22.02.2013 Views

RESEARCH FOR

RESEARCH FOR

RESEARCH FOR

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

probabIlIty dIstrIbutIons of Wave HeIgHts In tHe<br />

lItHuanIan coast<br />

egidijus Kasiulis<br />

Aleksandras Stulginskis University, Lithuania<br />

e-mail:egidijus.kasiulis@gmail.com<br />

abstract<br />

Since discovering that signals of random waves submit to the known laws of probability, this became widely used<br />

in engineering and energetics for probability distributions analysis of wave height. From an energetic point of view,<br />

it is necessary to know the average wave height in, for example, highly wavy (1% probability), medium wavy (25%<br />

probability) or non-wavy (95% probability) years. Whereas, maximum multi-year value of wave height characteristics<br />

is essential for engineering resistant wave energy converters that could withstand severe marine conditions. Average<br />

and maximum annual values of wave height data collected from Klaipėda coastal hydrometeorological station are used<br />

for this study. Probability distributions of average and maximum wave heights in the Lithuanian coast are analysed in<br />

this paper. The best fitting is obtained using HYFRAN and EASY FIT software. Both, a test for independence (Wald-<br />

Wolfowitz) and stationarity test (Kendall) are carried out for every time series using HYFRAN software. Maximum<br />

likehood method is selected for distribution estimation. Fitting is determined using chi-square test and the best fitting<br />

is verified with comparison (BIC and AIC) criterion. Fitting for one of the most commonly used distributions in the<br />

analysis of wave climate – Rayleigh distribution – cannot be determined with HYFRAN software. For this purpose,<br />

EASY FIT software is used additionally. The fit of the distribution is evaluated via the chi-square test similarly.<br />

Calculated wave heights based on lognormal probability distribution that fits best according to HYFRAN software<br />

are similar to those calculated using Rayleigh probability distribution.<br />

Key words: probability distributions, Baltic Sea, wave height.<br />

Introduction<br />

The statistical theory of random signals for<br />

electrical noise analysis was developed in 1944 by S.<br />

O. Rice (Rice, 1944), M. S. Longuet-Higgins applied<br />

this theory to the random water surface elevation<br />

of water waves. It emerged that the parameters of a<br />

random wave signal follow known probability laws<br />

(Longuet-Higgins, 1952). As a result, the analysis<br />

of probability distributions of wave heights in<br />

engineering and energetics is used to the present day.<br />

Gumbel distribution was used to determine wave<br />

climate and wind speed parameters during the period<br />

of tropical hurricanes in the Caribbean Sea in 2002<br />

(Calverley et al., 2002). Generalized extreme value<br />

(GEV) distribution was used for the coastal region of<br />

East Anglia (UK) in 2010 to analyze extreme events<br />

and estimate climate change implication on inshore<br />

waves and the occurrence of extreme events (Chini<br />

et al., 2010).<br />

An attempt to fit normal distribution for the Black<br />

Sea waves in Filyos region (Turkey), where a new sea<br />

port is planned, failed. It was estimated that statistical<br />

distribution of surface profile of the records containing<br />

extreme waves deviates from normal distribution<br />

(Bilyay et al., 2011).<br />

It is common that in various marine areas a number<br />

of distributions is used for different studies. For<br />

example, at the end of last century increased marine<br />

activities like offshore mineral and oil exploration,<br />

utilization of wave energy, construction of marine<br />

structures and harbors caused a demand for accurate<br />

information about wave climate in the Arabian Sea.<br />

WATER MANAGEMENT<br />

The available atlases of averaged visual wave statistics<br />

at that time provided information that differed from<br />

one another (Muraleedharan et al., 1990).<br />

In 1990 the comparative study of distributions<br />

of wave heights from these atlases was made for an<br />

area off Trivandrum. The long-term distributions of<br />

significant wave heights were tested with Weibull,<br />

Gumbel, Rayleigh, exponential and lognormal models.<br />

The best fit was obtained for Weibull probability<br />

density function (Muraleedharan et al., 1990).<br />

Weibull and Rayleigh probability distributions<br />

are most commonly used for analysis of wave<br />

heights. This is not accidental as Rayleigh probability<br />

distribution is often characterized as the special case<br />

of Weibull distribution.<br />

The demerits of Rayleigh distribution are that<br />

it does not always fit for the nearshore conditions<br />

and in some cases the enlarged values of maximum<br />

wave heights can be obtained. Therefore, Rayleigh<br />

distribution is often modified for the nearshore<br />

conditions (Thornton et al., 1983). There are even<br />

attempts to incorporate both distributions and to adapt<br />

complex Rayleigh-Weibull distribution (Mai et al.,<br />

2010).<br />

Forecasting of the wave heights while considering<br />

waves from energetic point of view is important. This<br />

is essential both for evaluating energetic potential<br />

and for designing and maintenance of wave energy<br />

converters.<br />

The objective of this article is to test probability<br />

distributions for wave heights and to obtain best fitting<br />

in the Lithuanian nearshore conditions. In particular<br />

152 ReseaRch foR RuRal Development 2012

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!