles, AMS=AMS)
Results <- cbind(Results, "Method" = rownames(Results))
Results_melt <- melt(Results, id.vars = "Method", variable.name = "Annuality")
Results_melt$Annuality <- as.numeric(as.character(Results_melt$Annuality))

cols <- c("AMS"="black",c(R1 = "#D7191C", R2 = "#FDAE61", R3 = "#1A9641", S1 = "#6BAED6", S2 = "#2171B5"),  "TMPS"="darkorchid1")

ggplot(Results_melt)+ 
  theme_bw()+
  geom_line(aes(x=Annuality, y=value, colour = Method ), size=1.5)+ 
  scale_color_manual(values=cols)+
  scale_x_log10()+
  labs(x="T[a]",y="Peak discharge [m³/s]")+
  annotation_logticks(sides = "b")





# Calculate 
TMPS_probs_from_AMS_quants <- pTMPS(q = round(AMS), 
  Flood_events = Floods_typed, Daily_discharge = Discharge, 
  return_TMPS = c("TMPS", "R1", "R2", "R3", "S1", "S2"))

TMPS_probs_from_AMS_quants[is.infinite(TMPS_probs_from_AMS_quants)] <- NA
TMPS_probs_from_AMS_quants <- round(TMPS_probs_from_AMS_quants, 1)
TMPS_probs_from_AMS_quants