WAFO { a Matlab toolbox for analysis of random waves and loads

Per Andreas Brodtkorb

Department of Marine Hydrodynamics, NTNU, Trondheim, Norway
Par Johannesson, Georg Lindgren, Igor Rychlik, Jesper Ryden and Eva Sjo
Centre for Mathematical Sciences, Lund University, Sweden

ABSTRACT in Rychlik and Lindgren (1993).

WAFO (Wave Analysis for Fatigue and Oceanography)
is a toolbox of Matlab routines for statistical analysis
and simulation of random waves and random loads. An
important part are routines for calculation of theoreti-
cal distributions of wave characteristics from observed
The algorithms are based on a specication of the random
waves by means of their (uni-directional or directional) spec-
trum, and on the underlying assumption of linear wave theory
and Gaussian distribution. However, a transformation of sea el-
evation data can be made to obtain a desired (horizontal) asym-
metric marginal distribution.
or theoretical power spectra of the sea. A new philoso-
phy concerning documentation is presented. The tool-
box contains scripts for generation of plots in recent
publications and scripts for comparison with results in
the literature. The theoretical density of rain
ow cycles new numerical routines have been introduced, and a consider-
can be computed from parameters of random loads.
KEYWORDS: Gaussian processes, wave spectra, esti-
mation, simulation, rain
A rst complete toolbox FAT (Fatigue Analysis Toolbox)
was presented in 1993 (Frendahl, Lindgren and Rychlik 1993);
this was followed by WAT (Wave Analysis Toolbox) in 1995
(Rychlik and Lindgren 1995) 1 , being extended with routines for
probabilistic-modelling problems in oceanography. In WAFO,
able improvement in computational speed and accuracy has been
achieved. WAFO allows treatment of more complicated prob-
lems; for example, spatial waves with time dynamics can be han-
ow cycles, fatigue dled, thus extending the analysis to random elds. Algorithms
ow analysis of switching Markov chains are included, as
INTRODUCTION
ow matrix. Many of the
In a random wave model, like that for Gaussian or transformed
Gaussian waves, the distribution of wave characteristics such as
wave period and crest-trough wave height can be calculated by
high accuracy for almost any spectral type. WAFO is a third-
generation package of Matlab routines for handling statistical
modelling, calculation and analysis of random waves and wave
characteristics and their statistical distributions. The package
also contains routines for cycle counting and computation in ran-
dom load models, in particular the rain
ow counting often used  transformation to Gaussian marginals (exact distribu-
in fatigue life prediction.
Random wave distributions are notoriously dicult to ob-
tain in explicit form from a random wave model, but numeri-
cal algorithms, based on the so-called regression approximation,
work well. This method to calculate wave distributions is the
only known method that gives correct answers valid for general
spectra. The theoretical background is reviewed in Lindgren
and Rychlik (1991) and computational aspects and algorithms
1 http://www.maths.lth.se/matstat/staff/georg/watinfo.html
Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb
for rain
well as for decomposition of the rain
new tools are the result of recent research, e.g. Rychlik, Johan-
nesson and Leadbetter (1997), Podgorski, Rychlik and Machado
(1999), Podgorski et. al (1999) and Johannesson (1999).
Further, WAFO has a modular structure, so users can eas-
ily add their own algorithms for special purposes. The modules
of the toolbox handle
 wave/load data analysis and estimation
 spectral distributions
tions)
 simple parametric approximations to wave characteristic
distributions
 simulation of Gaussian and Markovian wave/load time se-
ries
 extreme value and other statistical analysis
 cycle counting
Page: 1 of 8
  rain
ow cycle analysis and calculation The documentation directory /WAFO/docs/ contains all the
 fatigue life calculation
documentation available for the toolbox. The contents of any of
these les may be examined by typing its name for ascii les
 smoothing and visualization or viewing in ghostview for postscript les. Also each function
In the following section, we discuss in more detail the idea is well documented containing a help header describing how the
of the modular structure. That section is followed by an overview function works with a detailed list of input and output arguments
of the organization of WAFO, presenting some of the capabili- with examples of how to use the function.
ties of the toolbox. Finally, we give a number of examples to
demonstrate the use of some of the tools in WAFO for analysis
and modelling.
PHILOSOPHY { SOME FEATURES OF WAFO
The Matlab code to each function le also contains refer-
ences to related functions and a complete reference to published
articles from which the user can obtain further information if
such exist.
One important enhancement of the new toolbox is the use
of structure arrays in Matlab by which several types of data can
be stored as one object. This signicantly simplies the passing
A common problem with research involving complex scientic
(numerical) computations is that when researchers try to re-
produce and leverage their colleagues work, they often spend
a considerable amount of time just reproducing it.
Also a few months after they completed their own work,
authors are usually not capable of reproducing it without a great
deal of agony, due to the loss of the original input data and pa-
of input and output arguments of functions and also makes the
Matlab workspace much tidier when working with the new tool-
box compared to the old one. Three structures or object classes
are implemented and extensively used: the spectrum structure,
covariance structure and probability density function (hereafter
denoted pdf) structure. Thus the toolbox requires Matlab Ver-
sion 5 or newer and is portable to any computational environ-
rameter values etc. Thus many scientic articles are reproducible ment that supports Matlab, such as Unix or PC with MS Win-
in principle, but not in practice. dows.
To deal with this and to organize computational scientic All the les in the package are located in subdirectories
research and hence to conveniently transfer our technology, we under the main directory. The following directories are related
impose a simple ling discipline on the authors contributing to
the WAFO-toolbox. (A positive side eect of this discipline is a
reduced amount of errors which are prone to occur in computa-
tional science.)
This philosophy is adopted from the article by Matthias
to what has been discussed above. In the next section, we de-
scribe in more details the directories (or modules) which contain
routines for application.
/WAFO is the main directory containing dierent directories
Schwab et. al \Making scientic computations reproducible" for the WAFO software, datasets and documentation.
(http://sepwww.stanford.edu/research/redoc/). /WAFO/docs contains the documentation for the toolbox
The idea is to develop reproducible knowledge about the re- both in ascii and postscript format.
sults of the computational experiments (research) done at Lund /WAFO/paper is a subdirectory including scripts for repro-
University and to make it available to other researchers for their ducing gures in various articles and technical reports.
inspection, modication, re-use and criticism.
As a consequence, WAFO is freely available through the /WAFO/demos contains dierent demonstrations that illus-
Internet2 . Other researchers can here obtain the Matlab code trates and highlights certain aspects of WAFO.
which generated gures in articles and reproduce them. They /WAFO/data contains datasets used in the demo and paper
can if they wish modify the calculations by editing the under- scripts.
lying code, input arguments and parameter values. They can
use the algorithms on other data sets or they can try their own
methods on the same data sets and compare the methods in a
fast and easy fashion.
This is the reason of existence for the WAFO/paper directory
/WAFO/source/ contains mex and Fortran source les.
/WAFO/exec/.. contains Fortran compiled executables for
dierent computers and platforms.
which contains subdirectories including scripts for recreating g- ORGANIZATION OF WAFO
ures in published articles and technical reports. Each article
has its own subdirectory. The directories contain demonstration In this section, we make a brief presentation of each module.
scripts to generate individual gures and (possibly) specialized The text will not be a complete list of routines; such a list may
tools/functions not available in the ocial release of WAFO for be found at the web site for WAFO. We want to emphasize
generating these gures. that all routines in WAFO work together { the division into
Just like the WAFO/paper directory the WAFO/demos direc- sub-toolboxes is only to make it easier for the user to nd the
tory also contains dierent subdirectories with scripts producing routines for his actual problem.
gures. The only dierence is that it does not reproduce g-
ures from published articles but merely test and demonstrate Data analysis. The routines in this category treat data in the
various methodologies, highlight some features of WAFO, and form of time series. As examples of routines, we nd procedures
release code that approximately reproduces gures in other arti- for extraction of so-called turning points, from which troughs
cles. The important thing for both directories is not the printed and crests may be obtained, as well as procedures for estimation
gures, but the underlying algorithm and code. In addition, the of autocovariance function and one-sided spectral density. One
paper and demos scripts constitute an excellent starting point routine extracts wave heights and steepnesses. Numerous plot-
for the novel user to learn about WAFO. ting routines are included.
2 http://www.maths.lth.se/matstat/wafo/

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 2 of 8


Spectrum. Computation of spectral moments and covariance
functions, given a spectrum, is a necessary step for calculation
of exact probability distributions of wave characteristics. The
spectrum structure mentioned in the previous section allows
this calculation to be performed for directional spectra as well
as encountered spectra. We present routines for calculations
of commonly used frequency spectra S (!) , e.g. JONSWAP,
Torsethaugen. The spectra can be expressed in frequency as
well as wave number. Libraries of spreading functions D() , in
ow cycles. For
some cases allowed to be also frequency dependent, cf. Krogstad
ow
and Barstow (1999), are included.
Transformed Gaussian processes. WAFO is mainly in-
tended to model linear, Gaussian waves. For this cathegory of
waves, the exact distributions of wave characteristics can be
calculated, given a spectrum; for example
 pdf for wavelength (period)
 joint pdf for wavelength (period) and amplitude
distributions'; cf. Silverman (1986), Wand and Jones (1995).
They are, however, also applicable to multi-dimensional data,
and hence very useful for smoothing purposes when comparing
(theoretical) joint distributions of wave characteristics to data.
Cf. Silverman (1986), Wand and Jones (1995).
Discretization and cycle counting. After extraction of
the so-called sequence of turning points (the sequence of local
maxima and minima) from data, cycle counts can be obtained,
e.g. max2min cycles, trough-crest cycles, rain
decriptive statistics, the counting distribution and the rain
matrix are important; these can be obtained. Given a cycle
matrix, histograms for amplitude and range, respectively, are
possible to obtain.
Markov models. If the sequence of turning points forms
a Markov chain (MC), it is called a MC of turning points
(MCTP). An important property is the Markov matrix: the
expected histogram matrix of min2max and max2min cycles.
Given a rain
ow matrix of a MCTP, one can nd its Markov
 joint pdf of half wavelengths matrix. In WAFO, algorithms are implemented to calculate the
Routines for transformed Gaussian processes, cf. Rychlik et. al rain
ow matrix for a MC and a MCTP, cf. Frendahl and Rychlik
(1997), are included. (1993).

Parametric models. In WAFO, we have implemented certain Switching Markov models. In some applications, one wants
models for distributions of wave characteristics found in the to model data, whose properties change according to an under-
literature. For example, one nds
 approximations of the density (Tc ; Ac ) in a stationary
Gaussian transformed process (Cavanie, Arhan and Ezraty
(1976), Longuet-Higgins (1983))
 a model for the cdf/pdf of breaking limited wave heights
(Tayfun (1981))
 a model for the cdf/pdf of large wave heights (Tayfun
(1990))
Simulation of random processes. Ecient0 simulation of
a Gaussian process X (t) and its derivative X (t) , given the
spectral density or the auto-correlation function, can be per-
formed. A routine for simulation of a transformed Gaussian
process (and its derivative) is also included. For fast and exact
simulation, some routines use a technique with circulant em-
bedding of the covariance matrix, Dietrich and Newsam (1997).
lying, often unobserved process, called the regime process. The
state of the regime process controls which parameters to use and
when to switch the parameter values. The regime process can
be modelled by a Markov chain, and this is the fundamental
basis for the set of routines presented. For an application with
switching Markov models for fatigue problems, cf. Johannesson
(1998), (1999).
Fatigue and Damage. In WAFO, routines for calculation
of the ackumulated damage ackording to the Palmgren-Miner
rule have been implemented. It is possible to compute the total
damage from a cycle count as well as from a cycle matrix.
Plot routines, cycles and fatigue. This module contains
routines for visualization of cycle counts and cycle matrices and
various histograms.
More traditional spectral simulation methods (FFT) are also
used. Simulation of discrete Markov chains, Markov chains of
turning points, switching Markov chains etc. is possible.
Statistical tools and extreme value distributions. Certain
probability distributions are extensively used in ocean engineer-
ing, e.g. Rayleigh, Gumbel, Weibull. The generalized extreme-
value distributions (GEV) and generalized Pareto distributions
(GPD) are also important. For the distributions mentioned, it
is possible to estimate parameters, generate random variables,
evalute pdf and cumulative distribution function, and plot in
various probability papers. One cathegory of routines handles
Miscellaneous routines. We nd here various plot rou-
tines, algorithms for numerical integration, and functions for
documentation of WAFO with modules.
SOME APPLICATIONS OF WAFO
In this section we demonstrate in examples some of the capabil-
ities of WAFO. For further examples and knowledge about the
algorithms used in the routines, we refer to the tutorial and the
documentation in the routines. Note that the necessary Matlab
bivariate distributions. Besides having routines for estimation code for generation of the gures in this paper is found in the
of parameters etc. for the two-dimensional Weibull distribution, directory WAFO/paper.
bivariate modelling is possible. We start by dening a frequency spectrum, S (!) , which
will be used in many of the examples; we choose a Torsethaugen
Kernel-density-estimation tools. The routines in this spectrum with the parameters Hm0 = 6 m, Tp = 8 s, describing
cathegory complement the ones found in 'Data analysis' and, signicant wave height and primary peak period, respectively.
obviously, the routines in 'Statistical tools and extreme value The energy is divided between two peaks, corresponding to con-

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 3 of 8

tributions from wind and swell (Torsethaugen 1996). WAFO density. A wave characteristics as, for example, wave period, can
allows spectra to be dened simply by their parameters Hm0 be dened in several ways, and WAFO allows the user to choose
and Tp . between a number of denitions: trough-to-crest, down-to-up
crossing, up-to-up crossing etc.
Simulation from spectrum, estimation of spectrum In the numerical example, we have chosen to consider the
down-to-up crossing denition, thus rather a trough period. The
In Figure 1, we have simulated a sample path from S (!) . The wave periods can be extracted from the realization in Figure 1,
algorithm can equally well have the covariance function as input. and are shown as a histogram in Figure 3. This histogram may
The user species the number of wanted points in the simulation. be compared to the theoretical density, calculated from S (!) ,
5
and the estimated density Sest (!) ; see Figure 3.
4
0.3
3

2
Wave amplitude (m)

0.25
1

0 0.2
1

2 0.15

3
0.1
4

5
0 50 100 150 200 250 300 0.05
Time (s)
Figure 1. Part of a simulation from S (!) , a Torsethaugen
spectrum with Hm0 = 6 m, Tp = 8 s. Total number of points 0
0 2 4 6 8 10
= 1000 ,
t = 1:05 s. The rst 300 seconds are shown. Wave period, downtoup crossing (s)
Figure 3. Solid line: pdf for wave period, given S (!) . Dash-
A common situation is data given in form of a time series, dotted line: pdf for wave period, given Sest (!) . The histogram
for which one wants to estimate the related spectrum. From the shows the wave periods extracted from simulated data.
realization in Figure 1, we will now nd an estimate Sest (!)
and compare the result to the original Torsethaugen spectrum
S (!) . In Figure 2, the two spectra are displayed. The maximum
lag size of the Parzen window function used can be chosen by Directional spectra
the user or automatically by WAFO.
Spectral density In WAFO one nds means for evaluation and visualization of
7 fp = 0.79 (rad/s)
directional spectra, that is

6
S (!; ) = S (!)D(; !)
5
S(w) (m s/rad)

where S (!) is a frequency spectrum and D(; !) is a spread-

4
2

3
ing function. A number of common spreading functions can be
chosen by the user.
2 One way of visualizing S (!; ) is a polar plot. In Figure 4
we show the resulting directional spectrum (solid line) for the
1 Torsethaugen spectrum used above. The spreading function is
of the cos-2s type, that is (in the frequency dependent case)
0
0 0.5 1 1.5 2 2.5 3
Frequency (rad/s)
 
Figure 2. Solid: original spectrum. Dashed: spectrum esti- D() = 2p (s(s++1)1=2) cos2s 2
mated from data (1000 observations) and 95% condence inter-
val (dotted). Maximum lag size of the Parzen wind = 80 .

Probability distributions of wave characteristics with s=15 . Note that the two peaks can be distinguished. The
dash-dotted line is the corresponding result when the spread-
WAFO gives the possibility to compute exact probability dis- ing function is frequency dependent, cf. Krogstad and Barstow
tributions for a number of wave characteristics, given a spectral (1999).

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 4 of 8

 Directional Spectrum the notation for the parameters used in our examples is given.
Level curves at:
90 Tcf crest front wave period
Tcr crest rear wave period
1.5
1
2 120 60
3
Hd zero-downcrossing wave height
Ac crest amplitude
4
5 1

150 30
Tc crest period
0.5

180 0

Joint density of (Vcf,2Ac)

15
210 330 Level curves enclosing:

10
30
50
70
240 300 90
95
270 99
10 99.9

Figure 4. Directional spectrum. The frequency spectrum is a

amplitude [m]
Torsethaugen spectrum and the spreading function is of cos-2s
type with s = 15 . Solid line: directional spectrum. Dash-dotted
line: directional spectrum, using frequency dependent spreading
function. 5

Analysis of data from the North Sea

Brodtkorb et al. (1999) analysed data from Gullfaks C in the
North Sea. The wave parameters Vcf = Ac =Tcd (rise speed)
and Hd , were considered, cf. Figure 5, and a joint probability 0
distribution was tted to data. In Figure 6, the theoretical joint 0 1 2
Velocity (m/s)
3 4 5

density (solid line) has been calculated given an estimated spec- Figure 6. Solid line: theoretical joint density of Vcf and 2Ac .
trum with two peaks. The data are marked as dots. Further, Dashed line: kernel density estimate of Vcf and Hd . Dots:
the kernel density estimates (dashed line) are estimated from the data from the North Sea.
square root of the data using an Epanechnikov kernel and the
one-dimensional
p optimal
p value for the smoothing parameter for
both Vcf and Hd , see also Silverman (1986). In these cases
it also ensures that the kernel density estimate is zero for nega-
tive values due to the compact support of the kernel used. The
nal estimate is then obtained by a simple change of variables.
For the kernel density estimate in Figure 6, the contour levels of
the kernel density estimates enclose the given percentage of the
data and have been calculated by splitting the bi-variate data,
(Vcf ; Hd ) , into groups formed by ranking the data by density
estimate.

Analysis of data from the Japan Sea

Wavep

Time

Another example, nicely demonstrating how data and theoreti-

Ac
Wave Hu
Hd
cal distributions t together, is based on 15 hours of data from
Yura in the Japan Sea. We have calculated the theoretical joint
Tcf Tcr
pdf of Tcf and Ac given that Tc = 5 s and present the re-
Atd Atu sult in Figure 7. For the empirical distribution the waves with
Tc
4:8 < Tc < 5:2 are used. A one-peaked spectrum was esti-
mated from data and used as input for the calculation of the pdf.
In Figure 7 is alsop shown a kernel density estimate of the pdf,
Td

Tu
estimated from ( Ac ; Tcf ) using an Epanechnikov kernel and
Figure 5. Wave parameters, denitions. In the list below, the one-dimensional optimal value for the smoothing parameter.

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 5 of 8

 Joint density of (Tcf,Ac,Tc=5) which damage calculations and fatigue life predictions can be
performed. In WAFO there are numerous routines for evaluat-
Level curves enclosing: ing fatigue measured loads, as well as making theoretical calcula-
5 10 tions of distributions that are important for fatigue evaluation.
30
50 A powerful technique when analysing loads is to use Markov
70
90 models as approximations, especially to model the sequence of
4 95
99
TP by a Markov chain. For such models there exist many ex-
99.9 plicit results. Here we will use this Markov approximation for
computing the intensity of rain ow cycles and trough-crest cycles
amplitude [m]

3 for the Gaussian model in Figure 2.

For fatigue analysis the rain ow cycle, dened in Figure 9,
is often used.
2 Mk

0
0 1 2 3 4 5 mk = mrfc
k
period [s]
m+k
Figure 7. Solid line: theoretical joint density of Tcf and Ac Figure 9. Denition of the rain
ow cycle, as given by Rychlik
given that Tc = 5 . Dashed line: kernel density estimate. Dots: (1987). From each local maximum Mk one shall try to reach
data from the Japan Sea. above the same level, in the backward(left) and forward(right)
directions, with an as small downward excursion as possible. The
minimum, of mk and m+k , that represents the smallest devi-
Spatio-temporal waves. Distribution of extremal crest ation from the maximum Mk is dened as the corresponding
heights. rain
ow minimum mrfc k . The k :th rain
ow cycle is dened as
Consider spatial waves evolving in time, spatio-temporal waves (mrfc
k ; Mk ) .
or space-time waves, which, from a probabilistic point of view, The Markov model is dened by the min-max pdf, which is
are modelled as Gaussian random elds W (x; t) . A wave of
this type is called an extremal wave if its crest height attains
a local maximum in time. Podgorski et al. (1999) discussed
dierent aspects of extremal waves, and computed the joint den-
ow cy-
sity function of extremal crest height and half-wave length. A
comparison of the result for one-dimensional (spatial) waves and
ow cycles can be found
spatio-temporal waves was made; cf. Figure 8. The spectrum
used as input was a JONSWAP spectrum with frequencies cut
ow cycles found in the simulated load signal are shown.
obtained from the power spectral-density by using approxima-
tions in Slepian model processes, see e.g. Lindgren and Rych-
lik (1991) and references therein. For the Markov model
there is an explicit solution for the intensity of rain
cles, see Frendahl and Rychlik (1993). By using the rou-
tines in WAFO the intensity of rain
using Markov approximation; see Figure 10, where also the
rain
o outside (0:5!p ; 1:5!p ) and signicant crest height 3.8 m. 6
8 8

7 7

6 6 4

5 5
H1 [m]

H1 [m]

4 4 2
3 3
max

2 2
0
1 1 Level curves at:

0 4.2994e05
0 50 100 150 200 250 0 50 100 150 200 250 8.5987e05
L1 [m] L [m]
1 2 0.00017197
0.00042994

Figure 8. Isolines of the joint density function of crest height

0.00085987
0.0017197

and half-wave length for spatial waves (left) and extremal waves
0.0034395
4 0.0051592

(right) for a Gaussian eld W (x;t) with a JONSWAP spectrum.

0.006879

Levels: 10 3 [6 4 3 2 1 0.75 0.5 0.25].

6
6 4 2 0 2 4 6
Fatigue & Cycles & Markov models min

In fatigue applications the exact sample path is not important, Figure 10. Intensity of rain ow cycles computed from the psd
but only the tops and bottoms of the load, called the sequence of through Markov approximation, compared with the cycles found
turning points (TP). From the TP cycles can be extracted, from in the simulation.

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 6 of 8

The WAFO toolbox also contains routines for computing the
intensity of rain
ow cycles for a switching Markov chain of TP.
In oceanographic applications one is often interested in
trough-crest cycles, which are a minimum and a maximum sep-
arated by a zero crossing, see Figure 11.
Mktc Mk
0 mk
mtc
k tion of stationary Gaussian process through circulant embedding
REFERENCES
Brodtkorb, P.A., Myrhaug D., Rue, H. (1999). Joint Distri-
bution of Wave Height and Wave Crest Velocity from Recon-
structed Data. Proceedings of the 9th ISOPE Conference, Vol
III, pp. 66-73.
Cavanie, A., Arhan, M. and Ezraty, R. (1976). A statistical
relationship between individual heights and periods of storm
waves. Proceedings, BOSS'76, Trondheim, pp. 354{360.
Dietrich C.R. and Newsam G.N. (1997). Fast and exact simula-
Figure 11. Denition of the trough-crest cycles (mtck; Mktc ) , of the Covariance matrix. SIAM J. Sci. Compt., 18, pp. 1088-
compared with the denition of min-max cycles (mk ; Mk ) . 1107.
Also for the intensity of trough-crest cycles there exists an ex- Frendahl, M. and Rychlik, I. (1993). Rain
ow analysis: Markov
plicit solution for Markov models, see Rychlik et al. (1997). method. Int. J. Fatigue, 15(4), 265-272.
In Figure 12 the intensity of trough-crest cycles, computed by
WAFO, are compared with the cycles found in the simulated Frendahl, M., Lindgren G. and Rychlik I. (1993). Fatigue Anal-
load. ysis Toolbox. Dept. of Math. Stat., Lund University.
6 Johannesson, P. (1998). Rain
ow Cycles for Switching Pro-
cesses with Markov Structure. Probability in the Engineering
and Informational Sciences 12(2), pp. 143-175.
4

Johannesson, P. (1999). Rain ow Analysis of Switching Markov

Loads. PhD thesis, Mathematical Statistics, Centre for Mathe-
2
matical Sciences, Lund University.
Krogstad, H.E. and Barstow, S.F. (1999). Directional Distri-
max

0
butions in Ocean Wave Spectra. Proceedings of the 9th ISOPE
Conference, Vol III, pp. 79-86.
Level curves at:

0.0010513
0.0021027
2
Lindgren, G. and Rychlik, I. (1991). Slepian models and re-
0.0042054
0.010513
0.021027
0.042054 gression approximations in crossing and extreme value theory.
4
0.084108
0.12616
0.16822
International Statistical Review 59, 195-225.
Longuet-Higgins, M.S. (1983). On the joint distribution wave
6 periods and amplitudes in a random wave eld. Proc. R. Soc.
6 4 2 0
min
2 4 6 A389, pp 24{258.
Figure 12. Intensity of trough-crest cycles computed from the Podgorski, K., Rychlik I. and Machado, U. E. B. (1999). Exact
psd through Markov approximation, compared with the cycles Distributions for Apparent Waves in Irregular Seas, 1-54. Ocean.
found in the simulation. Engng. to appear.
Podgorski, K., Rychlik, I., Ryden, J. and Sjo, E. (1999). How
ACKNOWLEDGEMENTS big are the big waves? Proceedings of the 9th ISOPE Conference,
Vol III, pp. 53-60.
Many persons have put a great deal of eort into WAFO and its
predecessors FAT and WAT. Among others, we are grateful to
ow cycle count-
Mats Frendahl, Finn Lindgren, and Ulla Machado, all at Math-
ematical Statistics at Lund University, and Sylvie van Iseghem,
IFREMER, Brest.
Rychlik, I. (1987). A new denition of the rain
ing method. Int. J: Fatigue 9, 119-121.
Rychlik I. and Lindgren G. (1993). CROSSREG { a tech-
The wave data from Gullfaks C were prepared and
made available by Dr. S. Haver, Statoil. The wave data
from Yura, Japan Sea were prepared and made available by
Dr. Sc. H. Tomita, Ship Research Institute, Ministry of Trans-
port.
nique for rst passage and wave density analysis. Probability in
the Engineering and Informational Sciences 7, 125-148.
Rychlik I. and Lindgren G. (1995). WAVE Analysis Toolbox
{ a Tutorial. Dept. of Math. Stat., Lund University.
The rst author was nancially supported by the Norwe-
gian Research Council (NFR), Det Norske Veritas and Lund Uni- Rychlik I., Johannesson P., Leadbetter M. R. (1997). Modelling
versity. and Statistical Analysis of Ocean-Wave Data using transformed

Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb Page: 7 of 8

Gaussian processes. Marine Structures, 10, 13-47.
Silverman, B.W. (1986). Density estimation for statistics and
data analysis. Monographs on statistics and applied probability,
Chapman and Hall.
Tayfun, M.A. (1981). Breaking limited waveheights. Jour-
nal of the Waterway, port and coastal and ocean division 107,
pp. 59-69.
Paper no. ISOPE 2000-GFC-02 Per Andreas Brodtkorb
Tayfun, M.A. (1990). Distribution of large waveheights. Journal
of the Waterway, port and coastal and ocean division 116, pp.
686-707
Torsethaugen, K. (1996). Model for a doubly peaked wave
spectrum. Report No. STF22 A96204. SINTEF Civil and En-
vironm. Engineering, Trondheim.
Wand, M.P., and Jones, M.C. (1995). Kernel smoothing. Chap-
man and Hall.
Page: 8 of 8