Professional Documents
Culture Documents
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
Department of Ocean and Resources Engineering, University of Hawaii at Manoa, Honolulu, HI 96822, USA
Department of Meteorology, University of Hawaii at Manoa, Honolulu, HI 96822, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 3 November 2009
Accepted 20 July 2010
Available online 24 August 2010
Hawaii is subject to direct approach of swells from distant storms as well as seas generated by trade winds
passing through the islands. The archipelago creates a localized weather system that modies the wave
energy resources from the far eld. We implement a nested computational grid along the major Hawaiian
Islands in the global WaveWatch3 (WW3) model and utilize the Weather Research and Forecast (WRF)
model to provide high-resolution mesoscale wind forcing over the Hawaii region. Two hindcast case
studies representative of the year-round conditions provide a quantitative assessment of the regional wind
and wave patterns as well as the wave energy resources along the Hawaiian Island chain. These events of
approximately two weeks each have a range of wind speeds, ground swells, and wind waves for validation
of the model system with satellite and buoy measurements. The results demonstrate the wave energy
potential in Hawaii waters. While the episodic swell events have enormous power reaching 60 kW/m, the
wind waves, augmented by the local weather, provide a consistent energy resource of 15e25 kW/m
throughout the year.
2010 Elsevier Ltd. All rights reserved.
Keywords:
Mesoscale model
Spectral wave model
Waves in Hawaii
Wave energy
Wave power
Ocean observing system
1. Introduction
Hawaiis mid-Pacic location and year-round access to the ocean
make it a center for marine research and recreational activities.
Hidden in these activities are the wave energy resources that have
only recently attracted attention. Fig. 1 illustrates the wave climate
around the Hawaiian Islands. Extratropical storms near the Kuril
and Aleutian Islands generate northwest swells reaching 5 m
signicant wave height in Hawaii waters during the winter months
of NovembereMarch. The south facing shores experience more
gentle swell conditions associated with extratropical storms off
Antarctica during the summer months from May to October. In
addition, consistent trade winds generate wind waves from the
northeast to east throughout the year. Existing buoys operated by
the National Data Buoy Center (NDBC) provide point measurements
of the wave conditions either off the Hawaiian Island chain or at
nearshore locations. A numerical modeling approach can augment
the available measurements to provide spatial distributions of the
ocean wave resources for assessment [1e3].
The National Centers for Environmental Prediction (NCEP) operates WaveWatch3 (WW3) to provide 7.5 days of global wave forecasts
* Corresponding author.
E-mail addresses: stopa@hawaii.edu (J.E. Stopa), cheung@hawaii.edu (K.F.
Cheung), yileng@hawaii.edu (Y.-L. Chen).
0960-1481/$ e see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2010.07.014
555
Fig. 1. Wave climate and buoy resources around the Hawaiian Islands.
556
(w5.5 km) resolution. The model utilizes the ice concentration from
the FNL dataset and wind forcing from a combined FNL and Hawaii
WRF dataset. Tolman and Chalikov [23] concluded that the parameterization of the source term in WAM [24] underestimates the
transfer of energy from winds to waves for small fetches. The Hawaii
domain, which has intricate wind patterns caused by steep topographic gradient and narrow channels, would benet from the
source term of Tolman and Chalikov [23].
The Hawaii WW3 denes the boundary conditions for the Oahu
SWAN, which covers a domain of 201.65e202.40 E and
21.20 e21.75 N at 9-sec (approximately 280 m) resolution to
provide data at the nearshore wave buoys for model validation.
Wind forcing is not considered due to the small geographic size of
the computational domain. The SWAN model is similar to WW3 in
that it solves the action balance equation with parameterization of
nonlinear processes [6]. However, SWAN is better suited for
shallow water processes by including additional source terms for
triad waveewave interactions and depth-induced wave breaking as
well as the JONSWAP parameterization for dissipation due to
bottom friction [25]. In addition, SWAN accounts for some effects of
diffraction by including an additional term derived from the mildslope equation [26].
Both WW3 and SWAN provide directional wave spectra at the
grid points that can reveal multiple wave events occurring simultaneously. This is especially important for Hawaii as the ocean
always contains a mix of swell and wind wave events. Since wave
energy devices have narrow operating frequency ranges, it is
important to identify the energy levels of the wave components
separately. The latest release of WW3 (v3.14) implements the
Hanson and Phillips [27] algorithm to determine the wave
parameters for each spectral partition. The wave power per unit
crest length is
1
2
rgHrms
Cg
8
557
Fig. 3. Comparison of wind elds from Hawaii WRF and QuikSCAT for Case Study 1.
558
Fig. 4. Comparison of wave elds from Hawaii WW3 and altimetry for Case Study 1.
559
estimated from the slope of the leading edge of the returned signal
[32]. Typical errors are within 0.5 m or 10% of Hs, whichever is
larger. Caires and Sterl [33] have shown that wave heights derived
from buoys and altimeters have comparable instrumental error
variances. The signicant wave height derived from the Ku band is
considered in this study because the C band is more susceptible to
contamination by water vapor in the atmosphere. Similar to
QuikSCAT, Jason-1 and T/P are polar orbiting satellites. However,
these satellites orbit much slower and y over the same ground
track every 10 days. Along-track, gridded Hs values with resolution
of approximately 2.5 min (5.8 km) provide comparison with results
from the Hawaii WW3.
3. Waves in Hawaii
Fig. 5. Scatter plot of computed signicant wave height for Case Study 1.
Fig. 6. Comparison of signicant wave height from Hawaii WW3 and buoys for Case Study 1.
560
Fig. 7. Wave power distributions for swell and wind waves in Case Study 1.
561
Fig. 8. Comparison of wind elds from Hawaii WRF and QuikSCAT for Case Study 2.
of the swell ranges from the northwest to north. The swell in the
rst three days is the largest and two weaker swells occur during
the remainder of the case study. The winds are relatively calm in
the beginning and transition to moderate northeast trade conditions that generate waves within the Hawaii domain. This case
study allows testing of the numerical models and examination of
the wave conditions for a number of different episodes: weak
winds coupled with a swell from the North Pacic, moderate winds
coupled with a smaller swell, and locally generated wind waves
with a background swell.
The wind event can be broken into two distinct episodes with
three representative patterns as shown in Fig. 3. The data is plotted
in the original resolution along with every other QuikSCAT wind
vector and every sixth WRF wind vector. The QuikSCAT data is re-
562
Fig. 9. Comparison of wave elds from WW3 Hawaii and altimetry for Case Study 2.
Fig. 10. Scatter plot of computed signicant wave height for Case Study 2.
563
case study peaking on March 19e20. The model data matches the
average measurements very well, but does not capture the uctuations. As the winds in the Hawaii region are relatively calm, the
discrepancy is probably due to the low-resolution FNL wind forcing.
The computed swell from March 23 to 25 is consistently larger than
the buoy measurements most likely due to the generation source
term in WW3. The computed swell arrives slightly before the
record at buoy 51001 on March 23. The wave model predicts the
swell to have a longer wave period, which equates to a more
energetic sea state and faster propagation speed. This error propagates through Hawaii WW3 resulting in over-prediction of Hs at
51002, 51003, and Waimea. The Mokapu buoy, which is in the
shadow of Oahus northeast headland, does not register this swell.
The winds transition from calm to moderate and local waves begin
to develop on March 26. A common feature at buoys 51002, 51003,
Mokapu, and Waimea is the slight under-prediction of wave
heights during March 27e29. A possible explanation lies in the
ability of wind wave models to account for the abrupt change in
wind speed. The match is much better towards the end of the
period when the wind waves gradually develop. The high-resolution Hawaii WRF forcing in the Hawaii WW3 reproduces the small
uctuations at buoys 51002, 51003, Mokapu, and Waimea at the
end of the episode. The computed wave height at buoy 51001,
which is located outside the Hawaii WW3, is smooth throughout
this last episode due to the low-resolution FNL forcing.
With the model validated, we quantify the wave power around
the Hawaiian Islands and plot the results in Fig. 7 at the same time
steps as those in Fig. 3. The left column of panels represents the
primary swell and the right column represents the wind waves
determined from the algorithm of Hanson and Phillips [27]. The left
panel on March 20 corresponds to the peak of the northwest swell.
The wave power, which reaches 60 kW/m on the north facing shores
of the Hawaiian Islands, is on the same order of magnitude as
average conditions found in northwest Spain [1]. There are minimal
wind wave activities consistent with the weak local wind conditions
in the beginning of the case study. On March 25, the smaller
northwest swell has wave power in the range of 25e35 kW/m, and
local waves begin to develop with the trade winds from the east. The
northwest swell reduces to an estimated wave power of 15 kW/m or
less on April 1. The wind waves, on the other hand, become dominant and have 15e35 kW/m of wave power available. Because of the
return ow and low wind speed, the algorithm of Hanson and
Phillips [27] classies the wind waves in the wake of the trade winds
behind Hawaii Island as swell. Other than that the estimated power
is relatively large when compared to windewave regimes around
the world. Since the northeast and east trade winds are common
throughout the year, their speed up on the south side of Hawaii
Island provides a reliable resource for wave energy development.
3.2. Summer season
Hawaii experiences gentler south swells in the summer, when
the subtropical high-pressure generates persistent trade winds
from the east and northeast. In South Pacics winter, storms in the
roaring 40s generate most of the swells toward the Americas.
Hawaii receives swells from these sources through directional
spreading of the waves. The second case study from September 4 to
18, 2005 occurs in end of the summer season, when the systems
form closer to Australia and aim more wave energy toward Hawaii.
There are multiple swells throughout the case with the largest
around September 16e17 with signicant wave heights over 2 m
and peak periods over 18 s. Typical summer winds with speeds in
the 10e15 m/s range from the east persist throughout the entire
case study. Fig. 8 shows three representative patterns of the trade
wind ow across the Hawaiian Island chain. Alternating speed-up
564
Fig. 11. Comparison of signicant wave heights from Hawaii WW3 and buoys for Case Study 2.
and wake formations are evident leeward of the island chain. The
Hawaii WRF wind elds match the QuikSCAT mean ow very well
especially over the open ocean and in the exit region of the Alenuihaha Channel. QuikSCAT does not capture the island-induced
airow in coastal waters and the wake circulations off the leeward
coasts due to its spatial resolution. The rather constant trade winds
give rise to predominant wind waves throughout the period. This
case study allows testing of the numerical models and assessment
of the wave pattern in a typical summer event with moderate wind
waves and a range of south swell conditions.
Fig. 9 displays the representative wave patterns around the time
of the wind data snapshots for direct comparison. The rst and
second rows show the wind waves from the east and the last row
shows the south swell coupled with the wind waves. On September
6, the trade winds generate signicant wave heights of up to 3.5 m
in the exit region of Alenuihaha Channel. The Hawaii WW3
reproduces the shadow west of Hawaii Island, but slightly underestimates the wind waves to the south. On September 11, the wind
wave pattern continues. The computed wind waves compare well
with the altimetry data, with errors less than 0.5 m. On September
16, the wave eld displays the peak of the south swell coupled with
the moderate wind waves from the east. The energy from the south
swell lls the shadows of the wind waves on the leeward side of the
islands and vice versa. This results in a discontinuous pattern of the
peak wave direction in the domain. The regional model is able to
reproduce the wave heights relatively well in comparison to the
altimetry data. There are discrepancies in the southern part of the
domain where the wind waves and south swell combine to give
larger wave heights than the altimetry data. There were 10 satellite
passes from both the T/P and Jason-1 platforms during the entire
case study. Fig. 10 shows the overall comparison utilizing all the
altimetry data. The over prediction to the south of Hawaii Island is
shown as the group of points in the top of the plot. The overall
prediction is good with an error of typically within 0.5 m.
Fig. 11 shows the time series comparison of the computed and
measured signicant wave heights at the buoys. The overall agreement is reasonably good. There are instances of underestimation of
the wind wave heights at buoys 51002, 51003, and Mokapu due to
565
Fig. 12. Wave power distributions for swell and wind waves in Case Study 2.
the weaker modeled winds. The Hawaii WRF does not fully account
for all of the trade wind processes because the subtropical highpressure system is located outside the regional domain. These
errors, however, are less than 0.5 m in signicant wave height. The
largest discrepancies occur at buoys 51002 and 51003 at the peak of
the south swell during September 15e17. The wave model is predicting larger wave heights to the south of the Hawaiian Island chain
566
567