|
|
|
## CHAPTER5 contains the commands used in Chapter 5 of the tutorial
|
|
|
|
#
|
|
|
|
# CALL: Chapter5
|
|
|
|
#
|
|
|
|
# Some of the commands are edited for fast computation.
|
|
|
|
# Each set of commands is followed by a 'pause' command.
|
|
|
|
#
|
|
|
|
|
|
|
|
# Tested on Matlab 5.3
|
|
|
|
# History
|
|
|
|
# Added Return values by GL August 2008
|
|
|
|
# Revised pab sept2005
|
|
|
|
# Added sections -> easier to evaluate using cellmode evaluation.
|
|
|
|
# Created by GL July 13, 2000
|
|
|
|
# from commands used in Chapter 5
|
|
|
|
#
|
|
|
|
|
|
|
|
## Chapter 5 Extreme value analysis
|
|
|
|
|
|
|
|
## Section 5.1 Weibull and Gumbel papers
|
|
|
|
from __future__ import division
|
|
|
|
import numpy as np
|
|
|
|
import scipy.interpolate as si
|
|
|
|
from wafo.plotbackend import plotbackend as plt
|
|
|
|
import wafo.data as wd
|
|
|
|
import wafo.objects as wo
|
|
|
|
import wafo.stats as ws
|
|
|
|
import wafo.kdetools as wk
|
|
|
|
pstate = 'off'
|
|
|
|
|
|
|
|
# Significant wave-height data on Weibull paper,
|
|
|
|
|
|
|
|
fig = plt.figure()
|
|
|
|
ax = fig.add_subplot(111)
|
|
|
|
Hs = wd.atlantic()
|
|
|
|
wei = ws.weibull_min.fit(Hs)
|
|
|
|
tmp = ws.probplot(Hs, wei, ws.weibull_min, plot=ax)
|
|
|
|
plt.show()
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block = 1'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Significant wave-height data on Gumbel paper,
|
|
|
|
plt.clf()
|
|
|
|
ax = fig.add_subplot(111)
|
|
|
|
gum = ws.gumbel_r.fit(Hs)
|
|
|
|
tmp1 = ws.probplot(Hs, gum, ws.gumbel_r, plot=ax)
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
plt.show()
|
|
|
|
#disp('Block = 2'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Significant wave-height data on Normal probability paper,
|
|
|
|
plt.clf()
|
|
|
|
ax = fig.add_subplot(111)
|
|
|
|
phat = ws.norm.fit2(np.log(Hs))
|
|
|
|
phat.plotresq()
|
|
|
|
#tmp2 = ws.probplot(np.log(Hs), phat, ws.norm, plot=ax)
|
|
|
|
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
plt.show()
|
|
|
|
#disp('Block = 3'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Return values in the Gumbel distribution
|
|
|
|
plt.clf()
|
|
|
|
T = np.r_[1:100000]
|
|
|
|
sT = gum[0] - gum[1] * np.log(-np.log1p(-1./T))
|
|
|
|
plt.semilogx(T, sT)
|
|
|
|
plt.hold(True)
|
|
|
|
# ws.edf(Hs).plot()
|
|
|
|
Nmax = len(Hs)
|
|
|
|
N = np.r_[1:Nmax + 1]
|
|
|
|
|
|
|
|
plt.plot(Nmax/N, sorted(Hs, reverse=True), '.')
|
|
|
|
plt.title('Return values in the Gumbel model')
|
|
|
|
plt.xlabel('Return period')
|
|
|
|
plt.ylabel('Return value')
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
plt.show()
|
|
|
|
#disp('Block = 4'),pause(pstate)
|
|
|
|
|
|
|
|
## Section 5.2 Generalized Pareto and Extreme Value distributions
|
|
|
|
## Section 5.2.1 Generalized Extreme Value distribution
|
|
|
|
|
|
|
|
# Empirical distribution of significant wave-height with estimated
|
|
|
|
# Generalized Extreme Value distribution,
|
|
|
|
gev = ws.genextreme.fit2(Hs)
|
|
|
|
gev.plotfitsummary()
|
|
|
|
# wafostamp([],'(ER)')
|
|
|
|
# disp('Block = 5a'),pause(pstate)
|
|
|
|
|
|
|
|
plt.clf()
|
|
|
|
x = np.linspace(0,14,200)
|
|
|
|
kde = wk.TKDE(Hs, L2=0.5)(x, output='plot')
|
|
|
|
kde.plot()
|
|
|
|
plt.hold(True)
|
|
|
|
plt.plot(x, gev.pdf(x),'--')
|
|
|
|
# disp('Block = 5b'),pause(pstate)
|
|
|
|
|
|
|
|
# Analysis of yura87 wave data.
|
|
|
|
# Wave data interpolated (spline) and organized in 5-minute intervals
|
|
|
|
# Normalized to mean 0 and std = 1 to get stationary conditions.
|
|
|
|
# maximum level over each 5-minute interval analysed by GEV
|
|
|
|
xn = wd.yura87()
|
|
|
|
XI = np.r_[1:len(xn):0.25] - .99
|
|
|
|
N = len(XI)
|
|
|
|
N = N - np.mod(N, 4*60*5)
|
|
|
|
|
|
|
|
YI = si.interp1d(xn[:, 0], xn[:, 1], kind='linear')(XI)
|
|
|
|
YI = YI.reshape(4*60*5, N/(4*60*5)) # Each column holds 5 minutes of
|
|
|
|
# interpolated data.
|
|
|
|
Y5 = (YI - YI.mean(axis=0)) / YI.std(axis=0)
|
|
|
|
Y5M = Y5.maximum(axis=0)
|
|
|
|
Y5gev = ws.genextreme.fit2(Y5M,method='mps')
|
|
|
|
Y5gev.plotfitsummary()
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block = 6'),pause(pstate)
|
|
|
|
|
|
|
|
## Section 5.2.2 Generalized Pareto distribution
|
|
|
|
|
|
|
|
# Exceedances of significant wave-height data over level 3,
|
|
|
|
gpd3 = ws.genpareto.fit2(Hs[Hs>3]-3, floc=0)
|
|
|
|
gpd3.plotfitsummary()
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
|
|
|
|
##
|
|
|
|
plt.figure()
|
|
|
|
# Exceedances of significant wave-height data over level 7,
|
|
|
|
gpd7 = ws.genpareto.fit2(Hs(Hs>7), floc=7)
|
|
|
|
gpd7.plotfitsummary()
|
|
|
|
# wafostamp([],'(ER)')
|
|
|
|
# disp('Block = 6'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
#Simulates 100 values from the GEV distribution with parameters (0.3, 1, 2),
|
|
|
|
# then estimates the parameters using two different methods and plots the
|
|
|
|
# estimated distribution functions together with the empirical distribution.
|
|
|
|
Rgev = ws.genextreme.rvs(0.3,1,2,size=100)
|
|
|
|
gp = ws.genextreme.fit2(Rgev, method='mps');
|
|
|
|
gm = ws.genextreme.fit2(Rgev, *gp.par.tolist(), method='ml')
|
|
|
|
gm.plotfitsummary()
|
|
|
|
|
|
|
|
gp.plotecdf()
|
|
|
|
plt.hold(True)
|
|
|
|
plt.plot(x, gm.cdf(x), '--')
|
|
|
|
plt.hold(False)
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block =7'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# ;
|
|
|
|
Rgpd = ws.genpareto.rvs(0.4,0, 1,size=100)
|
|
|
|
gp = ws.genpareto.fit2(Rgpd, method='mps')
|
|
|
|
gml = ws.genpareto.fit2(Rgpd, method='ml')
|
|
|
|
|
|
|
|
gp.plotecdf()
|
|
|
|
x = sorted(Rgpd)
|
|
|
|
plt.hold(True)
|
|
|
|
plt.plot(x, gml.cdf(x))
|
|
|
|
# gm = fitgenpar(Rgpd,'method','mom','plotflag',0);
|
|
|
|
# plot(x,cdfgenpar(x,gm),'g--')
|
|
|
|
#gw = fitgenpar(Rgpd,'method','pwm','plotflag',0);
|
|
|
|
#plot(x,cdfgenpar(x,gw),'g:')
|
|
|
|
#gml = fitgenpar(Rgpd,'method','ml','plotflag',0);
|
|
|
|
#plot(x,cdfgenpar(x,gml),'--')
|
|
|
|
#gmps = fitgenpar(Rgpd,'method','mps','plotflag',0);
|
|
|
|
#plot(x,cdfgenpar(x,gmps),'r-.')
|
|
|
|
plt.hold(False)
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block = 8'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Return values for the GEV distribution
|
|
|
|
T = np.logspace(1, 5, 10);
|
|
|
|
#[sT, sTlo, sTup] = invgev(1./T,Y5gev,'lowertail',false,'proflog',true);
|
|
|
|
|
|
|
|
#T = 2:100000;
|
|
|
|
#k=Y5gev.params(1); mu=Y5gev.params(3); sigma=Y5gev.params(2);
|
|
|
|
#sT1 = invgev(1./T,Y5gev,'lowertail',false);
|
|
|
|
#sT=mu + sigma/k*(1-(-log(1-1./T)).^k);
|
|
|
|
plt.clf()
|
|
|
|
#plt.semilogx(T,sT,T,sTlo,'r',T,sTup,'r')
|
|
|
|
#plt.hold(True)
|
|
|
|
#N = np.r_[1:len(Y5M)]
|
|
|
|
#Nmax = max(N);
|
|
|
|
#plot(Nmax./N, sorted(Y5M,reverse=True), '.')
|
|
|
|
#plt.title('Return values in the GEV model')
|
|
|
|
#plt.xlabel('Return priod')
|
|
|
|
#plt.ylabel('Return value')
|
|
|
|
#plt.grid(True)
|
|
|
|
#disp('Block = 9'),pause(pstate)
|
|
|
|
|
|
|
|
## Section 5.3 POT-analysis
|
|
|
|
|
|
|
|
# Estimated expected exceedance over level u as function of u.
|
|
|
|
plt.clf()
|
|
|
|
|
|
|
|
mrl = ws.reslife(Hs,'umin',2,'umax',10,'Nu',200);
|
|
|
|
mrl.plot()
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block = 10'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Estimated distribution functions of monthly maxima
|
|
|
|
#with the POT method (solid),
|
|
|
|
# fitting a GEV (dashed) and the empirical distribution.
|
|
|
|
|
|
|
|
# POT- method
|
|
|
|
gpd7 = ws.genpareto.fit2(Hs(Hs>7)-7, method='mps', floc=0)
|
|
|
|
khat, loc, sigmahat = gpd7.par
|
|
|
|
|
|
|
|
muhat = len(Hs[Hs>7])/(7*3*2)
|
|
|
|
bhat = sigmahat/muhat**khat
|
|
|
|
ahat = 7-(bhat-sigmahat)/khat
|
|
|
|
x = np.linspace(5,15,200);
|
|
|
|
plt.plot(x,ws.genextreme.cdf(x, khat,bhat,ahat))
|
|
|
|
# disp('Block = 11'),pause(pstate)
|
|
|
|
|
|
|
|
##
|
|
|
|
# Since we have data to compute the monthly maxima mm over
|
|
|
|
#42 months we can also try to fit a
|
|
|
|
# GEV distribution directly:
|
|
|
|
mm = np.zeros((1,41))
|
|
|
|
for i in range(41):
|
|
|
|
mm[i] = max(Hs[((i-1)*14+1):i*14])
|
|
|
|
|
|
|
|
|
|
|
|
gev = ws.genextreme.fit2(mm)
|
|
|
|
|
|
|
|
|
|
|
|
plt.hold(True)
|
|
|
|
gev.plotecdf()
|
|
|
|
|
|
|
|
plt.hold(False)
|
|
|
|
#wafostamp([],'(ER)')
|
|
|
|
#disp('Block = 12, Last block'),pause(pstate)
|
|
|
|
|