introduction to r and extremesericg/rintro.pdf · quantiles from model simulated data 1regression...
TRANSCRIPT
Introduction to R and extRemes
Eric Gilleland Research Applications Laboratory
21 July 2014, Trieste, Italy
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Part I: Intro to R
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
R-Project website
• Documentation • Access to packages via CRAN • Search engines (including one specific to R) • other • http://www.R-project.org
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Advanced
• Reading and Writing Net CDF files http://www.image.ucar.edu/Software/Netcdf/
• A climate related precipitation example http://www.image.ucar.edu/~nychka/FrontrangePrecip/
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
.RData • All work conducted within a “dot” file
called .RData, which is referred to as the workspace and exists in the working directory
• save.image() • getwd() • setwd() • citation() • q() • Functions exist to export results outside of the
workspace
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Accessing Help Files • ?save.image • ?getwd • ?setwd • ?citation • ?q • To write to a file outside of .Rdata working
directory § ?write (text, csv, etc.) § ?write.table (text, csv, etc.)
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Object Oriented
• plot • methods(plot) • methods(class=“lm”) • print • summary • predict • length
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Logical Operators
• &, && (and) • |, || (or) • == • >=, >, <, <= • is.na, is.finite, is.numeric, is.logical,
is.element
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Logical Operators: Accessing Help Files
• ?”&”, ?”&&” • ?”|” • ?”==“ • ?is.na
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Arithmetic Operators
• + • - • * • / • ^, ** • %*% (matrix multiplication)
Assignment Operators
• <- § x <- 1:10
• -> § 1:10 -> x
• = § x = 1:10 § 1:10 = x
• ?assign
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Functions • help, ? • plot • Sys.time • Sys.time() • x <- Sys.time()
§ x is the result of the Sys.time call • y <- Sys.time
§ y is a copy of the function Sys.time • args(plot)
Some of my terminology
• x “gets” 1 to 10 § x <- 1:10 § x <- seq(1,10,1) § x <- seq(1, 10, , 10)
• plot x § plot(x)
• x “at” 3 § x[ 3 ]
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Some of my terminology • open paren
§ ( • close paren
§ ) • square brackets
§ open [ § close ]
• curly brackets / braces § open { § close }
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Packages • Thousands of users have written packages.
Most are freely available in the same place, and accessible (installation) from within an R session (provided you are connected to the internet).
• Must first install the package you want (need only do once), and load the library for each new R session.
• ?install.packages • ?update.packages • ?library • citation(“pkgname”)
Installing / Updating a package
• Need be done only once (per update) • Install extRemes
§ install.packages(“extRemes”) § Select a mirror § Installs extRemes and dependencies
• Updating already installed packages (all at once) § update.packages() § will prompt you for a mirror and may ask for
confirmation to update each package. UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Loading a package
• Must be done for every new R session (if you want to use the package)
• Load extRemes § library(“extRemes”) § loads all dependent packages as well
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Citing R packages
• citation(“pkgname”) • Example: to cite extRemes:
§ citation(“extRemes”)
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Some Atmospheric Science Oriented Packages
• Spatial Statistics § fields § spatstat, § sp § maps § Many More! (See the Spatial Data Task View on the R-Project web site)
• Extreme Value Analysis § extRemes § SpatialExtremes § texmex § Many More! (http://www.ral.ucar.edu/staff/ericg/softextreme.php)
• Forecast Verification / Model Evaluation § verification § SpatialVx
• RadioSonde • ncdf, netCDF • smoothie
Formulas in R
Left as an exercise. Should get somewhat familiar with these because they are used for modeling nonstationary EV models in extRemes (≥ 2.0). ?formula
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Types of R objects
• functions • vectors, matrices, arrays
§ numeric, integer, etc. § character § factor
• data frames • lists • other
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
vectors, matrices, arrays
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
x <- cbind(c(1,2,3), c(4,5,6))!y <- matrix(1:6, ncol = 2)!x!y!x == y!any(x != y)!x[, 1]!x[2, ] <- NA!is.na(x)!x – y!x[1:2, 3]!x[, -1]!
vectors, matrices, arrays
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
x <- rnorm(100)!!class(x)!!is.vector(x)!!is.matrix(x)!!x[ 1:10 ]!!plot(x)!
vectors, matrices, arrays
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
x <- array(1:24, dim = c(2, 3, 4))!!x!!x[, , 1] <- 0!!x!!apply(x, 1:2, sum)!
data frames
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
x <- data.frame(obs = 1:10, x = runif(10),! y = rnorm(10))!!x!!x[, “x”]!!x[, 2]!!plot(x)!!class(x)!!is.list(x)!
lists
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
x <- list(f = function(x) return(x ^2),! x = rnorm(10), y = 1:6)!!x!!x$f(3)!!x$x!!class(x)!
Conclusion of Part I: Intro to R
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Part II: Extreme Value Analysis with extRemes ≥ 2.0
Midwest flood 1993 (NCAR Digital Image Library, DI00578)
Block Maxima
“Il est impossible que l’improbable n’arrive jamais” --Emil Gumbel
Tutorial
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
http://www.ral.ucar.edu/staff/ericg/extRemes/extRemes2.pdf
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
library(extRemes)!!Zmax <- matrix(rnorm(100 * 1000), ! 1000, 100)!!dim(Zmax)!!Zmax <- apply(Zmax, 2, max)!!dim(Zmax)!
Simulate a sample of size 100 of maxima of standard normal distributed samples
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
length(Zmax)!!plot(Zmax, type = “h”, col = “darkblue”)!
Simulate a sample of size 100 of maxima of standard normal distributed samples
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
fit <- fevd(Zmax)!!fit!!plot(fit)!!ci(fit, type = “parameter”)!!distill(fit)!
Fit a GEV distribution to the simulated sample
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
data(SEPTsp)!!?SEPTsp!!par(mfrow = c(2, 2))!!!
Plot maximum winter temperature (°C) in Sept-Iles, Québec
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
plot(TMX1~ Year, data = SEPTsp,! type = "h", col = "darkblue")!!plot(TMX1~ AOindex, data = SEPTsp,! pch = 21, col = "darkblue", ! bg = "lightblue")!!plot(TMX1~ MDTR, data = SEPTsp,! pch = 21, col = "darkblue", ! bg = "lightblue")!!
Plot maximum winter temperature (°C) in Sept-Iles, Québec
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution Maximum winter temperature in Sept-Iles, Québec
1950 1960 1970 1980 1990
1520
25
Year
TMX1
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● ●
●
●
●
●
●
●
●
●
●●
−1 0 1 2 3
1520
25
AOindex
TMX1
●
●
●
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
● ●
●
●
●
●
●
●
●
●
●●
7 8 9 10 11
1520
25
MDTR
TMX1
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
fit0 <- fevd(TMX1, data = SEPTsp, ! units = "deg C")!!fit0!!plot(fit0)!!ci(fit0, type = "parameter")!!ci(fit0)!!
Fit a GEV distribution to maximum winter temperature in Sept-Iles, Québec
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution Fit a GEV distribution to maximum winter temperature in Sept-Iles, Québec
●
●●
●●●●●●●●●●●●●
●●●●●●●●
●●●●●●●
●●●●●●●●●●●
●●●
●
●●● ●
●
14 16 18 20 22 24 26 28
1520
25
Model Quantiles
Empi
rical
Qua
ntile
s
●
●
●
●● ●
● ●●●●●●●●●●
● ●●●●●●
●●● ●●●●●●●●● ●●
●●●● ●●●
●
●●
●●
●
15 20 25
1015
2025
30
TMX1 Empirical Quantiles
Qua
ntile
s fro
m M
odel
Sim
ulat
ed D
ata
1−1 lineregression line95% confidence bands
10 15 20 25 30
0.00
0.04
0.08
N = 51 Bandwidth = 1.448
Den
sity
EmpiricalModeled
2 5 10 50 200 1000
2025
3035
40
Return Period (years)
Ret
urn
Leve
l (de
g C
)
●●●
●●●●●●●●●●●●●●●
●●●●●●●●●●●●●
●●●●●●●●●
●●●●●●
●● ● ●
●
fevd(x = TMX1, data = SEPTsp, units = "deg C")
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
fit1 <- fevd(TMX1, data = SEPTsp, ! location.fun = ~AOindex, ! units = "deg C")!!fit1!!plot(fit1)!!lr.test(fit0, fit1)!!
Fit a GEV distribution to maximum winter temperature in Sept-Iles, Québec
Inclusion of AOindex in location parameter is not significant.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution Fit a GEV distribution to maximum winter temperature in Sept-Iles, Québec
●●
●●●●●●●●
●●●●●●●●●●
●●●●●
●●●●●●●●●●●●
●●●●
●
●●●
●● ● ●
●
●
−1 0 1 2 3 4−1
01
23
4
(Gumbel Scale)
(Standardized) Model Quantiles
Empi
rical
Res
idua
l Qua
ntile
s●
●
●●●
●● ●●●●●●●●●●
●●●
●●●●
●●● ●●
●●●●●●
● ●●●●
●● ● ●●
● ●●
●●
●
15 20 25
1520
25
TMX1 Empirical Quantiles
Qua
ntile
s fro
m M
odel
Sim
ulat
ed D
ata
1−1 lineregression line95% confidence bands
−2 0 2 4 6
0.00
0.10
0.20
0.30
Transformed Data
N = 51 Bandwidth = 0.5314
Den
sity
EmpiricalModeled
0 10 20 30 40 5015
2025
3035
index
Ret
urn
Leve
l (de
g C
) 2−year level20−year level100−year level
fevd(x = TMX1, data = SEPTsp, location.fun = ~STDTMAX, units = "deg C")
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution Fit a GEV distribution to maximum winter temperature in Sept-Iles, Québec
fit2 <- fevd(TMX1, data = SEPTsp, ! scale.fun = ~STDTMAX, ! use.phi = TRUE, units = "deg C")!!fit2!!plot(fit2)!!lr.test(fit0, fit2)!!
Addition of AOindex to scale parameter is not statistically significant.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
Addition from previous slide is a function of the following form. log(σ(AO index)) = φ0 + φ1 × AO index
Because: use.phi = TRUE !
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution
par(mfrow = c(2, 2))!!plot(TMN0~ Year, data = SEPTsp,! type = "h", col = "darkblue")!!plot(TMN0~ AOindex, data = SEPTsp,! pch = 21, col = "darkblue", ! bg = "lightblue")!!plot(TMN0~ MDTR, data = SEPTsp,! pch = 21, col = "darkblue", ! bg = "lightblue")!!!
Plot minimum winter temperature (°C) in Sept-Iles, Québec
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Generalized Extreme Value Distribution Fit GEV to (negative) minimum winter temperature (°C) in Sept-Iles, Québec
fit0 <- fevd(-TMN0 ~ 1, data = SEPTsp, ! units = “neg. deg. C”)!!fit0 !!plot(fit0)!!!The rest of the analysis of (negative) minimum
temperature is left as an exercise.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Part III: Extreme Value Analysis Frequency of extremes
photo from Wikipedia: http://en.wikipedia.org/wiki/Coligny_calendar
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Frequency of Extremes
Number of days that maximum daily temperature (°F) in Fort Collins, Colorado exceeds 95 °F.
data(FCwx)!!?FCwx!!tempGT95 <- aggregate(FCwx$MxT, ! by = list(FCwx$Year), ! function(x) sum(x > 95, ! na.rm = TRUE))!! class(tempGT95)!
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Frequency of Extremes
Number of days that maximum daily temperature (°F) in Fort Collins, Colorado exceeds 95 °F.
names(tempGT95)!!names(tempGT95) <- c(“Year”, “MxT”)!!tempGT95!
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Frequency of Extremes
Fit Poisson distribution to number of days that maximum daily temperature (°F) in Fort Collins, Colorado exceeds 95 °F.
plot(MxT ~ Year, data = tempGT95,! type = “h”, col = “darkblue”,! ylab = !“Number of days MxT > 95 deg. F”)!!fpois(tempGT95$MxT)!
Test for equality of mean and variance says that the two are statistically significantly different.
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Frequency of Extremes
Plot of number of days that maximum daily temperature (°F) in Fort Collins, Colorado exceeds 95 °F.
1900 1920 1940 1960 1980 2000
02
46
8
Year
Num
ber o
f day
s M
xT >
95
deg.
F
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Frequency of Extremes
Fit Poisson to number of days that maximum daily temperature (°F) in Fort Collins, Colorado exceeds 95 degrees F.
fit <- glm(tempGT95~yr, family = poisson())!!summary(fit)!! !
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Part IV: Extreme Value Analysis
Threshold Excesses
V.F.D. Pareto
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
par(mfrow = c(2, 2), oma = c(0,0,2,0))!!plot(Prec ~ Year, data = Denversp, ! type = "h", col = "darkblue", ! lwd = 2, cex.lab = 1.5, ! cex.axis = 1.5, ylab = "")!!plot(Prec ~ Day, data = Denversp, ! type = "h", col = "darkblue", ! lwd = 2, cex.lab = 1.5, ! cex.axis = 1.5, ylab = "")!!
Example: Denver, Colorado July hourly Precipitation (mm)
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
plot(Prec ~ Hour, data = Denversp,! type = "h", col = "darkblue", lwd = 2,! cex.lab = 1.5, cex.axis = 1.5, ! ylab = "")!!mtext("Precipitation (mm)\nDenver, Colorado", ! side = 3, outer = TRUE)!!
Example: Denver, Colorado July hourly Precipitation (mm)
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
50 60 70 80 90
0.0
0.5
1.0
1.5
Year0 5 10 20 30
0.0
0.5
1.0
1.5
Day
5 10 15 20
0.0
0.5
1.0
1.5
Hour
Precipitation (mm)Denver, Colorado
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
threshrange.plot(Denversp$Prec, ! r = c(0.1, 0.95))!!extremalindex(Denversp$Prec, threshold=0.5)!
A threshold of about 0.5 mm seems appropriate for these data. Threshold excesses appear to be independent with this threshold.
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
● ● ● ● ● ● ● ●
●
●
0.2 0.4 0.6 0.8
−2.0
−0.5
0.5
1.5
threshrange.plot(x = Denversp$Prec, r = c(0.1, 0.95))
repa
ram
eter
ized
scal
e
●●
● ● ● ● ● ●
●
●
0.2 0.4 0.6 0.8
−1.0
0.0
1.0
2.0
Threshold
shap
e
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
fitGP <- fevd(Prec, Denversp, threshold=0.5, ! type="GP", units="mm",! time.units="744/year”)!!fitGP!!plot(fitGP)!!
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
●●●●●●●●●
●●●●●
●●●●●●
● ● ● ●●
●
●
0.6 0.8 1.0 1.2
0.6
0.8
1.0
1.2
1.4
1.6
Model Quantiles
Empi
rical
Qua
ntile
s
●●●●●●
●●● ● ● ●●
●●●●
●●●
●●●●
●
●
●
0.6 0.8 1.0 1.2 1.4 1.6
0.5
1.0
1.5
2.0
Prec( > 0.5) Empirical Quantiles
Qua
ntile
s fro
m M
odel
Sim
ulat
ed D
ata
1−1 lineregression line95% confidence bands
0.0 0.5 1.0
0.0
1.0
2.0
3.0
N = 27 Bandwidth = 0.1129
Den
sity
EmpiricalModeled
2 5 10 50 200 500
0.5
1.0
1.5
2.0
2.5
Return Period (years)
Ret
urn
Leve
l (m
m)
●●●●●●●●●●●●●●●●●
●●●
●●●● ●
●
●
fevd(x = Prec, data = Denversp, threshold = 0.5, type = "GP", units = "mm", time.units = "744/year", verbose = TRUE)
Generalized Pareto
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Denver, Colorado July hourly Precipitation (mm)
ci(fitGP, type = "parameter")!!ci(fitGP)!
95% lower CI Estimate 95% upper CI σ(u = 0.5 mm) 0.16 0.32 0.47 ξ -0.48 -0.15 0.19 100-year return level (mm) 1.04 1.49 1.93
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Part V: Extreme Value Analysis Point Process
●
●
●
●
●
●
●
●
●
●
2 4 6 8 10
24
68
time
value
A
Siméon Denis Poisson
Point Process
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
fitPP <- fevd(Prec, Denversp, threshold=0.5, ! type="PP", units="mm”,! time.units="744/year")!!fitPP!!plot(fitPP)!
Point Process
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
●●●●●●●
●●●●●●●●● ●
●● ●
● ● ● ●●
●
●
0.6 0.8 1.0 1.2
0.6
1.0
1.4
Model Quantiles
Empi
rical
Qua
ntile
s
●●●●●●●●● ● ● ●●
●●●● ● ●●
●●
●●
●
●
●
0.6 0.8 1.0 1.2 1.4 1.6
0.5
1.0
1.5
2.0
Prec( > 0.5) Empirical Quantiles
Qua
ntile
s fro
m M
odel
Sim
ulat
ed D
ata
1−1 lineregression line95% confidence bands
0.0 0.5 1.0 1.5 2.0
0.0
0.4
0.8
1.2
N = 42 Bandwidth = 0.1264
Dens
ity
Empirical (year maxima)Modeled
0 1 2 3 4
0.0
1.0
2.0
3.0
Z plot
Expected ValuesUnder exponential(1)
Obs
erve
d Z_
k Va
lues
●●●●●●
●●●●●●●● ●
● ● ●● ● ●
●
●
●
●
●
●
1−1 lineregression line95% confidence bands
2 5 10 20 50 100 500
0.5
1.5
2.5
Return Levels based on approx.equivalent GEV
Return Period (years)
Retu
rn L
evel
(mm
)
● ●●●●●●●●●●●●●
●●●●●●●●●●●
●●●●●●●
●●●● ● ● ●
●
●
fevd(x = Prec, data = Denversp, threshold = 0.5, type = "PP", units = "mm", time.units = "744/year")
Point Process
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
ci(fitPP, type = “parameter”)!!ci(fitPP)!
95% lower CI Estimate 95% upper CI µ (mm) 0.21 0.36 0.51 σ (mm) 0.13 0.34 0.55 ξ -0.48 -0.15 0.19 100-year return level (mm) 1.09 1.48 1.88
Point Process
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Also vary the threshold by hour? And incorporate a diurnal cycle in the location parameter?
u <- numeric(dim(Denversp)[1])!u[ Denversp$Hour < 14 ] <- 0.001!u[ Denversp$Hour >= 14 ] <- 0.5!!fitPP3 <- fevd(Prec, Denversp, ! threshold = u, location.fun = ~Hour, ! type="PP", units="mm",! time.units = "744/year")!!fitPP3!plot(fitPP3)!
Ok, maybe not!
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F) data(Tphap)!!plot(MaxT ~ Year, data = Tphap, pch = 21, ! col = "darkblue", bg = "lightblue", ! lwd = 2, cex.lab = 1.5, cex.axis = 1.5, ! ylab = "")!!!
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
●●●●●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●
●
●
●●
●●
●
●●
●●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●●
●●●
●
●
●●
●
●
●●
●
●●●
●●●
●●
●●●●
●●
●●
●
●
●
●
●●●●
●
●
●
●
●
●●●
●
●
●
●
●●●●
●
●
●
●
●
●●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●●
●●
●
●●●
●
●●
●
●
●●
●
●
●●●
●●
●●
●
●●●●●
●
●●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●●
●●●
●●●●
●●●●
●
●
●●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●●●●
●●●●
●●
●
●
●●
●●●●
●●
●
●
●●
●
●●
●
●●
●
●
●
●●●●
●
●●
●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●●
●●
●
●
●
●
●
●
●●
●●●
●●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●●●●
●●
●
●●
●●
●
●●
●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●
●
●
●
●●
●
●●●
●●●●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●●●
●
●●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●
●●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●●●
●●●
●
●
●
●
●
●
●
●●
●●
●●●●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●●
●
●●
●●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●●
●●●
●
●
●●
●●●
●
●
●
●●●
●
●
●
●
●
●●
●
●●●●
●
●
●
●●
●
●●●
●
●●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●●
●●●●
●
●●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●●
●●
●
●
●●●●
●●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●●
●●●
●●
●●
●●
●
●
●
●
●●
●
●●
●●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●
●
●●●●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●
●●●
●
●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●
●
●●
●
●
●
●●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●●●
●●
●
●●
●●
●●
●
●●
●
●●
●
●
●●●●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●●●
●●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●●●
●
●
●
●
●●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●●●
●
●●
●●
●●
●
●●●
●
●
●
●
●●●
●●●●
●●●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●●●●
●●●●●
●
●
●●
●
●
●
●●
●
●
●
●●●●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●
●●
●
●●●
●
●●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●●
●●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●●
●
●
●●
●●
●●
●●
●
●●●●
●
●●
●●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●
●●
●
●●
●●
●●
●●●
●
●●●●
●
●
●
●●
●●●●
●
●
●●
●
●
●●●
●●
●●
●
●
●
●
●●
●●●●
●●●
●●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●●●●
●●●
●
●
●●
●●
●
●●
●
●●
●
●●
●
●
●●●●
●●
●●
●●
●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●
●●
●●●
●
●●
●
●●●
●
●●●●●
●
●
●
●
●
●●●
●●
●
●
●
●●
●
●
●●
●●
●●
●
●●●●
●●
●
●●●●●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●●●
●●●●●
●
●●●
●
●●
●
●●
●
●
●●
●●●
●●
●●●●●
●●●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●●●●●
●●●
●
●
●
●●●
●●
●
●●●
●
●●●
●
●
●
●
●
●●
●●●
●
●
●
●
●●●●●
●●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●●
●●●●●
●
●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●●
●●
●
●●
●●●
●
●
●
●
●●●
●●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●●
●
●
●●●
●●
●
●
●
●
●●
●
●
●●
●
●
●●●
●●●
●●●●●
●
●
●●
●●
●●●
●
●●
●
●
●●
●●
●
●●
●
●
●●●●●
●
●●●●
●
●
●●
●
●
●●●
●
●
●
●●●●●
●●●●
●●●●●●
●●
●●
●●
●●
●
●●
●
●
●
●
●
●
●●●●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●●
●●
●●●
●
●●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●●
●●●
●
●
●
●
●
●
●
●●
●●
●●
●
●●●●●
●
●
●
●●●
●●●
●
●
●
●●
●●
●
●
●●●
●
●
●●
●
●
●●●
●●
●
●●
●●
●●●●●
●
●
●
●●
●
●●●
●●
●
●
●●●●
●
●
●
● ●
●●
●●
●●●●●
●
●
●
●●
●
●●●
●●
●
●
●●●●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●●
●●
●●●
●●
●●●
●
●●
●
●●
●●
●●●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●●●●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●●●●
●
●
●●●●
●
●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●●●●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●●●
●●●
●●
●●●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●●
●
●●
●●
●●
●
●
●
●●●
●
●
●●●
●
●
●●
●
●
●
●●
●
●●●●
●●●
●●
●
●
●
●
●●
●
●
●
●
●●●●●●
●
●
●
●
●●
●●
●
●
●
●●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●●●●
●
●●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●●●●
●
●●●
●
●
●
●
●●
●●
●
●●
●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●●
●●●
●●●
●
●
●
●
50 60 70 80 90
8090
100
110
Year
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F)
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
threshrange.plot(Tphap$MaxT, ! r = c(105, 110), type = "PP")!
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F)
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F)
●
●● ●
●
●● ●
●
●●
●
●
●●
● ●
●●
●
105 106 107 108 109 110
115.0
116.0
threshrange.plot(x = Tphap$MaxT, r = c(105, 110), type = "PP", nint = 20)
location
●●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
105 106 107 108 109 110
0.8
1.2
1.6
scale
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
105 106 107 108 109 110
−0.4
−0.2
0.0
Threshold
shape
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F) extremalindex(Tphap$MaxT, threshold = 105)!
θ Number of Clusters Run Length 0.21 234 2
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F) y <- decluster(Tphap$MaxT, threshold = 105, ! r = 2)!!y!!plot(y)!
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F)
●●●●●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●
●
●
●●
●●
●
●●
●●
●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●●
●●●
●
●
●●
●
●
●●
●
●●●
●●●
●●
●●●●
●●
●●
●
●
●
●
●●●●
●
●
●
●
●
●●●
●
●
●
●
●●●●
●
●
●
●
●
●●●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●●
●●
●
●●●
●
●●
●
●
●●
●
●
●●●
●●
●●
●
●●●●●
●
●●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●●
●●●
●●●●
●●●●
●
●
●●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●●●●
●●●●
●●
●
●
●●
●●●●
●●
●
●
●●
●
●●
●
●●
●
●
●
●●●●
●
●●
●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●●●
●●
●
●
●
●
●
●
●●
●●●
●●
●
●
●
●
●●
●
●●
●●
●
●
●
●
●
●
●
●●●●
●●
●
●●
●●
●
●●
●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●
●
●●●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●
●
●
●
●●
●
●●●
●●●●●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●●●
●
●●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●
●●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●●●
●●●
●
●
●
●
●
●
●
●●
●●
●●●●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●●
●
●●
●●
●
●
●●
●●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●●●
●
●
●●
●●●
●
●
●●
●●●
●
●
●
●●●
●
●
●
●
●
●●
●
●●●●
●
●
●
●●
●
●●●
●
●●
●●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●●
●●●●
●
●●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●●
●●
●
●
●●●●
●●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●●
●●●
●●
●●
●●
●
●
●
●
●●
●
●●
●●●
●
●
●
●●
●●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●
●
●●●●
●●
●
●●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●
●●●
●
●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●
●
●●
●
●
●
●●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●●●
●●
●
●●
●●
●●
●
●●
●
●●
●
●
●●●●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●●●
●●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●
●●●
●
●
●
●
●●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●●●
●
●●
●●
●●
●
●●●
●
●
●
●
●●●
●●●●
●●●●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●●
●
●
●
●●●
●
●
●
●
●
●
●●●●
●●●●●
●
●
●●
●
●
●
●●
●
●
●
●●●●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●
●●
●
●●●
●
●●●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●●
●●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●●
●
●
●●
●●
●●
●●
●
●●●●
●
●●
●●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●
●●
●
●●
●●
●●
●●●
●
●●●●
●
●
●
●●
●●●●
●
●
●●
●
●
●●●
●●
●●
●
●
●
●
●●
●●●●
●●●
●●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●●●●
●●●
●
●
●●
●●
●
●●
●
●●
●
●●
●
●
●●●●
●●
●●
●●
●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●
●●
●●●
●
●●
●
●●●
●
●●●●●
●
●
●
●
●
●●●
●●
●
●
●
●●
●
●
●●
●●
●●
●
●●●●
●●
●
●●●●●●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●●●
●●●●●
●
●●●
●
●●
●
●●
●
●
●●
●●●
●●
●●●●●
●●●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●
●
●●●●●●
●●●
●
●
●
●●●
●●
●
●●●
●
●●●
●
●
●
●
●
●●
●●●
●
●
●
●
●●●●●
●●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●
●●
●●●●●
●
●●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●●
●
●
●●●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●
●●
●●
●
●●
●●●
●
●
●
●
●●●
●●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●●
●
●
●●●
●●
●
●
●
●
●●
●
●
●●
●
●
●●●
●●●
●●●●●
●
●
●●
●●
●●●
●
●●
●
●
●●
●●
●
●●
●
●
●●●●●
●
●●●●
●
●
●●
●
●
●●●
●
●
●
●●●●●
●●●●
●●●●●●
●●
●●
●●
●●
●
●●
●
●
●
●
●
●
●●●●
●
●
●
●●
●
●●
●
●●●
●
●
●
●
●●
●●
●●●
●
●●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●●●
●●
●
●
●
●
●
●
●
●
●
●●
●●
●●
●●
●
●●
●●●
●
●
●
●
●
●
●
●●
●●
●●
●
●●●●●
●
●
●
●●●
●●●
●
●
●
●●
●●
●
●
●●●
●
●
●●
●
●
●●●
●●
●
●●
●●
●●●●●
●
●
●
●●
●
●●●
●●
●
●
●●●●
●
●
●
●●
●●
●●
●●●●●
●
●
●
●●
●
●●●
●●
●
●
●●●●
●
●
●
●
●
●●●
●
●●
●
●
●
●●
●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●●●
●●
●
●
●
●
●●
●●
●●●
●●
●●●
●
●●
●
●●
●●
●●●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●
●●
●
●
●
●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●
●
●●
●
●
●●
●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●●●●●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●●●●
●
●
●●●●
●
●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●●
●●●●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●●●
●●●
●●
●●●●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●
●●
●
●●
●●
●●
●
●
●
●●●
●
●
●●●
●
●
●●
●
●
●
●●
●
●●●●
●●●
●●
●
●
●
●
●●
●
●
●
●
●●●●●●
●
●
●
●
●●
●●
●
●
●
●●●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●●●●
●
●●
●
●
●●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●●●
●●●●
●
●●●
●
●
●
●
●●
●●
●
●●
●
●
●
●●●
●
●
●
●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●●
●●●
●●●
●
●
●
●
0 500 1000 1500 2000 2500
8090
100
110
decluster.runs(x = Tphap$MaxT, threshold = 105, r = 2)
index
Tphap$MaxT
●●●●●●●●●●●●●●
●
●●●●
●
●
●
●●
●
●
●
●
●
●
●●●●●
●
●
●
●●
●●●●●●●●●●●
●
●●●●●●●
●
●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●●
●
●●
●
●
●
●
●●
●●●
●
●
●
●●●●●
●
●●●
●●●
●●
●●●●
●
●
●●
●
●
●●●●●●●
●
●●
●
●●●
●
●
●●
●
●●●●
●
●
●
●
●●●
●●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●
●●●●●●●●●●●
●
●●●●●●
●
●●●●●
●
●●●●●●●●●●●●●●●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●●●
●
●●●
●
●
●●●●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●●●●●●●
●
●
●
●
●
●
●●●●
●●
●
●
●
●●●●●●
●
●●●
●●●●
●
●●
●●
●
●
●
●
●●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●●●
●
●●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●●●●
●●
●
●●●●
●
●●
●
●●●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●●●
●●●
●●
●●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●
●
●
●
●●
●
●●●
●●●●●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●●●●●
●
●●
●
●
●
●●
●
●●
●●
●
●●
●
●
●
●
●
●
●
●●
●●●
●●●●
●
●●
●
●●●
●
●
●
●●
●
●●●
●●●
●●●
●
●
●
●
●
●
●
●●
●●
●●●●
●
●●
●
●●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●●●
●
●●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●
●
●●●
●
●
●●
●●●
●
●
●●
●●●
●
●
●
●●●
●
●
●
●
●
●●
●●●●●●●●●●
●
●●●●●●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●●●
●●●●●
●
●●●●●
●
●
●
●●●
●
●
●●●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●
●●●
●●
●
●
●●●●
●●
●
●●
●●●●
●
●
●
●
●●●●●
●
●●●●
●●
●
●
●●
●
●
●
●
●
●●
●●
●●●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●●●●
●
●
●
●
●
●
●
●●
●●●●●●●●●●●●●
●
●●●
●
●
●
●
●●
●
●
●
●
●●●
●
●
●●●
●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●●●●●●
●
●
●
●
●●
●
●●
●
●
●●
●
●●
●
●
●●●●
●
●●●●●●●●
●
●●●●●●●
●
●●
●●
●●●●
●
●●●
●
●
●●●●
●
●
●
●●
●●●●●●●●●●●●
●
●●●
●●●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●●
●●
●
●●●●
●
●
●
●
●
●●
●
●
●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●●●●
●
●●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●●●●
●
●●●●●●
●
●
●●
●
●●
●
●●●
●
●●●
●
●●
●
●●●●●
●
●●●●●●●●●●●
●
●●●
●●
●
●
●●
●
●●
●●
●
●●●●●●●●
●
●●●
●
●
●
●
●
●
●●●●
●●●●●
●
●
●
●●●
●
●●
●
●
●
●●●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●
●●
●
●●●
●
●●●●
●
●●●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●●●●●●
●
●
●
●
●●
●
●●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●●
●
●●●
●●
●
●
●●
●●
●●
●
●●●●●●
●
●●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●●●●●●●●
●
●●
●●●
●
●●
●
●
●
●●●●
●
●●
●
●●●●
●
●
●
●●●●●●●
●
●●●●
●
●●●●●●
●
●
●
●
●●
●●●●
●
●●
●●
●
●●
●
●
●
●●●●●
●
●
●
●
●●
●
●
●
●●●●●●●●●●●●●
●
●●●●
●
●●●●●●●●●●●●●●●●●●●●
●●●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●
●
●●●
●●
●●●●●●
●
●●●
●
●●●●●●●
●
●
●
●●●
●●
●
●
●●●●●●
●
●●●●
●
●●●●
●●
●
●●●●●●
●
●
●
●●
●
●●●
●
●
●●
●
●●●●●●●●●
●●●●
●
●
●●●●
●
●●
●
●●●●
●
●●
●
●●●●●
●●●●●●●●●●
●
●●●●
●
●●●
●●●●●
●●●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●●●
●
●●●●●●●
●
●
●
●
●
●
●●
●
●●●●●●
●
●●●
●
●
●●●●
●
●●●●●●●●●●●
●
●
●
●●●●
●
●
●
●●●●●●●
●
●
●
●●●
●●●●
●
●●●●●●
●
●●
●
●●●●●
●●
●●●●●●
●●
●
●
●●●
●●●
●
●●
●●
●
●
●
●
●
●●
●
●
●●●
●●●
●●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●●
●
●●●
●
●
●
●
●●
●
●●●●●●●●●●
●
●
●
●
●
●
●
●●●●●●●●●●●
●
●●●●
●
●
●●●
●●
●
●
●
●
●●
●
●
●●
●
●
●●●●●●●●●●●●●●●●●●●●●
●
●●●
●●
●●
●●
●
●●
●●●●●
●
●●●●●
●
●●
●
●
●●●
●
●
●●●●●●
●
●●●●●●●●●●●
●
●●●●●●●●●
●
●
●●●●●●●
●
●●●●●●●●●●●●●●
●
●●
●●
●●●
●
●●
●●●●
●
●
●
●
●
●
●●
●
●
●
●●●
●
●
●
●
●●●●●●●●●●●●●●
●
●●●●
●●
●●●
●
●●●●●
●
●
●
●
●
●
●●
●●
●●
●
●●●●●
●
●●●●
●
●●●●●
●
●●
●●
●
●●●●
●
●
●●
●
●●●●
●●
●
●
●
●●●●●●●
●
●
●
●●
●●●●●●●●●●●
●
●●●
●●
●
●
●●●●●●●
●
●
●
●●
●●●●●●●●●●●
●
●●●
●
●
●●●
●
●●
●
●●●
●
●●●●●●●●●
●
●●
●
●
●
●
●
●
●●●●●
●
●●
●●
●●
●
●●●●●●●●
●●
●●●●●●●●●●●
●
●●●●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●●
●
●
●
●●●●●●●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●●
●
●
●●●
●
●●
●
●
●●
●●
●●
●
●
●
●
●
●●
●
●
●●●●
●
●●●●●●●●●●●●●
●●
●
●
●
●●●●●
●
●
●
●●
●
●
●
●●
●
●●●●●
●●
●●●
●
●
●●
●
●
●●
●
●●●●●●
●
●●
●
●
●
●
●
●
●●●●
●
●
●●●●
●
●
●
●
●
●●●●●●●
●
●●●●●●●
●
●●●●●●
●
●
●
●
●
●●●●
●
●
●
●
●
●●
●
●●●●●●●●●●●●●
●
●
●
●
●
●●●●●
●
●●
●
●
●
●●
●
●●
●
●●
●
●
●●
●
●
●
●●●
●
●
●●●
●
●
●●
●
●
●●●●●●●●●●●●●●
●
●●●●
●
●
●
●●●●
●
●●●●●●
●●
●●
●
●
●
●●●●●●●●
●
●●●●●
●
●
●
●●
●●
●●●●
●
●●●
●
●●●●●●
●
●●●●●●●
●
●●●●●●●●
●
●●●
●
●●●●
●
●●●
●
●
●
●
●●●●
●
●●
●
●
●
●●●
●
●●●●
●
●●●
●
●
●●
●●●
●
●
●
●
●
●●
●
●●●●
●
●
●●●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●●●
●●●
●●●
●
●
●
●
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F) Tphap2 <- Tphap!Tphap2$MaxT.dc <- c(y)!!fit <- fevd(MaxT.dc, threshold = 105, ! data = Tphap2, type = "PP",! time.units = "62/year", units = ! "deg F")!!fit!!plot(fit)!!
Declustering
UCAR Confidential and Proprietary. © 2014, University Corporation for Atmospheric Research. All rights reserved.
Example: Sky Harbor airport, Phoenix, Arizona July to August maximum temperatures (°F)
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●
●●●●
● ●
106 108 110 112 114 116 118
106
110
114
118
Model Quantiles
Empi
rical
Qua
ntile
s
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●
●●●●●●●●●●● ●●●●●●●
●●●●●●●●●●●●●●●●●●●●● ●●●●●●●
●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●● ●●●●●●
●●●●●●●●●● ●●●●
●●●●●●●●●●●●● ●●
●● ●
●
106 108 110 112 114 116 118
106
110
114
118
MaxT.dc( > 105) Empirical Quantiles
Qua
ntile
s fro
m M
odel
Sim
ulat
ed D
ata
1−1 lineregression line95% confidence bands
105 110 115 120
0.00
0.10
N = 43 Bandwidth = 0.9037
Dens
ity
Empirical (year maxima)Modeled
0 1 2 3 4 5 6
12
34
5
Z plot
Expected ValuesUnder exponential(1)
Obs
erve
d Z_
k Va
lues
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●●●●●●●●●●
●●●●●●●●●●●●
●●●●●●●●●●●●● ●
● ●●
●●
●●
1−1 lineregression line95% confidence bands
2 5 10 20 50 100 500
113
115
117
119
Return Levels based on approx.equivalent GEV
Return Period (years)
Retu
rn L
evel
(deg
F)
●●●
●●
●●●●●●●●●
●●●●●●●●●●
●●●●●
●●●●●●●
● ● ●
● ●
fevd(x = MaxT.dc, data = Tphap2, threshold = 105, type = "PP", units = "deg F", time.units = "62/year")