This article appeared in a journal published by Elsevier.

The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elseviers archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights

Author's personal copy

Renewable Energy 57 (2013) 594e605

Contents lists available at SciVerse ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Wave energy resources: Wave climate and exploitation

Jesus Portilla a, *, Jeison Sosa a,1, Luigi Cavaleri b, 2
a b

Universidad San Francisco de Quito (USFQ), Av. Diego de Robles y Va Interocenica, Campus Cumbay, Quito, Ecuador Istituto di Scienze Marine, Consiglio Nazionale delle Ricerche (ISMAR-CNR), S. Polo 1364, 30125 Venice, Italy

a r t i c l e i n f o
Article history: Received 6 September 2012 Accepted 25 February 2013 Available online xxx Keywords: Wave energy Wave climate Renewable energy Spectral energy distribution Power generation

a b s t r a c t
In identifying the most convenient zones for harvesting wave energy, it is natural to be attracted by the areas where we nd the highest mean energy values. The obvious examples are the storm belts. A more careful analysis reveals that for practical use other factors need to be taken into account. Some of the main ones are the energy spread in frequency and direction, and its seasonality, without discussing the cost of the structure basically related to the conditions to be withstood. This reveals that other areas, in particular the equatorial ones, can be conveniently used, and be possibly advantageous from various points of view. Based on the results of the ECMWF ERA-Interim reanalysis and of altimeter data, we have carried out a comparative analysis between two locations with opposite characteristics, in the North Atlantic and in the Equatorial Pacic respectively. The quantied results conrm that less energetic, but more regular and less extreme, areas have a potential comparable to that of the classically considered storm belts. 2013 Elsevier Ltd. All rights reserved.

1. Introduction Harvesting wave energy from the ocean is obviously a subject of interest. Taking for granted the present level of related technology (e.g., [1e4]), it is necessary to establish which are potentially the most promising areas. At a rst glance, it is natural to associate these areas to the parts of the ocean where we nd the highest levels of wave energy. In this paper, we show that a deeper analysis of the situation is required for an optimal choice, both from the point of view of production and for the related economic analysis. There are many challenges involved in the practice of wave energy harvesting. Some of them are technical, because the nature of wave energy is oscillatory, while standard technologies for electricity production involve rotational or linear generators. Wave energy converters (WEC) are conceived for carrying out this transformation. Others challenges are environmental, because wave energy does not come in a regular form. A normal sea state is composed by the superposition of a number of monochromatic waves. In order to convert energy efciently, ideally an optimal WEC should be able to interact with all of the small and large wave components. In practice, from a more realistic point of view, WECs

* Corresponding author. Tel.: 593 2 297 1700x1046. E-mail addresses: jportilla@usfq.edu.ec, jportilla@ymail.com (J. sosa.jeison@gmail.com (J. Sosa), luigi.cavaleri@ismar.cnr.it (L. Cavaleri). 1 Tel.: 593 2297 1700x1046. 2 Tel.: 39 041 5216810; fax: 39 041 2602340. 0960-1481/$ e see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.renene.2013.02.032

Portilla),

are restricted to work in specic ranges of frequencies and directions (e.g., see Ref. [5]). Other challenges involve the harsh environmental conditions at sea. WECs are exposed to corrosive saline water and to strong forces inherent to the water motion. In comparison to air for wind energy, the water density is three orders of magnitude larger and its associated energy is proportionally higher. However, the forces on the mechanisms and the related construction costs increase as well, generally with a power >1. For this reason, WECs cannot operate under strong wave conditions. Whenever a high sea state is expected, the device has to stop operations and protect itself, going in the so-called survival mode. On the other hand, WECs cannot operate if the energy is too low. A minimum of energy is necessary to start-up the system. These aspects naturally affect the performance, and therefore also the economical return of a related project. Apart from the WEC technological complexity and the variety of concepts developed to convert wave energy into electricity, several issues affecting WEC operation can be associated with the wave climate, which is the focus of this study. The advantage is that at present many environmental variables are understood with a very good degree of condence. Wave variables in particular, are routinely monitored from space and forecast by numerical models. In addition, meteorological centres like the European Centre for Medium-Range Weather Forecasts (ECMWF, Reading, U.K.) archive data over long periods of time. In this paper we carry out a comparative evaluation of wave energy resources, taking into account aspects, like those mentioned

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

595

above, that can be linked to the wave climate. Namely, we focus our attention on the distribution of energy over frequency and direction, and on the start-up and survival conditions. These considerations are accounted for by straightforward parameterizations established from the perspective of the wave climate. The paper is organized in six sections. In Section 2 we describe the main data sources, in Section 3 the general background for the calculation of the wave power is briey presented, in Section 4 we dene the statistics required for the description of the wave climate in the selected areas, and Section 5 contains the evaluation of the wave resources considering the above aspects. The output is translated into standard economical parameters like the annual production and the capacity factor to make the link with engineering applications. Finally in Section 6 a brief discussion summarizes our main ndings and conclusions. 2. Data sources

352.5 in steps of 15 (see Ref. [27] for details). Deep-water source terms account for wind input, non-linear resonant interactions, and white-capping dissipation. The data has global coverage with a spatial resolution of 0.5 in latitude and longitude. Although the wave spectrum is the actual variable of the model, it is typical to provide users with outputs in the form of integral parameters like signicant wave height (Hm0), mean wave period (Tm1,0), and mean wave direction (qm). The wave power, object of this study, is another integral parameter of the wave spectrum and can therefore be obtained accordingly (see Section 3). The period we considered covers 21 years from 1989 to 2010. This relatively long time series is necessary in order to be statistically representative. For the comparison between model and altimeter data, the following statistical parameters are considered: bias, root mean square error (RMSE), and scatter index (SI) dened as follows:

BIAS
Two data sources are used in this study. Namely, wave model results from ECMWF, and signicant wave height (Hs) data from satellite altimeters. Altimeter data is very attractive (see, e.g., [6,7]). There is a practically global marine coverage and data is available with continuity since the early 90s. Global Hs distributions have been derived, among them the one by Young and Holland [8] and the permanently updated one by Oceanor [9]. However, a strong limitation is that the information contains only Hs. Some attempts have been made to also derive wave period (see, e.g., [10]), but without much conclusive evidence, particularly in swell dominated areas. In any case, altimeter data lacks the fundamental information about direction. This limitation is very restrictive, since a correct assessment of the wave energy requires the knowledge of its distribution in frequency and direction. The Synthetic Aperture Radar (SAR) could in principle overcome this. However, sparseness in time and space, as is the case for altimeters, is inherent to this data. Moreover, other limitations apply to the SAR data. The rst one is the rather low cut-off frequency, only sufciently long waves can be observed accurately, see Refs. [11,12]. The second is the directional ambiguity, the SAR imagette from which the spectrum is derived is a frozen picture of the moving sea surface and therefore ambiguous over 180 . Although several studies have been conducted to tackle this aspect (e.g., [13,14]), this is still an open issue. The alternative data source, model results, has quite different characteristics, the main one being its continuity and homogeneity both in time and space. This adds to the fact that it is possible to assemble large model data-sets from historical runs (hindcasts). At present, numerical wave models are very robust and include the most relevant physical processes of the wave evolution, (see Refs. [15e18]). The obvious solution is to make use of both the sources, complementing the sparse and discontinuous, but measured, altimeter data with the continuous and homogeneous model one. For the present study, we use the ECMWF data derived from the ERA-Interim archive [19,20]. ERA-Interim is a reanalysis project aimed at providing long-term meteorological and oceanographic data with uniform accuracy and sufcient resolution. The ERAInterim reanalysis products have been extensively validated and veried (see, among others [21e24]). These and other studies have shown the high quality of the model data, both for meteorological and wave parameters. The wave model operational at ECMWF is WAM Cycle IV [15,25,26]. WAM is a state-of-the-art spectral wave model used at several meteorological centres around the world. WAM solves the wave energy balance equation dened in the spectral domain. For ERA-Interim the wave spectrum is discretized into 24 directions and 36 frequency bins. The frequency ranges from 0.0345 to 0.5476 Hz in geometric sequence, and the directions from 7.5 to

N 1 X y xi N i1 i

(1)

( RMSE

N 1 X y xi 2 N i1 i

)1=2 (2)

SI

RMSE jx j

(3)

with x the measured and y the model data. We use data from three different altimeters, ENVISAT, Jason and Jason-2, their orbits have (had for ENVISAT) different characteristics with return time of 31 days for ENVISAT, and 10 for the Jasons. This corresponds to an equatorial distance between adjacent orbits of 90 and 277 km respectively. For our discussion and to analyse different wave climates, we focus our attention on two areas, one in the North Atlantic (NA), and the other one in the Equatorial Pacic (EP). The different latitudes of the considered areas lead to a different, larger in the case of the NA, number of altimeter data. Notwithstanding the previous validation of the ERA-Interim reanalysis results (see the already quoted references), we have chosen to carry a devoted validation in the two areas of interest. Therefore, altimeter data concerning NA and EP have been colocated with the corresponding wave model results. The colocation areas correspond to boxes of 20 in latitude and longitude centred at the analysis points (see Table 1). The overall results of the comparison are reported in Fig. 1 and Table 2. At a summary look, it is clear that the low values of bias, RMSE and SI ensure that model data can be used condently for the present purposes. Nevertheless, it is worthwhile to point out some of their characteristics. The bias values reveal some differences among the three altimeters, a fact already discussed in the literature (see, among others, [22]). The results in Table 2 suggest that ENVISAT provides on average larger Hs, whence a negative bias of the model. The larger RMSE values at NA reect the higher, on average, Hs at this location. Possibly the larger biases at EP derive from the less accurate, with respect to the storm belts, modelling of the propagation and possible attenuation of swell over large distances, see Refs. [15,28] for discussion on the subject. Note, however, how these

Table 1 Geographical coordinates of the considered locations. name NA EP Area North Atlantic Equatorial Pacic Longitude 26 W 93 W Latitude 40 N 1 S

Author's personal copy

596

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

ENTRIES:

15
HS model (m)

(a)

1-4 4 - 11 11 - 36 36 - 116 116 - 378

378 - 1236
1236 - 4050

P rg

ZN Ef cg f df
0

(4)

10

where r is the sea water density, g the gravity acceleration. The spectral energy density E(f) and the group velocity cg, are both frequency dependant. In deep water, however, where the group velocity expression reduces to cg g/4pf, equation [4] is often given in terms of the signicant wave height (Hm0) and the mean wave period (Tm1,0),

NUM=251169 YM=2.75 STDY=1.39 XM=2.85 STDX=1.42 SSLO=0.97 RMSE=0.22 BIAS=-0.10 CORR=0.99 SI=0.07

rg2 2 H T W=m 64p M0 M1;0

(5)

0 0

10 HS ENVISAT (m)

15
ENTRIES:

(b)
HS model (m)

4 3 2 1 0 0

1-4 4 - 11 11 - 36 36 - 118 118 - 388

388 - 1276
1276 - 4200
NUM=81179 YM=1.86 STDY=0.41 XM=1.94 STDX=0.42 SSLO=0.96 RMSE=0.14 BIAS=-0.08 CORR=0.96 SI=0.06

Two things need to be mentioned about this expression. The rst one is that we have lost any information about directionality, not only on the main wave direction, but also about their directional spreading. Even if we put a directional device with the right orientation, expression [5] provides a higher than truth value, the more we are dealing with wind sea the higher the deviation. The second point concerns the units of equation [5], [W/m]. These are units of power over units of longitude, e.g., meter of crest. However, the actual power available to a given device derives from a complex relation among the characteristics of the wave system and those (dimensions, array distribution, principle of operation) of the device itself. For instance, a point absorber is so called if small compared with the wave dimensions, in such conditions, and if working in resonance with the exciting wave, its absorption length is 3l/p, with l the wavelength of the monochromatic exciter wave (e.g., [29,30]). In this paper, we have elected not to discuss the characteristics of the possible devices to be used, but to limit ourselves to the analysis of the power actually available in a given area. Thus, it is clear that while equation [5] can provide a rst guess (in excess) of the potential energy at a given location, a more correct analysis needs to go into the details of the corresponding spectra. This is the subject of the next section. 4. Wave climate Wave energy is the result of energy transferred from the atmosphere by wind. The distribution of storms over the globe is not homogeneous, but it is rather concentrated in specic areas mainly located in the mid-latitudes. These areas, commonly referred to as storm-belts, are characterized by an intense meteorological activity including weather fronts and extra-tropical cyclones (see, e.g., [31e 33]). There are strong seasonal changes in storm activity, with intense winter periods and calmer conditions during summer. Due to the distribution of the land masses, this seasonality is more marked in the northern than in the southern hemisphere [29]. In the tropical zone, the typical meteorological conditions range from moderate to low, with the most relevant activity represented by the formation of tropical cyclones, which are mainly episodic events. In this context, most of the related wave energy is found where the winter storms take place, i.e. the extra-tropical zones (see Fig. 2). While this is true as overall amount of energy, energy

HS ENVISAT (m)
Fig. 1. Scatter-plots of Hs for ENVISAT altimeter vs. model at the analysis locations. a) North Atlantic, and b) Equatorial Pacic.

differences tend to disappear when we consider the overall average values (last row in Table 2). Finally, the very low values of SI, point to a very strong consistency of the data. This is clearly visible in Fig. 1. Therefore, for the scope of the present study, the model results are considered as providing a sufciently good representation of the variables of interest. 3. Wave power calculation Wave power is an integral parameter of the wave spectrum representing the rate with which energy reaches a certain location. It is computed as the integral of the product between the group velocity and the spectral energy density as:

Table 2 Comparison of signicant wave height, between model output and altimeter observations for different missions. Locations NA and EP (see Table 1). Mission Period N [#] NA ENVISAT JASON-1 JASON-2 all 2002e2010 2002e2009 2008e2010 e 251,169 272,985 57,943 e EP 81,179 81,377 17,279 e Bias [m] NA 0.10 0.04 0.02 0.01 EP 0.08 0.10 0.06 0.03 RMSE [m] NA 0.22 0.27 0.30 0.26 EP 0.14 0.16 0.16 0.15 SI [%] NA 7.5 9.9 10.6 9.3 EQ 7.1 8.7 8.6 8.1

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

597

Mean Wave Power [kW/m] (19892010)

60
50

5 60 0
60
30
10

40 latitude 20

40

30

20

20

20

20

0 20

10

10

30

20
60

30
40
60

40

60 150 100 50 0 longitude 50 100 150

Fig. 2. Global distribution of the mean global wave power [kW/m] for the period 1989e2010.

60

40

50

30

40

50

50

30 50

harvesting and a related optimal approach may require a more general consideration. In particular, wave energy can be transported as swell over long distances. Consequently, the wave climate at a specic location is made of the locally generated wind-sea plus the swell energy coming from somewhere else. This characteristic of waves is important and has to be taken into account. While a whole range of situations is possible, for the purpose of discussing how the problem should be approached, we have chosen to focus on two rather opposite situations, the North Atlantic (NA) and the Equatorial Pacic (EP). Their substantially different characteristics as wave power is concerned are clearly visible in Fig. 2. More specically, NA is characterized, especially in the winter months (October to March), by severe and prolonged storms, the wave heights in the most northerly part of this area being considered among the highest ones on the planet. On the contrary EP is virtually storm absent, the local energy being associated to the long period swell that constantly reaches the area. 4.1. Wave power time-series For a discussion about the possibility and the convenience of harvesting wave energy in different situations it is useful to step back from the summarizing data in Fig. 2 and to focus our attention on the original time series, at the chosen locations, NA and EP. Their geographical coordinates are given in Table 1. The NA point is near the Azores Islands, the EP one close to the Galapagos Islands, both

with surrounding depths larger than 3000 m. Purposely, and to be as general as possible, we have selected points in deep water. Focussing on shallow or coastal areas would have implied specic case analyses and a lack of generality. Nevertheless, it should be mentioned that the method of analysis we have followed is valid for any location at arbitrary water depths. In Fig. 3a the wave power time series for the whole period of study are given. A smaller time window is presented in Fig. 3b to show some details of the time series. Table 3 gives some statistics of the time series, including the ones for Hm0, Tm1,0, and wind speed (U10). From Figs. 2e4 and Table 3, the differences in wave power between NA and EP can be appreciated. The differences in seasonality are quite evident as well, particularly in Figs. 3 and 4. At NA large events occur during the northern winters (90th percentile 101.7), while wave activity during the northern summer is limited (10th percentile 6.9). The EP time-series show less energy content and less variability (90th percentile 31.6, 10th percentile 11.0).
Table 3 Statistical indicators for wave and wind conditions at the analysis locations in North Atlantic (NA) and Equatorial Pacic (EP) (see Table 1 for their geographical coordinates). Variable Wave power [kW/m] Mean Median Maximum Percentile 90 Minimum Percentile 10 Hm0 [m] Mean Median Maximum Percentile 90 Minimum Percentile 10 Tm-1,0 [s] Mean Median Maximum Percentile 90 Minimum Percentile 10 U10 wind speed [m/s] Mean Median Maximum Percentile 90 Minimum Percentile 10 NA 44.5 25.4 1118.6 101.7 1.2 6.9 2.7 2.4 12.5 4.4 0.6 1.4 9.2 9.0 16.4 11.4 5.2 7.3 7.5 7.1 23.9 12.3 0.0 3.1 EP 20.3 18.5 110.2 31.6 3.7 11.0 1.9 1.9 3.8 2.3 1.0 1.5 10.7 10.6 15.8 12.2 7.0 9.3 4.3 4.4 9.3 6.1 0.0 2.2

Fig. 3. Time series of wave power at the analysis locations. a) For the whole study period, b) for a time window including two boreal winters.

60

60

50

50

30

20

Author's personal copy

598

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

100 wave power [kW/m] 80 60 40 20 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Equatorial Pacific North Atlantic

Fig. 4. Seasonal variability of the monthly mean wave power at the analysis locations.

There is a weak seasonal variation at EP driven mainly by the seasonality of the southern hemisphere where most of the energetic local swell comes from Antarctica. At the NA location, in spite of the occasional large values of wave power, the mean value is comparatively low (44.5), slightly over twice the mean value of the EP (20.3). In contrast, the NA maximum in the 21 year period (1118.6) is about 10 times higher than at EP (110.2). The median values at the two locations (NA 25.4, EP 18.5) are in absolute terms more comparable. Having obtained from Figs. 3 and 4 a general idea of the seasonality of the two wave conditions, it is convenient to go further analysing the characteristics of the wave spectra at the two locations. We begin in the next (4.2) subsection with an analysis of the two-dimensional spectra. While more complete, the 2D information is not easily summarized at a glance into the relevant gures. Therefore in the following subsection (4.3) we focus on the statistics that we can derive from the frequency spectra. 4.2. 2D spectral statistics Statistical properties of the 2D spectrum have been calculated for the two locations. The statistics shown in Fig. 5 are calculated for
0 330 300 0.3 0.2 30 3.5 60 3 2.5 90 2 1.5 240 210 150 120 1 0.5
240 210
2

each frequency-direction bin independently. They include the mean, median spectrum, and the relative probability of occurrence above a certain threshold. This threshold corresponds to the 10th percentile of the spectral energy in all bins. Note that for the spectral plots we have followed the nautical convention. This means that zero degrees is the geographical north and waves shown in certain direction are going in that direction. Please note also that, for better visualization, due to the different footprint domain of the indicators, the frequency domain of the relative occurrence plots (panels 3 in Fig. 5) ranges from 0 to 0.6 Hz, while the range for the mean and the median (panels 1 and 2) goes from 0 to 0.3 Hz. From Fig. 5 (panels a) we see that the wave conditions at NA consist of two main wave systems. The predominant system is broad, both in frequency and in direction, with relatively low frequency. This indicates mature wind-sea characteristics, most likely corresponding to old wind-seas because of its proximity to the generation zone. Its propagation direction is about 130 (from Northwest). The secondary system is narrow, also with low frequencies. This suggests swell characteristics. The propagation direction is 0 (from South). The magnitude of this secondary system is low compared to the wind-sea-like system. Therefore, its footprint in the mean and median spectra (panels 5a1,a2) is weak, but it is clearly visible in the occurrence probability graph (panel 5a3). This graph shows also that waves occur over the whole range of spectral directions. These occurrences correspond most probably to occasional wind-sea events. They contain energy at high spectral frequencies that contribute only marginally to the wave climate. Panels 5b show that the wave conditions at EP are characterized by four different wave regimes. Two of them propagate towards the North, two towards the South, all with swell characteristics. The predominant system is the one propagating to 30 (from Southwest). The second one ows to 130 (from Northwest). The third system is the broadest one, propagating to
0.3
0

4
330 300

0 30 1 60 0.8 90 0.6 0.4 120 0.2 150

210 240 270 300 330

30 60

24.65 22.04 18.13 14.33 10.64 120 6.88 3.33

0.2

(a1)
0.1 270

(a2)
270

0.1

(a3)
0.2

0.6 0.4 0.2 90

150

mean [m2.s/rad] North Atlantic 40N 26W

median [m .s/rad] North atlantic 40N 26W

occurrence probability [%] NA 40N 26W

0 330 30 3 300 60 0.3 90 1.5 240 2.5 2

270 300 330

0 30 2 60 0.3 90 1 240 1.5

270 300 330

0 30 60 24.73 22.01 18.15 0.2 0.4 0.6 240 120 10.59 7.08 2.96 90 14.48

270

0.1

0.2

0.1

0.2

(b1)
210 150

120

1 0.5

(b2)
210 150

120 0.5

(b3)
210 150 occurrence probability [%] EP 1S 93W

mean [m2.s/rad] Equatorial Pacific 1S 93W

2 median [m .s/rad] Equatorial Pacific 1S 93W

Fig. 5. Statistics of the 2D wave spectrum for the analysis locations, NA (panels a), EP (panels b). The directions shown are ow direction, i.e. where the waves are going to.

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

599

340 (from Southeast). This is the most related to the local wind conditions. Finally, a marginal swell system propagates to about 180 (from North). From the panels 5b we see that, although the wind sea energy (high frequency bins) does occur frequently at EP (see panel b3), its magnitude is relatively low. Indeed, as seen in panels 5b1 and 5b2, most of the wave power is associated to the swell systems.

4.3. 1D spectral statistics Additionally to the information provided in the previous section, we now proceed to analyse some statistics of the non-directional spectra. These are presented in Fig. 6. Namely we show the mean and the median 1D spectra projected into frequency (panels 1), and into direction (panels 2), in panels 3 how much (in percent) each

Fig. 6. Statistics of the 1D spectrum for the two analysis locations NA (panels a), EP (panels b).

Author's personal copy

600

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

energy-frequency pixel contributes to the overall energy, and in panels 4 the percent occurrence probability of the different pixels. If not otherwise specied, all the quotations in this subsection refer to Fig. 6. From panels a1 and b1 it is obvious that the NA mean spectrum is much larger than EP. However, the corresponding median values are similar (actually larger at EP). This implies a much larger spread of the possible values at NA. At EP, panel b1, the minor wind system (see previous subsection) is visible as a small hump in the median spectrum. Although summarized into the integrated parameters, similar information was already available from Table 3. A similar larger spreading in direction at NA is clearly visible from a2 and b2. At EP the energy ows along well dened directions (swells). Additional to the overall information in these panels, Fig. 7a, b, offer a picture of the seasonal variability of the directional spectrum. At NA, although the main propagation direction is always consistent (from Northwest), there is a considerable difference between the peak in winter (at 112.5 ) and in all other seasons (at 142.5 ). At EP, during the boreal winter the Northwesterly wave system contributes most of the energy, while in all other seasons the South-westerly system is the dominant one. This

information is valuable when considering reorientation capabilities for the devices (see scenario C, Section 5.3). Having obtained a general idea about the spectra, now we go deeper in the analysis asking how much (a3, b3, in percents) a certain level of energy at a given frequency contributes to the overall budget and (a4, b4) how frequently (percents) this happens. At NA the energy (a3, a4) is dispersed in an approximate range from 0.07 to 0.3 Hz, although most of it is between 0.1 and 0.2 Hz. However, the most recurrent frequencies, but not the most energetic ones, are between 0.15 and 0.25 Hz. The highest values of energy density, above 5 m2 s, do contribute to the relative budget, but only marginally, as their frequency of occurrence is rather low. At EP the energy contribution (b3) and the frequency of occurrence (b4) distributions are more similar than at NA. This is favourable for energy harvesting. There are two ranges of practical interest, namely 0.07e0.11 Hz, energetic, but not very frequent waves, and the, less energetic, but more frequent, 0.13e0.21 Hz range. Note that most of the latter energy is associated to the swell as low frequency part of the related spectra, hence rather constant and well dened in direction. 5. Comparative analysis

North Atlantic

60 50 energy density [m .s]

2

5.1. Background and outline

DecJanFeb MarAprMay JunJulAug SepOctNov

40 30 20 10 0

(a)

50

100

150 200 250 frequency [Hz]

Equatorial Pacific

300

350

30 25 energy density [m .s]

2

DecJanFeb MarAprMay JunJulAug SepOctNov

20 15 10 5 0

From the analysis in the previous section, it is clear that integrated parameters, as the overall or mean energy, do not provide the information required for a proper estimate of the usable energy at a given location. This energy depends on its distribution in time, in frequency and direction, and, last but not least, on the characteristics of the device we are planning to use. While the latter information is fundamental in any exploitation plan, our present purpose is to provide a more general view of the problems and limitations encountered in exploiting the wave power in different areas. Granted the knowledge of the local wave climate, and without entering into the details of the single device, the usable power is a direct function of the installed capacity that can therefore be considered as a variable of the system, but not necessarily the only one. With this approach, it is possible to introduce in the analysis other practical operational conditions, like start-up and survivability thresholds, or conditions on directionality and frequency ranges of operation. The advantage is that the estimates are independent from the device. The disadvantage is that the units of the resulting variables remain as units of power over units of longitude. This would be a shortcoming if we wished to obtain actual economical gures, but it is immaterial for comparative purposes. In order to get closer to engineering applications, we express the integrated resource in terms of variables commonly used in the electricity generation sector, like annual electricity production and capacity factor, dened in the following. 5.2. Denitions Nominal installed capacity (IC): In power generation, the nominal capacity of any generating device refers to the total amount of electricity that would be produced if the device was operating at its full nameplate capacity during 100% of the time. It is expressed in units of power over the period considered, typically one year (i.e., MWh/m/year). Annual electricity production (AP): it is the actual total gross electricity generation produced by the device in a period of time (one year). It is measured in units of power over time (i.e., MWh/m/ year).

(b)

50

100

150

200

250

300

350

frequency [Hz]
Fig. 7. Seasonal variations of the directional spectrum. a) North Atlantic, b) Equatorial Pacic.

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

601

Capacity factor (CF): The capacity factor is the ratio between the annual electricity production and the nominal capacity. Since the annual production is associated to the revenues, and the nominal capacity is associated to the initial investment, CF gives an indication of how well is the investment used. 5.3. Assumptions For the evaluation of electricity production, we consider three basic scenarios. For different values of installed capacity, the energy is integrated over the whole study period and scaled down to one year in order to calculate the annual average. The three scenarios are now further detailed. Scenario 1: The blunt assumption is that for a given installed capacity, the device can convert the whole energy locally available, with as only upper limit the installed capacity. When there is extra energy available this cannot be converted. Scenario 2: The assumptions of scenario 1 are not entirely realistic in the sense that i) for the lower limit, the device cannot be triggered to work however low the available energy. In practice the device would start-up only above a certain threshold; ii) for the upper limit, if wave conditions are much stronger than those the device has been designed for, operations must be halted, and the device has to enter into survival mode. For our purposes these conditions can be represented as a function of IC. Three levels of lower and upper limits are considered (i.e., 10%, 20% and 30% of IC), tpgoto "and in order to simplify the analysis, for each case both the lower and upper limits are represented with the same percentage (the sensitivity of each threshold limit is assessed separately and presented in different gures). For instance, for the 20% case and a nominal installed capacity of 100 kW/m, if the resource is lower than 20 kW/m, there is no production because the device cannot be set to work. If the resource is higher than 120 kW/m, the production is again nil because the device has to enter survival mode. The system can only produce electricity between these two limits. Scenario 3: For a still more realistic scenario, besides the survival and trigger-to-action limits, we consider the directional and frequency limits within which a device is able to harvest energy. For instance, the WaveDragon [34] is an overtopping device with a frontal opening for the capture of waves and a pair of arms to enhance this capture. It is clearly a directional device. Other devices like attenuators (e.g., Pelamis [35]), are relatively long and must be aligned according to the wave propagation for operation. At the opposite, it is natural to quote the point absorber devices that in principle can capture wave energy from any direction. However, in practice, either due to the mooring system or because they are installed in arrays, they too will be subject to directional restrictions. The same is true for frequency. As an example, WECs are designed to resonate only within a nite range of frequencies. In this scenario, we represent these limitations setting frequency-direction limits to the energy of the available spectrum. We consider three cases: A) we assume that the range of directions within which the device is able to capture energy is 60 wide. Similarly, the range of frequencies is 0.09 Hz. Case B is more restrictive, the two ranges being reduced respectively to 45 and
Table 4 Settings for scenario 3. S&S are the startup and survival thresholds. Case Directional range of operation (deg) Restriction A B C C winter 60 45 60 60 NA 97.5 / 157.5 97.5 / 142.5 112.5 / 172.5 82.5 / 142.5 EP 7.5 / 52.5 7.5 / 52.5 7.5 / 52.5 112.5 / 172.5

0.075 Hz. In case C we analyse the possible capability of the device to change orientation, this in order to optimize energy capture according to the varying climate conditions. We acknowledge that the choice of the frequency-direction limits is rather arbitrary, based on the discrete frequency-direction grid of the model. However, we believe they represent standard enough situations, and these considered cases are intended to provide an idea of the sensitivity of the results also to these limitations. As for the start-up and survival thresholds, in the three cases, they are specied as in scenario 2, with a 20% value. With these assumptions, it is clear that the frequency (characteristic of the device) and direction (its orientation) at each location must be chosen carefully in order to optimize the capture of energy. Using the results of the spectral analysis in Sections 4.2 and 4.3 we can decide, for each considered case, the optimal strategy at the two selected locations. At NA, it is intuitive that we must focus on the main wave system. The limitations stem from the fact that the spectrum is broad in both frequency and direction. Quite a bit of energy is therefore lost. Direct inspection, and trials and errors indicate that for case A the best ranges at NA are 97.5 e157.5 in direction and 0.0673 Hz e 0.1586 Hz in frequency. For case B these values change to 97.5 e142.5 and 0.0673e0.1442 Hz respectively. At the EP location, the appropriate choice is to focus on the main wave system because it is the more energetic and seasonally consistent. Therefore, the corresponding values for A) are from 7.5 to 52.5 in direction, and from 0.0505 to 0.1442 Hz in frequency, and, case B), 7.5 e52.5 and 0.0556 Hz e 0.1311 Hz. For case C, looking at Fig. 7a, at NA, more energy is coming in winter from 112.5 , while during all other seasons the peak is at 142.5 . Similarly at EP (Fig. 7b), during the boreal winter most energy is coming from 127.5 , while during the other seasons the South-westerly wave system (from 22.5 ) is always more energetic. Therefore, in case C the device is reoriented on a seasonal basis according to this information. All other settings are the same as for case 3A. Table 4 summarizes the settings for this scenario. 5.4. Results Scenario 1 e For any installed capacity, IC there is also the limit to the annual electricity production AP given by the local wave climate. Therefore, as expected (see Fig. 8a), AP grows with IC. However, only till what provided by the local wave climate. At least from this point of view, it makes no sense to invest at EP into systems with IC > 30 kW/m. The opposite is true for NA where the availability of very severe storms makes the curve grow until very large values. The CF (capacity factor) curves (see Fig. 8b) have a completely different trend. The maximum values are found at low IC, when the device will be operational at nearly full capacity the whole year. For higher IC, idle or not full use periods will occur more often, with a consequent decrease of CF. This means that from the economical point of view, it is not interesting to increase much the installed capacity because the capacity factor drops quickly, making the investment less protable. The decay at EP is steeper because saturation of annual production is reached at lower IC.

Frequency range of operation (Hz) Restriction 0.090 0.075 0.090 0.090 NA 0.067 0.067 0.067 0.067 / / / / 0.159 0.144 0.159 0.159 EP 0.051 0.056 0.051 0.051 / / / / 0.144 0.131 0.144 0.144

S&S (%IC) 20 20 20 20

Author's personal copy

602

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

400 Annual production [MWh/m/year] Annual production [MWh/m/year]

300 250 200 150 100

20

(a)

10 20 30

300

(a)
200

10

100

50
30

50 100 150 Installed capacity [kW/m]

200

0 0

50

100

150

200

Installed capacity [kW/m] 0.7 0.6 Capacity factor 0.5 0.4 0.3 0.2 0.1
10 20 30 10 20 30

1 Equatorial Pacific North Atlantic 0.8 Capacity factor 0.6 0.4 0.2 0 0 Equatorial Pacific North Atlantic

(b)

(b)

50

100

150

200

Installed capacity [kW/m]

Fig. 8. Results of annual production (a) and capacity factor (b) for scenario 1.

0 0

50

100

150

200

Installed capacity [kW/m]

Fig. 9. Results of annual production (a) and capacity factor (b) for scenario 2.

It is instructive to make a quantitative example on the use of these curves. Let us assume IC 20 kW/m. From the two panels of Fig. 8 for NA we derive AP 144 (MWh/m/year) and CF 0.81. The corresponding values at EP are AP 148 and CF 0.83. Hence the two locations behave very similarly. However, at NA, there is still room for extra capacity, but at the price of a reduced CF. The reason is the relatively slow growth of the NA curve in panel 7a. It follows that, to double the production up to 300 (MWh/m/year), the installed capacity needs to be four times larger, with a corresponding CF 0.42. Scenario 2 e The main consequence of the introduction of startup and survival restrictions is that the curves of both AP and CF (Fig. 9a, b) show a growth-decay behaviour clearly different from the previous scenario. While Fig. 9 offers a general view of the combined effect of the two thresholds, it is important to understand also their individual effect. This is provided by Fig. 10aed. The slower growth of AP derives mainly from the low energy waves not being usable due to the threshold for start-up. In Fig. 10a, as the magnitude of the lower threshold increases, AP decreases. The upper threshold has the counter effect and it is also less strong (Fig. 10c). Please note the different AP vertical scale with respect to scenario 1 in Fig. 8. As expected, the higher IC and the higher the percent for triggering the system to work (10, 20, or 30%), the higher the threshold, hence the stronger the effect. This is more evident at NA (compare panels 8a and 9a) because here we rely substantially on high energy systems. Conversely, for moderate and

large IC values, the introduction of the survival limit forces the system to disregard large seasonal values. It has therefore the effect of saturating the AP curves at higher IC values (Fig. 9c). Obviously, the growth-decay characteristics of the AP curves in panel 9a are reected into similar ones on how CF depends on IC (panel 9b). Again, note the different vertical scale with respect to panel 8b, and also the strong reduction of the peak values, especially at NA. Similarly, the sensitivity of the CF curve to the thresholds is consistent with the sensitivity of the AP curve (Fig. 10b, d). Given the curves in Fig. 9a, b, we can derive for instance indications on the optimal values of installed capacity (IC) which do not necessarily match for the two indicators AP and CF. For the 20% threshold, the optima IC at EP are 46 and 24 kW/m, at NA these values correspond to 160 and 30 kW/m. It is important to note that at NA the AP curve attens well before the peak is reached (around 90 kW/m). This saturation behaviour and the slower growth of the curve are inherent to the NA wave climate. This anticipates the fact that the optimal installed values will most likely be for much lower IC as it is reected by the CF curve (Fig. 9b). Scenario 3 - The main effect of introducing restrictions in the frequency-direction domain with respect to the previous scenario is a strong reduction in absolute values of both AP and CF (see Fig. 11a, b). This is especially true for cases A and B. The energy harvest in case C, is signicantly higher at both locations (compared to case A), indicating rst that orientation is a crucial parameter,

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

603

Sensitivity to lower threshold

Sensitivity to lower threshold

300 Annual production [MWh/m/year] 250 200 150 100 50 0 0 Equatorial Pacific North Atlantic

0.7 0.6 Capacity factor 0.5 0.4 0.3 0.2 0.1

200
50% 5%

Equatorial Pacific North Atlantic

5%

(a)

(b)

50%

50 100 150 Installed capacity [kW/m]

Sensitivity to upper threshold

0 0

50 100 150 Installed capacity [kW/m]

Sensitivity to upper threshold

200

300 Annual production [MWh/m/year] 250 200 150 100 50 0 0 Equatorial Pacific North Atlantic 50 100 150 Installed capacity [kW/m] 200
100%

1 Equatorial Pacific North Atlantic 0.8 Capacity factor

(c)
5%

0.6 0.4

100%

(d)

5%

0.2 0 0

50 100 150 Installed capacity [kW/m]

200

Fig. 10. Sensitivity to lower (a, b) and upper (c, d) cut-off thresholds for scenario 2.

and most importantly that reorientation of the device according to the seasonal wave climate conditions will have a signicant impact both in terms of production and capacity factor. Regarding the shape, the curves keep their growth-decay type. All this is natural because in this scenario, only part of the energy in the spectrum is available for conversion, and the survival and threshold restrictions apply in the same manner as in scenario 2. For practical use, the most relevant fact is the quantication of the impact of the different restrictions at the two considered locations. Again, please note the different vertical scales with respect to Fig. 9a, b. In Fig. 11a, the AP drops to about 30% at NA for cases A and B, and only to about 75% for case C. At EP, the production drops to about 50% (compared to scenario 2), while for case C it drops only to about 85%. A similar decrease is observed for CF in cases A and B, while this indicator stays at levels comparable to scenario 2 for case C (Fig. 11b). Comparing the performance at the two locations, perhaps the most remarkable fact is that the NA production curves are lower than the EP ones for cases A and B, and comparable (in the grow section and also in the maxima) for case C. Given the much higher energy potentially available at NA, this may be surprising. Basically this stems from the different characteristics of the two wave systems. As detailed in Section 4, when discussing the directional spectra, the wind-sea like systems at NA contains energy that is spread in the frequency-direction domain. On the contrary the energy of the swell systems at EP is more concentrated, a fact that enhances its possible exploitation. 6. Summary and discussion We make a comparative assessment of the wave energy practically available for conversion at two locations characterized by

rather opposite wave climates. In general, the economic viability of a WEC project shall depend on the balance between associated costs and benets. Both depend on the characteristics of the local wave climate and on the capacity of the system to be installed. Without focussing on a specic device, we have preferred a more general approach that is valid, mutatis mutandis, for any device to be installed at a specic location. In addition, to avoid any consideration specic to coastal geometry, we have purposely kept our analysis in the open ocean. However, the method of analysis should be valid for any location at arbitrary water depths, provided that the data sources are representative for the location. They should come, for instance, from a nearshore wave model that accounts properly for shallow water effects (e.g., bottom friction, refraction, depth induced breaking, .), or alternatively from local wave spectral measurements. This is important to note, since many wave energy extraction applications will be located in shallow waters. To be as general as possible, we have used the installed capacity IC of the system as a variable, and explored how the annual production of electricity AP and the capacity factor CF (an indicator of how efciently is the system used) vary when varying IC. We have used three possible scenarios, with progressively increasing and realistic restrictions. Starting with no restriction, except of course IC, we have then introduced minimum and maximum thresholds representing respectively the trigger into action and the survival conditions. Finally we have considered specic windows in frequency and direction where the system can actually operate, looking additionally into the possible reorienting capability of the device. Granted the amplitude of their range (Hz and degrees), for each location the specic choice has been optimized according to the local wave climate.

Author's personal copy

604

J. Portilla et al. / Renewable Energy 57 (2013) 594e605

200
Annual production [MWh/m/year]
C

150
C

(a)
A B A

100

50

50 100 150 Installed capacity [kW/m]

200

0.7 0.6 Capacity factor 0.5 0.4 0.3 0.2 0.1 0

A C

Equatorial Pacific North Atlantic

(b)

meaningful assessment can only be made by taking into account the most relevant aspects of that complex system. Purposely, and because of the uncertainty in the related gures, we have not considered the economical aspect of the problem. On a parallel, but different subject, the cost of an offshore wind energy system is currently quoted at 2500 Euro/kW (or w3000 US$/ kW, Brian Holmes, personal communication). However, there are so many feed-in tariffs on green trading, renewable energy certicates, etc., that any gure becomes a little nebulous. In any case, it is clear that the costs of construction and maintenance of a structure increase with the severity of the conditions of operation and, much more so, with the ones to withstand for survival. The implications for the two areas we have considered are obvious. Our purpose was not to force a choice between the two opposite situations. Many other factors control the choice of where to operate, rst of all what is available in the seas surrounding a specic country. We meant to show how to interpret the local wave climate, not only for its power, but also for its distribution in frequency, direction and time, to derive a better idea of the true potential in a certain zone. It turns out that, besides the classically considered storm belts, other areas can be similarly attractive simply because the lower energy level is compensated by a greater potential efciency due to wave characteristics more uniform in time and to the lack of severe conditions, which is of course advantageous also from the economical point of view. Acknowledgements

B C A B

50 100 150 Installed capacity [kW/m]

200

Fig. 11. Results of annual production (a) and capacity factor (b) for scenario 3.

The results show that the mean energy locally available, however signicant, is not a sufcient indicator for a rational nal decision. The spreading of energy in frequency and direction and its seasonality play a fundamental role in dictating what can be harvested by a given system and, most of all, the efciency of the overall investment. It turns out that, granted the interest for the stormy areas, also the equatorial zone can be protably, and in some aspects convenient for electricity production, using the local wave motion. The basic reason for this is that the higher average energy in the storm belts are due to infrequent large winter storms which can only be captured by installing larger capacity devices. However, most likely these devices will be idle in summer. Conversely, in the equatorial zone, there is a permanent and relatively constant, albeit lower, wave motion, but at low frequencies, hence with large wave power. In this regard, the average wave power as a mapping indicator is misleading. Other indicators can be found that are more objective. For instance, Barstow et al. [36] presented two. The rst is the ratio of the minimum monthly mean power to the annual mean power (their gure 4.6). This indicator quanties the seasonal variability and it is therefore associated to the economic advantage of the project. The second is the ratio of extreme wave height to the mean wave height (their gure 4.8). This indicator gives an idea about the ratio of construction costs over revenues. It should be mentioned that these two indicators show a picture of the global resource that is in opposition to that brought across by the average power. However, given the complexity of wave energy, an objective and

The model data used for the overall statistical analysis has been obtained from the ERA-Interim archive of the European Centre for Medium-Range Weather Forecasts. Altimeter data has been obtained from the GlobWave project. We gratefully acknowledge the help of Luciana Bertotti in retrieving the data and of Claire Nicklin and Cesar Zambrano for revising the text. J.Portilla acknowledges the nancial support of USFQ, INOCAR (project CD-INOCARLOG031-11), and the GlobWave project. He is also grateful for the stimulating support and benets derived from attending the WISE meetings and discussing with many distinguished colleagues. J.Sosa acknowledges the support of the Marine Renewable Energy Center. L.Cavaleri acknowledges the support of the EU funded project MyWave (SPA 2011.1.5e03) within the 7th Framework Programme. We are also very thankful with the anonymous reviewers for their committed and constructive comments, which were very valuable in the preparation of the nal manuscript. References
[1] Cruz J. Ocean wave energy, current status and future perspectives (green energy and technology). Springer; 2008. p. 431. [2] Drew B, Plummer AR, Sahinkaya MN. A review of wave energy converter technology. Journal of Power and Energy 2009;223(8):887e902.. http:// dx.doi.org/10.1243/09576509JPE782. [3] Falco AFO. Wave energy utilization: a review of the technologies. Renewable and Sustainable Energy Reviews 2010;14:899e918. [4] EMEC, The European Marine Energy Centre Ltd (online), http://www.emec. org.uk/wave_energy_developers.asp, [accessed 06/2012]. [5] Korde UA. Efcient primary energy conversion in irregular waves. Ocean Engineering 1999;26(7):625e51. [6] GlobWave DUE, Deliverable D.16, satellite wave data quality report, GlobWave/DD/SWDQR, Issue 1.2; 2010. [7] Queffeulou P, Croiz-Fillon D. Global altimeter SWH data set, May 2010. Technical report. Ifremer; 2010. [8] Young IR, Holland GJ. Atlas of the oceans: wind and wave climate. Elsevier Science, ISBN 0-08-042435-X; 1996. p. 241. [9] Oceanor (online), http://www.oceanor.com/Services/wwa_info/, [accessed 31.07.12]. [10] Gommenginger CP, Srokosz MA, Challenor PG. Measuring ocean wave period with satellite altimetry: a simple empirical model. Geophysical Research Letters 2003;30(22):2150.. http://dx.doi.org/10.1029/2003GLO17743.

Author's personal copy

J. Portilla et al. / Renewable Energy 57 (2013) 594e605 [11] Hasselmann K, Hasselmann S. On the nonlinear mapping of an ocean wave spectrum into a synthetic aperture radar image spectrum and its inversion. Journal of Geophysical Research Oceans 1991;96(C6):10,713e10,729.. http:// dx.doi.org/10.1029/91JC00302. [12] GlobWave DUE. Deliverable D.9, wave data handbook. GlobWave/DD/WDH Issue 1.0 2011. [13] Schulz-Stelleneth J, Lehner S, Hoja D. A parametric scheme for the retrieval of two-dimensional ocean wave spectra from synthetic aperture radar look cross spectra. Journal of Geophysical Research Oceans 2005;110:C05004.. http://dx.doi.org/10.1029/2004JC002822. [14] Johnsen H, Engen G, Collard F, Kerbaol V, Chapron B. Envisat ASAR wave mode productsdquality assessment and algorithm upgrade. In: Proceedings of SEASAR 2006. Eur. Space Agency Spec. Publ., ESA SP-613; 2006. [15] WISE Group. Wave modelling e the state of the art. Progress in Oceanography 2007;75(4):603e74.. http://dx.doi.org/10.1016/j.pocean.2007.05.005. [16] Holthuijsen LH. Waves in oceanic and coastal waters. Cambridge University Press; 2007. [17] Janssen PAEM. Progress in ocean wave forecasting. Journal of Computational Physics 2008;227:3572e94. [18] Ardhuin F, Rogers E, Babanin A, Filipot J-F, Magne R, Roland A, et al. Semiempirical dissipation source functions for ocean waves. Part I: denitions, calibration and validations. Journal of Physical Oceanography 2010;40:1917e41. [19] Simmons A, Uppala S, Dee D, Kobayashi S. ERA-Interim: new ECMWF reanalysis products from 1989 onwards in ECMWF Newsletter No. 110 2006. [20] Dee D, Uppala S, Simmons A, Berrisford P, Poli P, Kobayashi S, et al. The ERAInterim reanalysis: conguration and performance of the data assimilation system. Quarterly Journal of the Royal Meteorological Society 2011;137:553e 97.. http://dx.doi.org/10.1002/qj.828. [21] Caires S, Sterl A. Validation of ocean wind and wave data using triple collocation. Journal of Geophysical Research 2003;108:3098,.. http://dx.doi.org/ 10.1029/2002JC001491. [22] Janssen PAEM, Abdalla Saleh, Hersbach Hans, Bidlot Jean-Raymond. Error estimation of buoy, satellite, and model wave height data. Journal of Atmospheric and Oceanic Technology 2007;24:1665e77. [23] Bidlot JR. Intercomparison of operational wave forecasting systems against buoys: data from ECMWF, MetO_ce, FNMOC, MSC, NCEP, MeteoFrance, DWD, BoM, JMA, KMA, Puerto del Estado, DMI. November 2009 to January 2010.

605

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33] [34] [35] [36]

Technical report, Joint WMO-IOC Technical Commission for Oceanography and Marine Meteorology, 2010. Tavolato C, IsaksenL. Data usage and quality control in ERA-40, ERA-Interim and the operational DA system, ECMWF Technical report, ERA Report Series; 2011. WAMDI group, Hasselmann S, Hasselmann K, Bauer E, Janssen P, Komen G, et al. The WAM model e a third generation ocean wave prediction model. Journal of Physical Oceanography 1988;18:1775e810. Komen GJ, Cavaleri L, Donelan M, Hasselmann K, Hasselmann S, Janssen PAEM. Dynamics and modelling of ocean waves. Cambridge, England: Cambridge University Press; 1994. ECMWF, IFS. Documentation e Cy36r1, operational implementation 26 January 2010. PART VII: ECMWF WAVE MODEL, http://www.ecmwf.int/ research/ifsdocs/CY36r1/WAVES/IFSPart7.pdf; 2011. Ardhuin F, Chapron B, Collard F. Observation of swell dissipation across oceans. Geophysical Research Letters 2009;36:L06607.. http://dx.doi.org/ 10.1029/2008GL037030. Falnes J. Ocean waves and oscillating systems: linear interactions including wave-energy extraction. Cambridge, UK: Cambridge University Press, ISBN 982-207-002-0; 2002. Salter SH, Cruz J, Lucas J, Pascal R. Wave powered desalination, in macroengineering seawater in unique environments. Berlin Heidelberg: Springer. http://dx.doi.org/10.1007/978-3-642-14779-1_29; 2011. p. 657e74. Keable M, Simmonds I, Keay K. Distribution and temporal variability of 500 hPa cyclone characteristics in the Southern Hemisphere. International Journal of Climatology 2002;22:131e50. Hoskins BJ, Hodges KI. New perspectives on the Northern Hemisphere winter storm tracks. Journal of the Atmospheric Sciences 2002;59(6). ISSN: 15200469:1041e61. Alves JH. Numerical modeling of ocean swell contributions to the global windwave climate. Ocean Modelling 2006;11:98e122. WaveDragon (online), http://www.wavedragon.net/, [accessed 2012]. Pelamis Wave Power (on line), http://www.pelamiswave.com/, [accessed 2012]. Barstow S, Mrk G, Mollison D, Cruz J. The wave energy resource [Chapter 4]. In: Cruz J, editor. Ocean wave energy. Berlin: Springer. p. 93e132, http://dx. doi.org/10.1007/978-3-540-74895-3; 2007. 2008.