Tm

I

ube

tew

a

JOURNAL OF RENEWABLE AND SUSTAINABLE ENERGY 2, 043108 �2010�

1

he wave energy resource along Australia’s Southernargin

M. A. Hemera� and D. A. GriffinCentre for Australian Weather and Climate Research: A Partnership between CSIRO andthe Bureau of Meterolology and the CSIRO Wealth for Oceans National ResearchFlagship, Hobart, Tasmania 7001, Australia

�Received 23 March 2010; accepted 26 June 2010; published online 5 August 2010�

The Southern Australian margin is one of the most energetic regions in the worldsuitable for the extraction of wave energy for electricity generation. We have pro-duced a data set in which the deep-water wave energy resource for the region isdescribed by three representative deep-water wave states, equivalent to the 10th,50th, and 90th percentiles of the deep-water wave energy flux, derived from ar-chives of the USA National Oceanic and Atmospheric Administration �NOAA�WaveWatch III �NWW3� operational wave model. The Simulating WAves Near-shore �SWAN� wave model is then applied along the full Southern Australianmargin to propagate these representative wave states into the near-shore regionto quantify the effects of shallow water processes such as refraction, shoaling,and bottom friction. The wave energy incident on the 25-m isobath��30–50 kW /m� is approximately 35%–50% less than the World Energy Councilestimates of offshore wave energy but is approximately 20% greater than the en-ergy observed from long-term buoy deployments on the midshelf. The latter dis-crepancy is attributed to an overestimation of significant wave height along theSouthern Australian margin by the NWW3 model. The near-shore model applied inthis study adequately simulates the attenuation of wave heights across the conti-nental shelf when compared with estimates of wave height attenuation obtainedfrom the Topex satellite altimeter. The attenuation of wave energy across the con-tinental shelf reduces the estimates of offshore wave energy as given by the WorldEnergy Council; however the wave energy resource incident on the Southern Aus-tralian margin remains considerable. We estimate that if 10% of the incident near-shore energy in this region, which is an ambitious target when conversion effi-ciency is considered, were converted to electricity, approximately 130 TW h/yr�one-half of Australia’s total present-day electricity consumption� would beproduced. © 2010 American Institute of Physics.�doi:10.1063/1.3464753�

. INTRODUCTION

The Australian Government has set a 20% target for renewable energy by 2020 to expand these of renewable energy as part of its commitment to reduce Australia’s greenhouse gas emissionsy 60% on 2000 levels by 2050.1 To achieve this goal, an additional 45,000 GW h/yr of renewablenergy will have to be produced by 2020.

Wave energy has a number of advantages over other renewable energy sources. Environmen-ally, impacts occur during the construction and installation processes. Once in operation, wavenergy has the potential to provide a clean source of energy with no greenhouse gas emissionshile posing minimal impacts on the environment.2 A noteworthy advantage of wave energy

�

Author to whom correspondence should be addressed. Electronic mail: mark.hemer@csiro.au.

2, 043108-1941-7012/2010/2�4�/043108/15/$30.00 © 2010 American Institute of Physics

cistscnrtegK

wmearetpib

itdeNsmsctroS

anpmt

I

swccewp

043108-2 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

onversion devices is that many have a low visibility profile. Seafloor designs are essentiallynvisible, while surface designs protrude only a few meters above the ocean surface and are barelyeen when installed several kilometers offshore. These features make wave energy more appealingo nearby residents who have objected to wind turbines �and solar farms� on the grounds that theypoil coastal areas of high recreational value. Objections to wave energy farms have been made byompeting industry sectors �e.g., transport and fishers� on grounds of the potential hazard toavigation and access to resources, recreational users �e.g., surfers� on grounds of the potential foreducing incident coastal wave energy, and public utilities commissions �e.g., electricity corpora-ions� on grounds of the high cost per unit energy associated with high maintenance expenses inarly stages of the industry. Uncertainties remain on the effects of long-term deployment of waveeneration devices on the marine and coastal environments. The Wave Hub site in the Unitedingdom is providing a useful test bed for research in this field.3

A fundamental aspect to tapping wave energy is resource characterization. Initial estimates ofave energy potential have identified the Southern margin of the Australian continent as one of theost energetic wave climates suitable for wave energy generation.4,5 The World Energy Council4

stimates, however, are based on deep-water information, while wave energy generation systemsre typically positioned relatively near to shore, with the “offshore” systems positioned in depthanges of 30–50 m.6 Wave energy developers therefore require assessments of near-shore wavenergy that take into account the various transformations that take place as waves propagate acrosshe continental shelf. In this work, we aim to refine the estimates of near-shore wave energyotential for the world-class wave energy resource along the Southern Australian margin, takingnto account the redistribution of wave energy density over the best available shallow waterathymetry �at 0.01° spatial resolution� for this 3000 km stretch of coast.

We have used the long-term archives of the USA National Oceanic and Atmospheric Admin-stration �NOAA� WaveWatch III �NWW3� operational wave model,7 which Hemer et al.8 iden-ified as the most suitable wave model for the Australian region, to characterize the offshore,eep-water wave climate. Using this analysis of the deep-water wave data, the dominant andxtreme sea-states were identified and propagated into the near-shore using the Simulating WAvesearshore �SWAN� wave model,9 which accounts for refraction, shoaling, bottom friction, and

heltering from coasts and islands. SWAN and WaveWatch III are both widely used spectral waveodels that have been validated in a wide range of situations. Both models are governed by the

ame principle, where the evolution of the wave spectrum in space and time is described byonservation of action density being balanced by source terms representing generation, dissipa-ion, and wave-wave interaction processes. The action density is the energy density divided by theelative frequency. WaveWatch III tends to be more efficient at global scales, whereas SWANffers advantages at smaller scales10 and its specific consideration of shallow water processes. TheWAN model was implemented for this study to make use of these advantages.

The resultant map of wave energy resource is expected to complement the other issues thatrise when determining suitable locations to extract wave energy such as proximity to ports,avigation routes, and aquaculture and fishing areas, as well as proximity to the energy grid. Thisaper is organized in five sections as follows. This Introduction is followed by Sec. II outlining theethods applied. Results of the study are presented in Sec. III. Section IV discusses the results in

he context of Australia’s energy use, and Sec. V presents the main conclusions of the study.

I. METHODS

Hemer et al.8 assessed a number of available wave models for the Australian region for theiruitability to describing the Australian wave climate. The archived output of NOAA’s operationalave model, an implementation of the WaveWatch III model �NWW3 �Ref. 7��, was shown to

ompare best with waverider buoy data from 16 sites located on exposed sections of the Australianoast. However, NWW3 has two distinct limitations as an information resource for assessing wavenergy potential. First, the model archives include only integrated wave parameters: significantave height �Hs�, which describes the largest 1/3 of the waves in the instantaneous wave field;

eak wave period �Tp�, which describes the most energetic waves in the spectrum; and peak wave

ds�da

A

euawoAcETaw

m1i

wfuAwczacP�f=b

�Sdwn

E1w

043108-3 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

irection ��p�, which describes the direction of waves corresponding to Tp. The directional wavepectra are not available. Second, the spatial resolution of the NWW3 model is 1° latitude

1.25° longitude, and thus any climatology derived directly from the NWW3 archives is notirectly applicable to the near-shore zone until the cross-shelf wave transformations are taken intoccount.

. Deep-water wave energy

Spectral wave measurements along the South-Western Australian margin �here defined asxtending from Perth to Hobart� indicate that the wave climate in this region is dominated bynimodal energy within the swell spectral band �i.e., waves with period greater than 8 s �Refs. 11nd 12�� The region is considered near enough to the extratropical storms, which generate theaves in the Southern Ocean, so that a little separation between the sea and swell bands isbserved in the wave spectrum.11 The wave climate along the South-Western margin of theustralian continent is therefore adequately described as a unimodal wave climate, which can be

haracterized by the integrated wave parameters in the NWW3 archives. In contrast, the South-astern Australian margin receives waves from a large latitudinal range13 spanning the Southernasman Sea to the Coral Sea in the North. As a consequence, a bimodal wave state often exists, sorchives of spectrally integrated parameters �such as from NWW3� cannot accurately describe theave climate along the South-Eastern Australian coast.

Using the 10-yr �1997–2006�, 6-hourly archives of Hs, Tp, and �p from the NWW3 waveodel, a time-series of wave energy flux, EF, is calculated at all NWW3 grid points in the domain

10–155° E, 30–45° S, assuming a theoretical Pierson–Moskowitz14 �PM64� spectral shape,15

.e.,

EF = �1��2EPM�Hs,Tp,��cg��,h�d� , �1�

here EPM�Hs ,Tp ,�� is the PM64 wave energy spectrum as a function of Hs, Tp, and waverequency � and cg is the group wave speed as a function of � and water depth h and is determinedsing the approach described in Hemer et al.16 Water depth is derived from the Geoscienceustralia �GA� 0.01° bathymetry database,17 averaged over the NWW3 grid cell. Tp, the onlyave period parameter archived from NWW3, is not the preferred wave period parameter for

onducting energy resource assessments. Ideally, the wave energy period, Te, derived from theeroth and first negative moments of the frequency spectrum, is used for wave energy resourcessessments, as Te corresponds to the weighted average of the wave energy. We overcome this dataonstraint by assuming a theoretical PM64 spectrum. Assuming deep-water wave conditions, theM64 spectrum introduces the approximate relationship between Tp and the mean wave period Tz

Tp=1.4Tz�, such that Eq. �1� reduces to the International Energy Agency’s recommended equationor determining wave energy flux from integrated wave parameters18 �i.e., EF=0.577 TzHs

2

0.412 TpHs2�. By using Eq. �1�, the need to assume deep-water wave conditions, which may not

e appropriate for long swell as occurs on the South Australian continental shelf, is removed.The choice of a unimodal PM64 spectrum is appropriate for the fully developed unimodal sea

swell� observed along Australia’s South-Western margin. In regions with bimodal seas �e.g., theouth-Eastern Australian continental margin�, the recent development of theoretical multipeakedirectional wave spectra19 may enable extension of the approach taken in this study. In this study,e have limited our analysis to unimodal seas applicable to the South-Western Australian conti-ental margin.

The 10th, 50th, and 90th percentiles of EF are determined at all NWW3 grid points �EF,10,m,

F,50,m, and EF,90,m, respectively� for each of the 12 months, m, of the year, and for the whole0-yr archive �m= ��. These parameters summarize the spatial and seasonal distribution of the

ave energy resource for the Southern Australia deep-water region.

B

spwaapA

m0ar

izeb

Fci

Fr

043108-4 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

. Near-shore wave energy

As waves travel over the continental shelf toward shallower water, they start to feel theea-bed once the water depth is less than half the wavelength.20 A number of depth-dependentrocesses become important, including refraction, shoaling, and bottom friction. These shallowater processes may modify the spatial distribution of wave energy, typically observed as an

ttenuation of energy across the continental shelf. To quantify the wave energy in the near-shorend identify locations optimal for wave energy extraction, we applied the SWAN wave model toropagate specified wave conditions from deepwater into the near-shore regions of the Southernustralian margin.

Ten midresolution �0.05° and �5 km� model domains were configured along the Southernargin of the Australian continent �Fig. 1�. Bathymetry of each domain was derived from the

.01° GA bathymetry. Each of the 0.05° model domains was chosen so that the seaward boundaryligned with the center of NWW3 grid points near to the continental shelf edge. A directionalesolution of 5° and 35 frequency bins over the range of 0.033–0.5 Hz were specified.

Hemer21 calibrated the wave model for the grid which contains the Cape de Couedic waver-der buoy. Model parametrizations of bed friction, wind generation, and whitecapping parametri-ations, the directional spreading of the boundary input directional wave spectrum, and the pres-nce of quadruplet or triad wave-wave interactions were investigated. Calibration was carried outy minimizing error in the wave height simulation, being considered more important than the

IG. 2. Map of Topex tracks over Southern Australian margin, used to assess attenuation of wave height over theontinental shelf. The positions of the DW and SW points are indicated by circles on each track. The Topex pass number

IG. 1. SWAN grids applied for near-shore wave energy atlas. Coarse line represents 0.05° grid domains; fine linesepresent 0.01° grid domains. The 200-m bathymetric contour is also shown.

s indicated.

wstfswbwttt

slce

wdamp

Pbp

0ttrd

043108-5 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

ave period. Using each parametrization, the model was run in stationary mode for 44 individualea-states, with the timing of these runs corresponding to the time of satellite altimeter passes overhe model domain. Hemer21 found that in this region, wave conditions were well described usingorcing from swell boundary conditions only �with no local wind forcing�, with a root-mean-quare error �Hs� of 0.5 m. In these circumstances, wind-growth and whitecapping source termsithin SWAN are unnecessary, and Hemer21 found that the calibration was largely insensitive toed friction parametrization at these scales. The frictional drag law of Collins22 was implemented,ith a friction parameter, f , of 0.01 defined. Under these conditions, SWAN was most sensitive to

he directional spreading factor �cosine power, p� of the incoming swell at the open boundaries ofhe domain, and a value of p=0.2 provided best model-data agreement. Following these calibra-ion experiments, the same parameters were chosen for all model grids.

The SWAN wave model was run in stationary mode for 39 representative sea-states, corre-ponding to EF,10,m, EF,50,m, and EF,90,m �for m=1. .12 and � as defined above� for each midreso-ution domain. For the EF,X,m case �where X is the percentile 10, 50, or 90�, representative waveonditions are chosen as the mean of those waves, which lie within the representative band ofnergy flux, i.e.,

Hs,X,m = mean�Hs�tX,m�� ,

Tp,X,m = mean�Tp�tX,m�� ,

qp,X,m = arg�mean�ei�p�tX,m��� ,

here tX,m are the times in the 10 year archive for which the month is m �m=1, . . . ,12 or m= �, asefined above� and EF is within five percentile levels of the Xth percentile of EF. For example, thennual wave height associated with the 10th percentile of wave energy flux �EF,10,�� was deter-ined as the mean of all wave heights for which wave energy flux was greater than the 5th

ercentile of all EF values and less than the 15th percentile of all EF values.Boundary conditions for the midresolution domain runs were then defined by assuming a

M64-shaped spectrum and linearly interpolating between NWW3 grid cells along the openoundary. Landward of the nearest grid cell to the coast is assumed to have constant waveroperties.

Fine resolution �0.01°� near-shore wave properties were determined by nesting a total of 47.01° model domains within the ten midresolution 0.05° grids, which spanned the Southern Aus-ralian coast �Fig. 1�. Default one-way SWAN nesting was used to force the 0.01° model runs, andhe same SWAN parameters were used within the 0.01° model runs, as tuned for the parent 0.05°un. Therefore, in total, 2223 stationary SWAN model runs were carried out made up from 57

TABLE I. Overview of waverider buoy data sets used for validation.

LocationLatitude

�° S�Longitude

�° E�Water depth

�m� Dates Data source

Cape Sorell �CS� 42.15 145.01666 100 11-Jul-1985 CSIRO

24-Sep-1992

42.12 145.03 100 23-Mar-1998 BoM

31-Dec-2006

Cape de Couedic �CdC� 36.07 136.62 80 01-Nov-2000 BoM

31-Dec-2006

Cape Naturaliste �CN� 33.44 114.77 50 07-Nov-1998 WADPI

31-Dec-2006

omains �ten 0.05° grids+47 0.01° grids� for three sea-state scenarios �10th, 50th, and 90th

pcR

C

ta�omw

tm

Fpia

043108-6 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

ercentiles� for 13 temporal periods �m=1, . . . ,12 and m= �, as defined above�. The computationalost was approximately 40 h on a single 3.16 GHz Intel Xeon CPU, requiring approximately 6 GBAM, and produced approximately 4 GB data.

. Validation of wave height and energy flux maps

Waverider buoy data from three sites located along the Southern Australian margin were usedo validate the wave energy maps �Table I�. Again, only integrated wave parameters �Hs and Tp�re archived from the waveriders, and consequently wave energy flux was determined using Eq.1�, assuming a PM64 spectral shape. 10th, 50th, and 90th percentiles of wave energy flux at eachf the three sites were determined for m=1, . . . ,12 and m= � �as defined above� and compared toodeled EF values from the closest 0.01° model grid cell. The corresponding Hs and Tp valuesere also compared. No directional data is available for model comparison.

In order to assess the ability of the SWAN wave model to attenuate the wave heights acrosshe continental shelf, we compared atlas-derived significant wave heights with wave heights esti-

ated from TOPEX/Poseidon satellite altimeter records. Comparisons were made at offshore

IG. 3. Contour plots of the West Tasmania �43° S, 143.75° E� Hs vs Tp �left� and Hs �radial distance� vs �p �angle� �right�robability density functions, using NWW3 daily estimates of Hs, Tp, and �p for 1997–2006. The heavy black circlesndicate the 10th, 50th, and 90th percentile values �small, medium, and large circles, respectively� of Hs, Tp, and �p applieds boundary conditions for the 0.05° SWAN model.

FIG. 4. As for Fig. 3 for South Australian site �37° S, 135° E�.

�wa�HTpcteedft

I

A

043108-7 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

DW� sites and continental shelf �SW� sites, which lie on eight specified ascending satellite tracks,hich pass over the Southern Australian margin �Fig. 2�. Significant wave heights were obtained

t each of the 16 locations from the altimeter record for all available cycles between 1 and 481from October 1992 to October 2005�. Data were rejected using the same procedure as specified inemer et al.23 and corrected using the calibration equations derived by Challenor and Cotton.24

he altimeter passes over each point once every 10 days. At each point, the 10th, 50th, and 90thercentiles of wave height were determined �m= ��. These altimeter-derived HS percentiles wereompared with corresponding estimates from the atlas, as was the ratio of the DW:SW HS in ordero compare the modeled and observed rates of cross-shelf attenuation as well as the absolutenergy levels. Recall that the atlas-derived percentile values are derived from percentiles of thenergy flux and not just HS. We have measured the impact of the other factors �period andirection� in the energy flux by comparing variously derived percentiles using time-series datarom the NWW3 model and found less than 1% difference between the different ways of defininghe HS percentiles, so we treated the definitions as if they are equivalent.

II. RESULTS

. Adequacy of method to describe distribution of sea-states

Figures 3–6 display plots of NWW3 Hs versus Tp and Hs versus �p bivariate probability

FIG. 5. As for Fig. 3 for West Australian site �34° S, 113.75° E�.

FIG. 6. As for Fig. 3 for New South Wales site �34° S, 152.5° E�.

dc�tattcp

Fdel

043108-8 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

istribution functions �PDF� for four example NWW3 grid cells that were used to define boundaryonditions for the near-shore models. Also shown on these plots are the corresponding Hs, Tp, and

p values associated with the three energy flux percentiles. For the Southern margin, �Figs. 3–5�,he wave conditions associated with the 10th, 50th, and 90th percentiles of wave energy fluxdequately describes the distribution of wave properties at these sites, as evidenced by the fact thathe black dots lie in, or close to, high density regions of the wave parameter space. The same is notrue, however, for the South-Eastern region �Fig. 6�, where the distribution of wave properties islearly bimodal, comprising a mix of eastward and westward propagation and short and long

IG. 7. Monthly time-series comparison of wave energy flux between SWAN model output �solid line� and waverider buoyata �dashed line� at Cape Sorell �top�, Cape de Couedic �middle�, and Cape Naturaliste �bottom�. Upper two curves ofach plot correspond to 90th percentile values. The middle two curves correspond to the 50th percentile values, and theower two curves correspond to the 10th percentile values.

eriods. Recalling that the data set analyzed here is unimodal on a day by day basis, this property

ofswcfitSo

B

wf�6opted

tloa

tabs

T�p

E

C

C

C

H

C

C

C

T

C

C

C

043108-9 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

f bimodality can only have come from the time dimension of the NWW3 data set. Analysis of theull spectral information would doubtlessly show an even greater spread of the wave directiontatistics, which points to the relative importance of local wind forcing. Our omission of localind forcing in the nested SWAN models is consistent with a focus on areas where the wave

limate is unimodal and dominated by remotely generated waves. These conditions are not satis-ed East of the Southern tip of Tasmania, so we have restricted the domain of our output data sets

o the region where our approach is most valid, namely, the South-Western shelf as defined above.ince this is the region of greatest wave energy in Australia, we think it is an appropriate regionn which to focus for the purposes of the Australian renewable energy industry.

. Validation of wave height and energy flux maps

The wave energy flux predicted by the near-shore high-resolution SWAN models exceeds theave energy flux estimated from buoy data, for each of the three sites for almost the entire year,

or each sea-state case �10th, 50th, and 90th EF percentiles; Fig. 7�. For the 50th percentilemedian� all-month estimates of energy flux, the relative errors of the model are 27%, 29%, and% at the three buoy locations �Table II�. This is due to the modeled wave heights exceeding thebserved heights by 16%, 17%, and 9%, an error that is partially offset by the modeled waveeriods being too short �by 4%, 6%, and 9%� compared to the observations. The relative errors ofhe 10th percentile values are larger, while those of the 90th percentile values are smaller. Forxample, the 90th percentile model energy flux at Cape Sorell is 10% larger than the buoy estimateue to the height being 8% too great but the period 4% too short �Table II�.

In the deep ocean, where the best available validation data type is satellite altimetry, we findhat the model estimates of wave height exceed the satellite estimates �Table III� by 5%–10% at allocations for the 10th percentile values and West of 135° E for the median values. The relative sizef the 90th percentile values is inconsistent �alternating from track to track�, but the relative errorsre all less than 8%; many less than 5%.

Recall that a principal motive of the present work was to model the attenuation of HS acrosshe continental shelf for all locations across the South of Australia. Table III lists the ratio of deepnd shallow water wave heights at the locations of the satellite tracks. The model and altimeter-ased estimates of the attenuation rate agree well, with consistent values of about 0.9 for narrow-

ABLE II. Comparison of SWAN model and waverider buoy estimates of the energy flux �EF�, significant wave heightHS�, and peak period �TP� at three locations: Cape Sorell �CS�, Cape de Couedic �CdC�, and Cape Naturaliste �CN�. Theercentage error �PE� of the model is defined as 100��model-buoy� /buoy.

10th percentile 50th percentile 90th percentile

Atlas BuoyPE�%� Atlas Buoy

PE�%� Atlas Buoy

PE�%�

F �kW/m�S 19.4 12.6 53 51.6 40.6 27 136.0 123.4 10

dC 16.5 11.0 49 43.9 33.9 29 118.6 99.0 20

N 13.8 12.3 13 41.7 39.1 6 117.1 114.8 2

s �m�S 2.11 1.60 32 3.16 2.73 16 4.90 4.53 8

dC 1.94 1.47 32 2.88 2.45 17 4.51 3.94 14

N 1.81 1.54 17 2.73 2.50 9 4.37 4.09 7

p �s�S 10.2 11.5 �11 11.9 12.4 �4 12.9 13.5 �4

dC 10.2 11.6 �12 11.9 12.7 �6 12.9 13.9 �7

N 9.5 11.3 �16 11.9 13.1 �9 12.9 14.6 �11

helf regions and 0.7 or 0.8 where the shelf is wider.

C

1ssl�aFrE

cdnpptTidennob

043108-10 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

. Wave energy maps

Our wave energy resource atlas consists of 156 maps of 0.01° resolution over the domain13–147° E, 44.5–29° S. The 156 maps are made up from four variables �wave energy flux,ignificant wave height, peak wave period, and peak wave direction� for 39 representative sea-tates. These sea-states are defined as representing one of three �10th, 50th, or 90th� percentileevels of the wave energy flux for each of the individual months and for all months combinedm=1, . . . ,12 and m= �, respectively�. Figure 8 shows maps of the wave energy flux for the threell-month �m= �� percentile levels. Results for individual months are viewable via the Australianederal Government’s Renewable Energy Atlas �http://www.environment.gov.au/sustainability/enewable/atlas/index.html� and a CSIRO Marine and Atmospheric Research Ocean Renewablenergy website �http://www.marine.csiro.au/~griffin/ORE, see Supplementary Material25�.

An inspection of the maps reveals both the value and limitations of our approach. The prin-ipal motivation of this work was to provide accurate estimates across Southern Australia of theegree of attenuation suffered by the deep ocean waves as they cross the continental shelf into theear-shore. This effect is seen quite clearly. There are also, however, a number of modeling and/orlotting artifacts that the user of the maps should bear in mind. An obvious one is that the waveroperties are discontinuous where the various model domains overlap. A second is the quantiza-ion of the peak wave period in the nested SWAN models �see Tp plots in supplementary material�.he least obvious artifact is that islands and headlands cast unrealistically sharp “shadows.” This

s because the 50th percentile map shows, for example, what the wave properties would be on aay when the offshore wave height was equal to the median value and the direction and periodqual to the corresponding values. The shadows would be more diffuse if we had run the manyested models for many actual years and then computed the statistics from the time histories of theested models. Such an approach would be much more computer intensive, but the varying anglesf incidence would allow time-averaged rather than “snapshot” estimates of the wave properties to

TABLE III. Significant wave heights �meter� from Atlas and Topex altimeter along Australian Southern margin.DW points are deep-water wave values �i.e., taken from nearest NWW3 grid point�, SW points are points on thecontinental shelf at a location where altimeter data remains reliable �i.e., sufficiently far from the coast�. Theattenuation ratio �HS_SW/HS_DW� is shown in brackets. Altimeter derived wave heights are corrected accord-ing to Challenor and Cotton �Ref. 24�.

TP pass Data

10th pc 50th 90th

HS_DW HS_SW HS_DW HS_SW HS_DW HS_SW

225 TP 2.22 1.94 �0.87� 3.70 3.41 �0.92� 5.65 5.33 �0.94�ATLAS 2.32 2.16 �0.93� 3.59 3.25 �0.91� 5.71 5.04 �0.88�

73 TP 2.01 1.79 �0.89� 3.24 2.90 �0.90� 5.29 4.64 �0.88�ATLAS 2.10 2.01 �0.96� 3.14 2.98 �0.95� 5.05 4.57 �0.91�

251 TP 1.97 1.69 �0.86� 3.23 2.88 �0.89� 4.91 4.55 �0.93�ATLAS 2.13 1.96 �0.94� 3.23 2.97 �0.92� 5.19 4.58 �0.88�

23 TP 1.81 1.68 �0.93� 2.93 2.60 �0.89� 4.85 4.57 �0.94�ATLAS 2.10 1.87 �0.89� 3.10 2.77 �0.90� 4.76 4.36 �0.92�

201 TP 1.86 1.59 �0.85� 2.96 2.39 �0.81� 4.37 3.55 �0.81�ATLAS 2.13 1.76 �0.82� 3.09 2.49 �0.81� 4.74 3.87 �0.82�

125 TP 1.90 1.38 �0.72� 2.96 2.11 �0.71� 4.81 3.20 �0.67�ATLAS 2.13 1.64 �0.77� 3.06 2.20 �0.72� 4.69 3.29 �0.70�

227 TP 2.18 1.94 �0.89� 3.20 2.78 �0.87� 4.74 4.16 �0.88�ATLAS 2.32 2.00 �0.86� 3.30 2.58 �0.78� 4.94 3.86 �0.78�

75 TP 2.18 2.01 �0.92� 3.33 3.02 �0.91� 5.70 5.14 �0.90�ATLAS 2.45 2.24 �0.91� 3.54 3.20 �0.90� 5.27 4.99 �0.95�

e calculated.

Fa

043108-11 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

IG. 8. Maps of wave energy flux at three levels: 10th, 50th, and 90th �all-month� percentiles. Inset numbers show valuest five example near-shore locations �circled�.

D

St1ttete

I

caGetvi

�otwWmtW

043108-12 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

. Along-coast integral of the wave energy flux

Integration of the flux propagating across the full length of the 25-m depth contour from 29°in the West to 148° E was carried out on all wave energy flux maps �Table IV�. Summer is the

ime of least wave energy, with the 10th, 50th, and 90th percentile integrated energy flux being 45,01, and 228 GW in January. In July, these rise to 97, 210, and 468 GW. These figures indicatehat the resource is fairly reliable: it is only for about 10% of the time �and here we should recallhat the NWWIII archive comprises estimates of wave properties at intervals of 6 h� that thenergy flux falls to less than half the median. Conversely, it is only for about 10% of the time thathe energy flux exceeds twice the median value. Averaged over the whole year, Australia’s South-rn coastline has a sustained wave energy resource of 146 GW �1329 TW h/yr�.

V. DISCUSSION

In 2005–2006, Australia generated around 254 TW h/yr of electricity from an estimated totalapacity of 50 GW.26 Presently, approximately 75% of electricity generation is from coal, withbout 55% from black coal. The wave energy resource along Australia’s Southern margin �146W� is approximately three times Australia’s total installed capacity. If 10% of the incident wave

nergy could be converted to electricity, approximately one-half of Australia’s present-day elec-ricity needs could be satisfied. Assessing the engineering feasibility, the cost or impact of con-erting this amount of wave energy to electricity is outside the scope of the present study. Our aims simply to assess the extent of the energy resource and to characterize its variability.

Our modeled wave energy estimates exceed buoy-derived estimates by a considerable margin13%–53%� when waves are small and by a significant margin �2%–29%� at other times. Thisverestimation of wave energy is consistent with the 5%–10% overestimation of Hs observed athe deep-water sites where a comparison between model values �effectively at the boundaries athich NWW3 input data are specified� and satellite altimeter data has been carried out �Table III�.e therefore attribute this error to the 5%–10% overestimation of Hs in the region by the NWW3odel. An alternative data set that could have been used to provide the offshore boundary condi-

ions for the nested domain SWAN models in this study is the one analyzed by Hughes and Heap.527

TABLE IV. Integrated energy flux on Australian coast for each of the representative sea-states for all months�GW�. Integrated along the 25-m contour for the stretch of coast from 29° S on the West Australian coast�approximate location of Geraldton, WA� to 148° E �Southern tip of Tasmania�, including that part of BassStrait, West of 148° E.

10th pc 50th pc 90th pc

Jan 45 101 228

Feb 47 110 248

Mar 50 120 276

Apr 60 136 352

May 64 168 408

Jun 75 178 443

Jul 97 210 468

Aug 96 211 466

Sep 90 211 486

Oct 68 160 374

Nov 46 105 247

Dec 44 107 236

Annual mean 59 146 372

Annual total �TW h/yr� 572 1329 3094

e did not use it because the discrepancy from buoy observations is much greater.

raa�swtlWi

wTN

Tt�dp�wwtbws

ayrt

043108-13 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

A problem common to many sources of renewable energy is the unpredictability of yield,aising concerns regarding their capacity to produce base-load electricity demand. Without stor-ge, wind and solar energy systems are susceptible to the rapid temporal variability of theirssociated natural processes. The weather systems responsible for this variability have scales of1000 km, so whole regions are impacted at once, not just individual farms. Five representative

ites along the Southern Australian margin are chosen as potential sites for wave energy extraction,hich are proximal to the existing Australian grid �Table V�. Four of these sites would connect to

he Australian National electricity Grid �NEM�, while the fifth would connect to the West Austra-ian grid only. We use time-series of EF computed directly from the NWW3 archives at each site.

e sum the energy across the four NEM sites, and across all five sites, and compute percentile,.e.,

E1P = �x=1�NEFx��P,

here x represents site 1 , . . . ,N �where N=4 or 5� and P indicates percentile �10, 50, and 90�.hese values are compared with the sums of percentiles presented in this paper �from deep-waterWW3 values�, i.e.,

E2P = x=1�N�EFP�x� .

he comparison provides a measure of the spatial variability of the incident wave energy availableo the national electricity grid. Summing over the NEM sites only during low wave conditions10th percentile case�, E110 is 8% larger than E210, indicating that during low wave conditions,istributing wave energy devices along the full extent of the NEM grid along the South coastrovides only a minimal increase in energy available for the grid. If summed over all five sitesi.e., if the WA and NEM grids were connected�, E110 is 26% larger than E210, indicating thathen low wave conditions are experienced along the SA, Vic, and Tas coasts, a larger portion ofave energy could be input to the grid from the Western Australia sites. These results are consis-

ent with those presented by Hemer et al.,16 which showed only a 4–5 h lag between wave eventseing observed off the South Australian coast and off the West coast of Tasmania, and indicateave energy devices connected to the NEM will likely be coincidentally low. Energy storage

ystems would be required in order to provide a consistent power supply from wave energy alone.The Australian Government has set a 20% target for renewable energy by 2020, requiring an

dditional 45 000 GW h. While an economic analysis of wave generation in Australian waters iset to be carried out, possibly with the development of this atlas, the massive availability ofesource, particularly when considered within the context of other renewable resources, suggests

TABLE V. Sum of available energy from sites along the Australian National Electricity Grid, including theWestern Australian Electricity Grid. N is the number of sites over which sum is performed. E1x and E2x aredefined in the text. The bracketed number is the percentage increase of E1x over E2x.

NLat

�° S�Lon�° E�

E110

�kW/m�E210

�kW/m�E150

�kW/m�E250

�kW/m�E190

�kW/m�E290

�kW/m�

4 35 133.75 64 �8%� 59 169 �2%� 165 471��2%� 478

38 140

39 142.5

42 145

5 35 116.25 103 �26%� 76 235 �10%� 212 572 ��8%� 622

hat the 20% target is certainly achievable.

V

5gae

ttpprottwg

wwwracswmreSSofvisi

bwtes

A

gmwRal

043108-14 M. Hemer and D. Griffin J. Renewable Sustainable Energy 2, 043108 �2010�

. CONCLUSIONS

Our near-shore wave energy resource maps describe a resource that is approximately 30%–0% less than the World Energy Council estimates for the Southern Australian margin but 50%reater than the estimates of Hughes and Heap.5 Our model-derived wave energy estimates arepproximately 20% greater than estimates from buoy data. We attribute this to a 5%–10% over-stimation of significant wave height in the region by the NWW3 model.

In the region of the Australian National Electricity Grid, the wave energy resource is essen-ially spatially uniform. When wave heights become low, they become so over the full stretch ofhe NEM, with only a short �12–24 h� delay across the whole Southern margin. Local storage orroduction offsetting rather than long distance transmission is therefore perhaps the more appro-riate way of managing the variability. Unlike some forms of renewable energy, however, theange of the energy yield fluctuations is not extreme. The yield is half the median value, only 10%f the time. It is also probably fair to say, although we have not proven this in this study, that theime-scale of variation and/or the unpredictable component of that is somewhat longer for waveshan it is for wind or solar. We attribute this to the simple fact that waves result from the action ofinds averaged over a very large distance away from the diurnally varying heating and topo-raphic effects of land.

The data set presented in this study aims to provide a suitable data set for informing policyith details of the available wave energy in Australia’s priority near-shore regions for harnessingave energy. The data will also further refine the most suitable locations for commercialization ofave energy converters, of interest to the wave energy industry. However, this user-group will

equire wave energy time-series data, providing, for example, details of threshold exceedancesbove and below which a device fails to operate �not available from this study� to enable aomplete assessment of the suitability of the wave energy resource for site-specific locations forpecific devices. A high spatial resolution wave transformation model as applied in this study, runith time-variable boundary and local wind conditions for a period of approximately 10 yr orore, should be regarded as the next step to adequately define the wave climate for wave energy

esource assessments. The required model should be applied with a single domain, so that mod-ling artifacts observed in the present study are removed. Unstructured grids are now supported byWAN �Ref. 28� and may prove suitable for future wave energy assessments for the extensiveouthern Australian domain investigated here. To support the required modeling effort, we rec-mmend greater investment in in situ observations along the Southern Australian margin in theorm of directional waverider buoys �or similar�. Presently, there is a dearth of directional obser-ations in this region. Hemer and Bye29 identified offshore wave direction as the single mostmportant parameter dictating attenuation of wave energy across the South Australian continentalhelf. Proper verification of wave models in this region would therefore ensure that wave directions suitably represented, which is not possible with the currently available observational data.

The Australian Government seeks to produce 45 000 GW h/yr of additional renewable energyefore 2020. The Southern margin of the Australian continent provides an abundant resource ofave energy, which can contribute to this target. Typical annual mean wave energy fluxes along

he Southern margin are approximately 50 kW/m �432 GW h/yr/km�. The total required renewablenergy quota could be achieved if 10% of the available wave energy resource over a 1000 kmection of the Southern Australian margin were converted to electricity.

CKNOWLEDGMENTS

This paper is a contribution from the Commonwealth Scientific and Industrial Research Or-anisation �CSIRO� Wealth for Oceans Flagship and the Centre for Australian Weather and Cli-ate Research �CAWCR�. The research was funded with assistance from the Australian Common-ealth Department of Environment, Water, Heritage and the Arts as part of the Australianenewable Energy Atlas. The authors acknowledge the supply of data from U.S. National Oceannd Atmosphere Administration, the Australian Bureau of Meteorology, and the Western Austra-

ian Government Department of Primary Industries and TUDelft for the use of the SWAN wave

mte

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

043108-15 Southern Australia’s wave energy J. Renewable Sustainable Energy 2, 043108 �2010�

odel. CAWCR is a partnership between the CSIRO and the Bureau of Meteorology. We alsohank Dr. Jim Gunson �CSIRO Marine and Atmospheric Research� and three anonymous review-rs for their comments on an earlier version of this manuscript.

1 Energy Supply Association of Australia, Australia’s Renewable Energy Target 2009, http://www.esaa.com.au/images/stories/FactSheets/2009ret.pdf �Accessed 24 March 2010�.

2 A. H. Clément, P. McCullen, A. Falcão, A. Fiorentino, F. Gardner, K. Hammarlund, G. Lemonis, T. Lewis, K. Nielsen,S. Petroncini, M.-T. Pontes, P. Schild, B. O. Sjöström, H. C. Sørensen, and T. Thorpe, Renewable Sustainable EnergyRev. 6, 405 �2002�.

3 D. Greaves, G. Smith, M. Attrill, M. Belmont, A. Chadwick, D. Conley, A. Eccleston, B. Godley, N. Harrington, C. L.Hor, P. Hosegood, L. Johanning, D. Millar, D. Reeve, J. Williams, J. Wolfram, J. Xu, A. Zobaa, and Q. Zou, Proceedingsof the ICE–Maritime Engineering 162, 187 �2009�.

4 World Energy Council, 2007 Survey of Energy Resources, World Energy Council 2007, London, United Kingdom, 600pp. Available at http://www.worldenergy.org/documents/ser2007_final_online_version_1.pdf �accessed 4 August 2009�.

5 M. G. Hughes and A. Heap, Renewable Energy 35, 1783 �2010�.6 G. Iglesias, M. Lopez, and R. Carballo, Renewable Energy 34, 2323 �2009�.7 H. L. Tolman, B. Balasubramaniyan, L. D. Burroughs, D. V. Chalikov, and Y. Y. Chao, Weather Forecast. 17, 311�2002�.

8 M. A. Hemer, J. A. Church, and J. R. Hunter, J. Coastal Res. 50, 433 �2007�.9 N. Booij, R. C. Ris, and L. H. Holthuijsen, J. Geophys. Res. 104, 7649 �1999�.0 W. E. Rogers, J. M. Kaihatu, L. Hsu, R. E. Jensen, J. D. Dykes, and K. T. Holland, Coastal Eng. 54, 1 �2007�.1 D. G. Provis and R. K. Steedman, Australasian Conference on Coastal and Ocean Engineering, IEAust., Christchurch,New Zealand, 1985.

2 M. A. Hemer, G. Heinson, J. A. T. Bye, and A. White, The Wind-Driven Air-Sea Interface. Electromagnetic and AcousticSensing, Wave Dynamics and Turbulent Fluxes, edited by M. L. Banner �University of NSW, Sydney, 1999�, pp. 49–58.

3 A. D. Short and N. L. Treneman, Aust. J. Mar. Freshwater Res. 43, 765 �1992�.4 W. J. Pierson and L. Moskowitz, J. Geophys. Res. 69, 5181 �1964�.5 L. H. Holthuijsen, Waves in Oceanic and Coastal Waters �Cambridge University Press, Cambridge, UK, 2007�.6 M. A. Hemer and I. Simmonds, Cont. Shelf Res. 28, 634 �2008�.7 P. Petkovic and C. Buchanan, Australian Bathymetry and Topography Grid (Digital Dataset) �Geoscience Australia,Canberra, 2002�.

8 K. Nielsen, “Development of recommended practices for testing and evaluating ocean energy systems,” in InternationalEnergy Agency Technical Report, Annex II, 2003 �http://www.iea-oceans.org/_fich/6/IEA-OES_Annex_II_2003_�2406�.pdf �accessed 18 June 2010�.

9 A. V. Boukhanovsky and C. Guedes Soares, Appl. Ocean. Res. 31, 132 �2009�.0 S. Pond and G. L. Pickard, Introductory Dynamical Oceanography, 2nd Ed. �Butterworth-Heinemann, Oxford, 1983�.1 M. A. Hemer, J. Coastal Res. SI56, 228 �2009�.2 J. I. Collins, J. Geophys. Res. 77, 2693 �1972�.3 M. A. Hemer, J. A. Church, and J. R. Hunter, Int. J. Climatol. 30, 475 �2010�.4 P. G. Challenor and P. D. Cotton, “The joint calibration of altimeter and in-situ wave heights,” Technical Report No. 12,World Meteorological Organisation Document Number WMO/TD-No.1081, JCOMM, 2002.

5 See supplementary material at http://dx.doi.org/10.1063/1.3464753 for figures of each of the 156 maps of the presentedwave energy resource atlas.

6 ABARE, Energy in Australia 2008, in Australian Government Department of Resources, Energy and Tourism �ABARE,Canberra, 2008�, http://www.abareconomics.com/publications_html/energy/energy_08/energyAUS08.

7 M. A. Hemer, K. McInnes, J. O’Grady, J. A. Church, and J. R. Hunter, “Inter-annual variability and trends in the offshoreAustralian wave climate,” in Final Report for Department of Climate Change Surface Ocean Wave Variability Project,CSIRO Australia, 2008, 119 pp., http://www.climatechange.gov.au/~/media/publications/coastline/wave-climate.ashx

8 M. Zijlema, Coastal Engineering 57, 267 �2010�.9 M. A. Hemer and J. A. T. Bye, Trans. Royal. Soc. Sth. Aust. 123, 107 �1999�.