Ocean Dynamics (2013) 63:961–977

DOI 10.1007/s10236-013-0640-2

Wave climate simulation for southern region

of the South China Sea
Ali Mirzaei & Fredolin Tangang & Liew Juneng &
Muzneena Ahmad Mustapha & Mohd Lokman Husain &
Mohd Fadzil Akhir

Received: 6 February 2013 / Accepted: 19 June 2013 / Published online: 9 July 2013
# Springer-Verlag Berlin Heidelberg 2013

Abstract This study investigates long-term variability and covering the Sunda shelf region. However, the trend appears
wave characteristic trends in the southern region of the South to be largely influenced by large El Niño signals.
China Sea (SCS). We implemented the state-of-the art
WAVEWATCH III spectral wave model to simulate a 31- Keywords South China Sea . WAVEWATCH III™ .
year wave hindcast. The simulation results were used to Significant wave height . Niño3.4 . El Niño Modoki
assess the inter-annual variability and long-term changes in
the SCS wave climate for the period 1979 to 2009. The
model was forced with Climate Forecast System Reanalysis 1 Introduction
winds and validated against altimeter data and limited available
measurements from an Acoustic Wave and Current recorder Climate change is recognised as a major environmental issue
located offshore of Terengganu, Malaysia. The mean annual influencing natural and human systems. Coastal areas are
significant wave height and peak wave period indicate the most affected by a range of hazards connected with climate
occurrence of higher wave heights and wave periods in the change, such as increasing inundation due to storms and sea
central SCS and lower in the Sunda shelf region. Consistent level rise resulting in coastal erosion. Moreover, increased
with wind patterns, the wave direction also shows southeast- storminess can change the characteristics of waves. Surface
erly (northwesterly) waves during the summer (winter) mon- ocean waves are formed when winds blow across the water
soon. This detailed hindcast demonstrates strong inter-annual surface transferring their energy to the water. As surface
variability of wave heights, especially during the winter winds over the ocean might be changing both in strength
months in the SCS. Significant wave height correlated nega- and patterns due to climate change, ocean waves could be
tively with Niño3.4 index during winter, spring and autumn affected by anthropogenic forcing of the climate system
seasons but became positive in the summer monsoon. Such (Wang et al. 2004). Ocean wave and storm surge character-
correlations correspond well with surface wind anomalies over istics of the future climate are also expected to be different
the SCS during El Nino events. During El Niño Modoki, the from that of the present (Mori et al. 2010). In assessing the
summer time positive correlation extends northeastwards to impact of changes in waves on coastal and offshore environ-
cover the entire domain. Although significant positive trends ments, it is crucial to understand the spatial and temporal
were found at 95 % confidence levels during May, July and variability of the wave climate. Long-term changes in wave
September, there is significant negative trend in December characteristics can modify the profile of the coast and the
shape of sandy beaches through changes in nearshore circu-
Responsible Editor: Roger Proctor
lation and sediment transport characteristics (Wright et al.
1991; Bosserelle et al. 2012). Understanding the present
A. Mirzaei : F. Tangang (*) : L. Juneng : M. A. Mustapha
wave climate is important in the context of assessing the
Research Centre for Tropical Climate Change System (IKLIM),
Faculty of Science and Technology, Universiti Kebangsaan long-term and future changes of waves associated with an-
Malaysia, 43000 Bangi, Selangor, Malaysia thropogenic forcing. Recent studies on several climate
e-mail: [email protected] change scenarios suggested shifts of storm tracks and an
M. L. Husain : M. F. Akhir
increase of storm frequency and hence, changes in the wave
Institute of Oceanography and Environment, Universiti Malaysia characteristics (IPCC, 2007). Moreover, investigating the
Terengganu, Kuala Terengganu, Malaysia present wave climate would provide valuable information
962 Ocean Dynamics (2013) 63:961–977

on the possible impact of waves to coastal zones, such as other investigations, numerical wave models have been used
coastal flooding, inundation and erosion. to predict the future wave climate under various global
A combination of data analysis, including statistical and warming emission scenarios (Wang et al. 2004; Wang and
numerical modelling has been employed to assess the present Swail 2006; Kriezi and Broman 2008; Mori et al. 2010;
wave climate. In this study, a numerical wave model was used Brown et al. 2012; Zacharioudaki et al. 2011).
to understand the long-term changes and spatial variability of The SCS is a semi-enclosed marginal sea located in the
waves in the southern region of the South China Sea (SCS). A western Pacific and covers an area from the Malacca Straits
number of studies have employed numerical wave models to including Singapore, to the Straits of Taiwan (Fig. 1). Mon-
simulate ocean surface wave characteristics in coastal areas soon winds and synoptic systems, such as fronts and tropical
with emphasis on long-term extreme wave climate impacts cyclones dominate the SCS weather and climate (Chu et al.
(Lionello and Sanna 2005; Music and Nickovic 2008; 1999, 2000). The SCS is an important natural resource for the
Panchang et al. 2008; Cuchiara et al. 2009; Hemer et al. surrounding nations in terms of: oil and gas production, tour-
2010). Swail et al. (1999) used a global spectral ocean wave ism and recreation, commerce, navigation and fisheries
model for two long-term (40 years) wave hindcasts to analyse (Salahuddin and Curtis 2011). However, the information on
the trends and variability of changes in waves. Regionally, potential long-term changes of wave properties in the SCS is
Bosserelle et al. (2009) implemented a third generation spec- fragmented and incomplete. It appears that long-term wave
tral wave model (WAVEWATCH III) with a 40-year simula- characteristics in the SCS have not been recorded systemati-
tion to quantify the inter-annual variability and longer-term cally. The number of studies on waves and their characteristics
changes in the wave climate around Western Australia. In are limited and only cover a short period of time (Qi et al.

Fig. 1 Bathymetry and 45˚

definition for three mosaic nested M1 = 0.31° × 0.31°
computational domains

40˚

35˚ (m)
8000

30˚
6000
M2 = 0.25° × 0.25°

25˚ 4000

2000
20˚

M3 = 0.15° × 0.15° 0
15˚
−2000

10˚
−4000

5˚
−6000

0˚
−8000

−5˚

−10˚
90˚ 95˚ 100˚ 105˚ 110˚ 115˚ 120˚ 125˚ 130˚ 135˚ 140˚ 145˚
Ocean Dynamics (2013) 63:961–977 963

2010; Chu et al. 2004; Chunxia et al. 2008; Jiang et al. 2010). wind products available, including those of the National Centres
As climate change is becoming more imminent, there is a need for Environmental Prediction (NCEP) and the European Centre
to assess the future wave climate in the SCS under various for Medium-Range Weather Forecasts. Climate Forecast System
greenhouse gases emission scenarios and thus, the basis for Reanalysis (CFSR) is an NCEP product based on a high-
such as assessment is the understanding of the present wave resolution global coupled atmosphere–ocean–land surface–sea
climate. This research seeks to quantify the changes in wave ice system with a resolution of ∼38 km (T382) and temporal
climate and variability in the southern extent of the SCS over coverage of 31 years from 1979 to 2009 (Saha et al. 2010). The
the past 31 years. CFSR data are a blended product of conventional and satellite
The rest of this paper is organised as follows: The data set observations and available at an hourly time resolution in trop-
and methodology used in this study and the model setup and ical regions. A comparison with the QuickSCAT climatology for
validation are all described in Section 2. The results, followed a period between September 1999 to October 2009 indicated
by a detailed discussion, are given in Section 3, and Section 4 smaller errors for the CFSR compared with the NCEP/NCAR
concludes the paper. reanalysis (R1) and NCEP/DOE reanalysis (R2) (Xue et al.
2010).

2 Data and method 2.2 Wave model setup

2.1 Atmospheric data This study employs WAVEWATCH III™ III™ (WW3),
which is an operational third generation wave model from
Reliable and accurate global high-resolution (both temporally NCEP/NOAA (Tolman 2002, 2009). WW3 has been widely
and spatially) surface wind data are crucial for driving ocean validated against global network buoys and altimeter data
wave models. The availability of such data is limited and (Tolman 2002; Bidlot et al. 2007; Hanson et al. 2009;
restricted to only small regions that have comprehensive marine Chawla et al. 2009). The main physics of WW3 is mostly
observation networks. However, there are several reanalysis based on the WAM model (WAMDI group 1988) but

Laos
es
pin

16˚ Thailand South China Sea 16˚

ilip
Ph

Vietnam

b
Combodia

012
12˚ 12˚
038

190
114
d
an
Isl
an
law

Gulf of Thailand
Pa

c
001

077

153

229

051

127

203

8˚ Sulu Sea 8˚

AWAC a
Terengganu Sabah
064

140

216

Natuna Island Brunei

Peninsular
M
4˚ al
ac
Malaysia
Siantan Island
4˚
c a Celebes Sea
St
ra
Jemaja Island
it Tioman Island Sarawak

Singapore

0˚ Sumatra
Karimata Strait
Indonesia 0˚

100˚ 104˚ 108˚ 112˚ 116˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
Fig. 2 The area of focus of this study (left) and the coverage of Topex satellite tracks in the SCS (right)
964 Ocean Dynamics (2013) 63:961–977

6 Pass 038 6 Pass 012

5 5

4 4
Model

Model
3 3

2 2

C.D. = 0.85 C.D. = 0.86

1 Bias = 0.09 1 Bias = −0.05
RMSE = 0.36 RMSE = 0.32

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 Pass 051 6 Pass 064

5 5

4 4

Model
Model

3 3

2 2

C.D. = 0.89 C.D. = 0.81

1 Bias = 0.12
1 Bias = −0.09
RMSE = 0.35 RMSE = 0.28

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 6 Pass 127
Pass 114

5 5

4 4
Model
Model

3 3

2 2

C.D. = 0.89 C.D. = 0.72

1 1 Bias = −0.10
Bias = 0.16
RMSE = 0.38 RMSE = 0.44

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 6 Pass 166
Pass 077

5 5

4 4
Model
Model

3 3

2 2

C.D. = 0.87 C.D. = 0.59

1 Bias = −0.04 1 Bias = −0.12
RMSE = 0.30 RMSE = 0.35

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter
Fig. 3 Scatter diagram of Altimeter (along the tracks) and predicted significant wave height (model) in metres
Ocean Dynamics (2013) 63:961–977 965

6 Pass 190
6 Pass 242

5 5

4 4

Model
Model

3 3

2 2

C.D. = 0.93 C.D. = 0.73

1 Bias = 0.03
1 Bias = −0.07
RMSE = 0.35 RMSE = 0.27

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 Pass 216 6 Pass 229

5 5

4 4

Model
Model

3 3

2 2

C.D. = 0.88 C.D. = 0.90

1 1 Bias = 0.02
Bias = 0.03
RMSE = 0.37 RMSE = 0.36

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 Pass 179 6 Pass 001

5 5

4 4
Model

Model

3 3

2 2

C.D. = 0.65 C.D. = 0.66

1 Bias = −0.12 1 Bias = −0.14
RMSE = 0.27 RMSE = 0.37

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter

6 Pass 203 6 Pass 153

5 5

4 4
Model

Model

3 3

2 2

C.D. = 0.81 C.D. = 0.87

1 Bias = −0.10 1 Bias = 0.08
RMSE = 0.31 RMSE = 0.40

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Altimeter Altimeter
Fig. 4 Same as Fig. 3 except for different tracks
966 Ocean Dynamics (2013) 63:961–977

subsequent improvement on its physics for shallow waters altimeter data are available in high-resolution and are averaged
increases the model’s ability to simulate shallow water pro- every five records along each track yearly and compared with
cesses, such as refraction and shoaling. In areas where the simulated data that are interpolated along their appropriate tracks.
swell is generated nearby, WW3 captures the wave height However, the altimeter data have limitations, especially with data
variability better than WAM (Padilla-Hernadez et al. 2004; sparseness, because the satellite only returns to the same point
NCEP 2006). WW3 also uses sub-grids to better resolve after 10 days or more and only the sea surface directly below the
islands in the simulation for improved estimation of refrac- satellite is measured.
tion and wave energy reduction due to depth-induced break- Chu et al. (2004) validated the WW3 model against
ing and obstructions blocking (Tolman 2009). TOPEX/Poseidon wave measurements in the SCS and they
A multi-grid WW3, which is a mosaic of grids with two- found good agreement between the simulated and observed
way exchange of information, was implemented to simulate data. Figures 3 and 4 represent scatter plots of simulated
the wave characteristics in the southern SCS (Fig. 1) from versus altimeter significant wave height (SWH) for all se-
1979 to 2009. The outer computational domain M1, covers lected tracks. The computed statistics for the entire year of
the entire SCS and parts of the adjacent seas from 90° to 145° 2009, including the coefficient of determination (C.D.), bias
E and 10° S to 45° N with 0.3×0.3° spatial grid resolution. and root mean square error (RMSE) are displayed. General-
Two nested domains of M2 and M3 with computational grids ly, based on the scatter plots, the model performed reason-
of medium (0.25×0.25°) and fine (0.15×0.15°) resolutions, ably well in simulating the SWH. For most tracks, the sim-
respectively, are embedded within M1 to assess wave pa- ulated SWH matches the observed with C.D. ranges between
rameters approaching the southern extent of the SCS coast-
lines. The use of high spatial resolution (e.g. M3 at 0.15°
resolution) is expected to resolve well the underlying ba-
3
thymetry for the M3 domain, which is large enough to also −− Model C.D. = 0.96
−− AWAC
properly account for the swell components. Bias = 0.13
RMSE = 0.22
To better resolve the islands and shoals in the computation,
the three computational grids are embedded with an obstacle
Significant wave height (m)

grid that is generated as described by Chawla and Tolman 2

(2007). The bathymetric data was obtained from the ETOPO2

(NGDC 2006) at 2-min resolution (≈0.033°) (Fig. 1). The
WW3 model automatically interpolates the forcing data in both
space and time using a quadratic scheme proportional to every 1
grid resolution. As the SCS is located in a tropical area, no ice
coverage is used in the computation.
Different configurations of the model were examined to
produce wave parameters. For input and dissipation, the 0
Tolman–Chalikov (TC) parameterisation was used because 0 200 400 600 800 1000
Two hourly records
record
it performed better than WAM4 in simulating wave param- 8
eters for the SCS (Tolman 2009). The TC approach also −− Model C.D. = 0.98
−− AWAC Bias = 0.28
reduces computation time. Additionally, all domains were RMSE = 0.75

set based on shallow water physics with default values.

6
Mean wave period (s)

2.3 Model validation

4
The results of the WW3 model were validated with satellite
altimeter data and limited 3-month Acoustic Wave and Current
(AWAC) directional wave data recorded nearshore at Terengga-
nu (102.92° E and 5.5° N) (Fig. 2a). The AWAC wave measure- 2

ments are available in two-hourly intervals covering a period

from January–April 2009 and were provided by the Institute
of Oceanography and Environment, Universiti Malaysia Tereng- 0
0 200 400 600 800 1000
ganu. The TOPEX/Poseidon (1992 to 2005) and JASON (2005 Two hourly records
record
to present) data for the SCS were obtained from the French Fig. 5 Comparison between modelled and measured (AWAC) signif-
Research Institute for Exploration of the Sea (Ifremer). Figure 2b icant wave height and mean wave period during the period January–
shows the 16 satellite tracks considered in this study. The March 2009
Ocean Dynamics (2013) 63:961–977 967

0.8 and 0.9 and averaged RMSEs of 0.35 m. For these tracks, value of 0.96 (0.22 m) is generally larger (smaller) compared
the range of bias is between −0.1 and 0.1 m. The model with the value based on the satellite tracks.
underestimated the SWH in the Malacca Straits and Gulf of Due to the nearshore location of the AWAC, the insufficient
Thailand (Passes 001 and 166) with smaller C.D.s of 0.66 resolution of the model resulted in an underestimation of the
and 0.59, respectively. In the case of Malacca Straits, insuf- wave height. On the other hand, short-lived local winds could
ficient model resolution reduces its performance. In addition, also be a reason why the model did not accurately reproduce
higher resolution of wind forcing is also required to capture short period waves in some regions (Bosserelle et al. 2009). In
wave characteristics in such a narrow water channel. More- addition, the complexity of topography and existence of
over, the wave characteristics can be influenced by strong islands could cause refraction and blocking of waves, which
tidal currents, especially along Peninsular Malaysia, the Gulf leads to an underestimation of wave heights and wave periods
of Thailand and the west coast of Borneo (Fang et al. 1999); nearshore and in coastal areas.
however, tidal effects were not included in the simulation.
In addition to satellite altimetry data, the model was also 2.4 Analysis methods
validated against a 3-month record of nearshore wave measure-
ments at Terengganu (Fig. 5). The measurements were carried The WW3 model has been set to output several parameters at
out using a Nortek acoustic Doppler profiler and directional 3-hourly intervals for the 31-year hindcast. These outputs
wave gauge (AWAC) for a period of January–April 2009. comprised three main wave parameters: the significant wave
Figure 4 shows the time series comparison of SWH and mean height, peak wave period and mean wave direction. These
wave period between the model and AWAC data. The compar- parameters were used to investigate the inter-annual variabil-
ison shows the reasonable ability of the model to reproduce the ity and long-term changes of the wave climate in the south-
wave parameters in the location with C.D. values of 0.96 and ern SCS. Seasonal mean significant wave heights were de-
0.98 for Hs and TM, respectively. For Hs, the C.D. (RMSE) termined by combining: the December–January–February

Fig. 6 Mean significant wave

height for the period 1979–2009
1.8
model hindcast
16˚
1.2

(m)
2

1.2
1.8
1.5

1.5

12˚
1.8

1.5

2
0.9 1.

8˚ 1
1.
5
0.9

1.2
1. 2

4˚ 0.9 0.9

0
0˚

100˚ 104˚ 108˚ 112˚ 116˚ 120˚

968 Ocean Dynamics (2013) 63:961–977

Fig. 7 Mean 90th percentile

wave height for the period 1979–
2009 model hindcast 6
16˚ 6
(m)
7

6
12˚

4
4
8˚ 4
4

5
3

4˚
2

5
4

4
1
0˚

100˚ 104˚ 108˚ 112˚ 116˚ 120˚

Fig. 8 Seasonal mean wave DJF MAM

height for the period 1979–2009
model hindcast 16˚ 16˚

1
2.5

12˚ 12˚

8˚ 8˚
1.5

1
5
2.
1.5

4˚ 2 4˚
1

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JJA SON
2
16˚ 16˚
1.5

12˚ 12˚
1.5

1 1.5

8˚ 8˚

1
4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
m

0 1 2 3
Ocean Dynamics (2013) 63:961–977 969

(DJF) monthly mean for winter, March–April–May (MAM) considered statistically significant. The relationship between
for spring, June–July–August (JJA) for summer and Septem- wave height variability and El Nino-Southern Oscillation
ber–October–November (SON) for autumn. Simulated pa- (ENSO) was also investigated by correlating ENSO indices
rameters were averaged annually and identified as the mean (Nino 3.4 and El Nino Modoki) with Hs.
annual wave height (Hs), peak wave period (Tp) and mean
wave direction (MWD). Moreover, the mean annual of the
99th (H99) and 90th (H90) percentiles of wave height, which 3 Results and discussion
are known as the largest 1 and 10 % of the parameters,
respectively, were also calculated. To compute the annual In terms of the long-term mean Hs, the maximum values
mean wave direction, the mean wave direction was first split feature prominently in the central region of the SCS (Fig. 6)
into meridional and zonal components and averaged sepa- where winds can increase from 5 knots to over 40 knots in a
rately and then recombined. Inter-annual variability of Hs matter of hours during the monsoon seasons. Hs decreases
was determined by calculating the standard deviation of the gradually in the northeast–southwest direction towards the
monthly mean values of Hs. Sunda shelf where the values are relatively lower. In the
Trends in the time series of Hs, Tp, MWD and H90 were southern region of the SCS and in the Gulf of Thailand, Hs
computed for each grid point of the M3 domain. The trend ranges from 0.5 to 1 m. The lower values of Hs in the Gulf of
analysis of the year-grouped wave height climate was made by Thailand might be due to shadowing effects of Indo-China.
the linear least square fitting of time series of Hs and H90, In the coastal region along Peninsular Malaysia and the west
while the p values were computed to determine the signifi- coast of Borneo, Hs values were lower than 0.5 m. The
cance of the trend. Additionally, trends in monthly wave height reduction of Hs in these regions could be due to complex
were computed to identify the significance of changes during bathymetry and insufficient model resolution. The bathym-
every month. Trends at 95 % confidence (p value<0.05) were etry in the SCS (Fig. 1) plays an important role because large

Fig. 9 Mean peak wave period

7

Tp, for the period 1979–2009

model hindcast
16˚
(s)

7
8

7
7
12˚
6

8˚ 5 4
6

4˚
6 2
4

1
5
6

4 5
0
0˚ 5
4

100˚ 104˚ 108˚ 112˚ 116˚ 120˚

970 Ocean Dynamics (2013) 63:961–977

amounts of wave energy, generated in the central parts of the SCS obstructs wave propagations resulting in a reduction of
SCS, is reduced due to apparent differences of topography Hs and H90. In addition, the tides and surges change the mean
between the Sunda shelf and the central SCS. On Sunda water depth and current fields (Wolf and Prandle 1999);
shelf, the gradual depth reduction allows the swell to travel therefore, lower Hs in the coastal areas could also be associ-
to coastal regions of Peninsular Malaysia with no sudden ated with tidal currents, which were not included in the
bathymetric interaction, except in areas where islands are simulation. Figure 8 depicts the seasonal mean Hs. Of the
located. However, along the offshore region of Borneo, seasons, the winter monsoon (DJF) has the highest mean Hs,
sudden reductions in depth reduce the wave energy impact ranging from 2 m in the southern part of the SCS to 3 m in the
on coastal areas due to breaking and refraction. Moreover, central region (Fig. 8). In comparison, the magnitudes are
this region might still be influenced by extreme waves gen- less than 1.5 m during summer monsoon (JJA). This obvious
erated by typhoons tracking over the central and northern contrast is because of higher intensity and variability of
regions of the SCS. However, whether they are accurately winds during the winter monsoon. The MAM (SON) transi-
resolved in the CFSR winds is beyond the scope of this tional season has a similar spatial pattern to that of the
paper. summer (winter).
The climatological spatial pattern of the 90th percentiles The climatological mean peak wave period Tp, has high
(Fig. 7) resembles mean Hs with the maximum values values in the central SCS (>12° N) and decreases slightly
reaching 4 m in the central SCS. Along the coast of Vietnam, towards the Sunda shelf and coastal region of Borneo
H90 can reach as high as 3 m. However, these values reduce (Fig. 9). Similar to the mean Hs, the Tp values decrease while
to between 1–2 m in the Gulf of Thailand and southern the waves travel towards the Gulf of Thailand due to the
regions of the SCS (Fig. 7). The obstruction effects of several shadowing of Indo-China. Moreover, the reduction of Tp in
islands in the southern region of the SCS were well resolved. southern regions of the study domain is due to the presence
The existence of multiple islands in southern regions of the of several islands, which prevent waves with longer

Fig. 10 Mean MWD for the DJF MAM

period 1979–2009 model
hindcast 16˚ 16˚

12˚ 12˚

8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JJA SON

16˚ 16˚

12˚ 12˚

8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚

deg
0 60 120 180 240 300 360
Ocean Dynamics (2013) 63:961–977 971

JAN FEB MAR

0.4

0.2
0.4
16˚ 16˚ 16˚

0.4
12˚ 12˚ 12˚

0.4

0.4

0.2
0.2

0.4

0.2
0.2
8˚ 8˚ 8˚

0.2
0.2

0.4
0.2
0.2
4˚ 4˚ 4˚

0.2
0.2
0.2

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
APR MAY JUN

0.2
16˚ 16˚ 16˚
0.2

0.4

12˚ 12˚ 12˚

0.4

0.2
0.2

0.2

0.2
8˚ 8˚ 8˚

4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JUL AUG SEP
0.4

16˚ 0.4
16˚ 16˚ 0.2

0.2
12˚ 12˚ 0.
4
12˚
0.2

0.2
0.4
0.2

0.2
8˚ 8˚ 8˚
0.2

4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
OCT NOV DEC

0.4
0.6
16˚ 16˚ 16˚
0.4

0.6

12˚ 0.4 12˚ 12˚

0.4

0.2
0.4

0.2
0.2

8˚ 8˚ 8˚
0.
2

0.
0.4

0.2 0.4
4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
m
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Fig. 11 Inter-annual wave height variability for the period 1979–2009 model hindcast

wavelengths from propagating into this area. This explains The MWD is predominantly influenced by the monsoonal
why the largest values of T90 (not shown) are simulated at the winds over this region. Starting around November, the Sibe-
offshore Riau Islands. Along the eastern coast of Peninsular rian High pushes the airflow around an anticyclone that
Malaysia, Tp has an almost constant value of 6.5 s. However, carries air from Siberia across China and out over the SCS
T90 reaches around 9 s along Peninsular Malaysia coasts. during the northeast monsoon (DJF). The mean MWD is
972 Ocean Dynamics (2013) 63:961–977

Fig. 12 Correlation of seasonal DJF MAM

mean Hs with Niño 3.4 for winter
(DJF); Spring (MAM); Summer 16˚ 16˚
(JJA) and Autumn (SON) at 90 %
confidence level. Dotted shading
shows non-significant 12˚ 12˚
correlations at 90 % confidence
level
8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JJA SON

16˚ 16˚

12˚ 12˚

8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
Correlation

−0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5

predominantly northeasterly during DJF and MAM (2005) indicated that the anomalous low-level winds over the
(Fig. 10). However, the mean MWD in the Gulf of Thailand SCS are largely modulated by ENSO, particularly during
becomes more southerly in the shadow of the Indo-China winter. Moreover, since the mid-1980s, there has been a
region. During the SON transitional season, the mean MWD tendency for maximum warming to occur at the central
gradually becomes westerly in the central SCS and turns equatorial Pacific Ocean resulting in a different event known
southward along Peninsular Malaysia. At the beginning of as El Nino Modoki (Ashok et al. 2007; Weng et al. 2007,
the summer monsoon (JJA), the wind over the SCS becomes 2009; Ashok and Yamagata 2009), which generates a differ-
predominantly south-southwesterly (Lau et al. 1998). There- ent response over the SCS (Feng et al. 2010). The El Nino
fore, as we expected, the MWD becomes southwesterly Modoki occurrences are represented by the El Niño Modoki
starting late May and reverts back to northeasterly around (EMI) index (Ashok et al. 2007); hence, it is of interest to
mid-September. investigate further the relationship between wave climates in
The standard deviation of the monthly mean significant the SCS and these phenomena.
wave heights for each calendar month was used to measure The mean values of significant wave height for each
the inter-annual variability in that month. The inter-annual season were correlated with the Niño 3.4 and EMI climate
variability of significant wave height in the southern sector indices. The correlation between the Niño 3.4 and mean Hs
of the SCS is greatest during the winter months, especially shows significant negative values in the central sector (>12°
December, with an elongated high standard deviation area N) of the SCS during the winter (DJF) as shown in Fig. 12.
extending from the north, in the northeasterly–southeasterly This is consistent with the work of Juneng and Tangang
direction, to south along the coastal region of Peninsular (2005), which indicated weaker winter monsoon winds over
Malaysia (Fig. 11). The high inter-annual variability ob- the SCS during El Nino. Weaker winds correspond to weaker
served in December could be forced by coherent large- waves and lower significant wave heights. However, in
scale phenomena, such as the ENSO. Juneng and Tangang summertime (JJA), positive correlation values are observed
Ocean Dynamics (2013) 63:961–977 973

Fig. 13 Correlation of seasonal DJF MAM

mean Hs with El Niño Modoki
(EMI) for winter (DJF); Spring 16˚ 16˚
(MAM); Summer (JJA) and
Autumn (SON) at 90 %
confidence level. Dotted shading 12˚ 12˚
shows non-significant
correlations at 90 % confidence
level 1 8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JJA SON

16˚ 16˚

12˚ 12˚

8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
Correlation

−0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3 0.4 0.5

for the entire southern SCS and become significant in the weakly but positively with EMI. This could be associated
offshore region of Peninsular Malaysia and the Karimata with the existence of the anomalous cyclonic circulation over
Straits (Fig. 12). Juneng and Tangang (2005) indicated the eastern Peninsular Malaysia during El Nino Modoki
strengthening of summer monsoonal winds over the south- (Feng et al. 2010). This cyclonic circulation strengthens the
ern region of the SCS during El Nino years. Furthermore, Hs winter monsoonal winds over this region during El Nino
in the Gulf of Thailand exhibits positive correlation with Modoki and hence, enhances the wave height. During tran-
Niño 3.4, except during SON (when there is a negative sitional seasons (i.e. SON and MAM), the relationship be-
correlation). Similarly, in the Celebes Sea region, the corre- tween Hs and EMI is non-significant for almost the entire
lation values are positive in all seasons except SON. The study domain. However, in summer, the area that has posi-
strong ENSO signals in the wave climate in this sea could be tive and significant correlation expands northward to cover
due to its open connection with the western Pacific Ocean, the entire domain, except for the northern region of the
which allows the swell from the Pacific Ocean to easily Vietnamese coast. The differences of correlation strength
penetrate the area. In contrast with this, the relationship is and spatial distribution between Nino3.4 and EMI are
weaker in the Sulu Sea due to the blocking effect of the attributed to the changes in the low-level wind field
Philippine group of islands. over the SCS during El Nino Modoki, which are asso-
Figure 13 indicates the correlation values between Hs and ciated with the westward shifting of the anomalous anti-
the EMI. During SON, DJF and MAM, the correlation cyclonic circulation over the western north Pacific re-
values are weaker compared with the conventional El Nino. gion (Feng et al. 2010).
Moreover, in the winter season, the significant negative Long-term trends in the surface wind influences the Hs
correlation can only be found over the northern region of trend (Young 1999; Young et al. 2011; Bosserelle et al.
the study domain. In the southern region, Hs correlates only 2009). Over the SCS, Juneng and Tangang (2010) indicated
974 Ocean Dynamics (2013) 63:961–977

JAN FEB MAR

16˚ 16˚

12˚ 12˚

8˚ 8˚

4˚ 4˚

0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
APR MAY JUN
16˚ 16˚ 16˚

12˚ 12˚ 12˚

8˚ 8˚ 8˚

4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
JUL AUG SEP
16˚ 16˚ 16˚

12˚ 12˚ 12˚

8˚ 8˚ 8˚

4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚
OCT NOV DEC
16˚ 16˚ 16˚

12˚ 12˚ 12˚

8˚ 8˚ 8˚

4˚ 4˚ 4˚

0˚ 0˚ 0˚

100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚ 100˚ 105˚ 110˚ 115˚ 120˚

m/year
−0.020 −0.015 −0.010 −0.005 0.000 0.005 0.010 0.015 0.020
Fig. 14 Trend in monthly mean wave height for the period 1979–2009 for each month of year. Significant trends at 90 % confidence level are
bounded by a solid red line. Dotted shading shows non-significant trends at 90 % confidence level

that the zonal wind has increased significantly over a period depicts a spatial distribution of significant trends of Hs at the
from 1962–2007. However, if a period from 1979 to 2007 95 % statistical confidence level with areas of non-
were to be considered, the trend might not be as prominent significant changes shaded with dots. Generally, increasing
(Juneng and Tangang 2010, Fig. 2f). In the present study, we positive trends of Hs are found to be significant in the central
quantified the long-term trends in the time series of Hs from region of the SCS during May, July and September (Fig. 14).
1979 to 2009 by a monthly mean trend analysis. Figure 14 In contrast, significant negative trends of Hs dominate the
Ocean Dynamics (2013) 63:961–977 975

7 4
a (Winter) a (Summer)
Significant wave height (m)

Significant wave height (m)

6
3
5

4 2

3
1
2

1 0
1980 1984 1988 1992 1996 2000 2004 2008 1980 1984 1988 1992 1996 2000 2004 2008
7 4
b (Winter) b (Summer)
Significant wave height (m)

Significant wave height (m)

6
3
5

4 2

3
1
2

1 0
1980 1984 1988 1992 1996 2000 2004 2008 1980 1984 1988 1992 1996 2000 2004 2008
7 4
c (Winter) c (Summer)
Significant wave height (m)

Significant wave height (m)

6
3
5

4 2

3
1
2

1 0
1980 1984 1988 1992 1996 2000 2004 2008 1980 1984 1988 1992 1996 2000 2004 2008
Fig. 15 Comparison of averaged 90th percentile of significant wave height in the three bounded areas (a, b and c) shown in Fig. 2 during December
and July between ERA-INTERIM (blue line) and model (red line)

southern region of the SCS during December. Moreover, 4 Conclusions

significant negative trends of Hs can also be found during
June and October in southern and central parts of the SCS. In A third generation wave model (WW3) was applied to sim-
other months, the trends are insignificant. ulate wave climate in the southern extent of the SCS during
We investigated further these trends by examining the the 31-year period from 1979 to 2009. The outputs were well
time series of three box-areas, as indicated in Fig. 2. validated against available in situ and altimeter data and
Figure 15 shows the time series of the 90th percentiles of hence, the computed wave characteristics from the model
Hs for the three areas during July and December. For com- outputs provide reasonable estimates of wave climate in the
parison, the corresponding time series of the ERA-INTERIM region in the absence of long-term wave observations. The
are also shown. As expected, the modelled 90th percentiles spatial distribution of the modelled seasonal mean Hs (and
of Hs are higher than those of the ERA-INTERIM due to the H90) indicates higher values in the northern region of the
WW3 finer grid resolution. Moreover, the two time series are study domain, which decrease southwestwards. In the most
also mostly in phase with respect to the inter-annual southern regions, lower amplitudes of Hs are associated with
variability. However, the modelled fluctuations are shadowing, island blocking and refraction. Seasonally, the
greater during winter, especially in conjunction with values are higher during winter (DJF) and lower during
the 1997/98 El Nino. The large inter-annual fluctuations summer (JJA) and spring (MAM). Also, the results demon-
and relatively short record could be the reason why strate strong inter-annual variability, especially during the
most of the trends of the modelled 90th percentiles of winter months. The mean Hs correlates negatively with the
Hs are not significant. Nino3.4 index for almost the entire region during winter,
976 Ocean Dynamics (2013) 63:961–977

spring and autumn seasons. However, the correlation sign re- Hanson JL, Tracy BA, Tolma HL, Scott DR (2009) Pacific hindcast
performance of three numerical wave model. J Atmos Ocean
verses during the summer monsoon. During El Nino Modoki,
Technol 26:1614–1633
the positive correlation during summer extends northeastwards Hemer MA, Church JA, Hunter JR (2010) Variability and trends in the
to cover almost the entire domain. The trend analysis indicates directional wave climate of the southern hemisphere. Int J
positive and negative long-term trends of mean Hs depending Climatol 30:475–491. doi:10.1002/joc.1900
IPCC AR4 (2007) Climate change 2007, Fourth assessment report of
on the month and location. However, the long-term trends
the intergovernmental panel on climate change. Cambridge
could be masked by the large inter-annual variability. University Press, Cambridge
Jiang LF, Zhang ZX, Qi YQ (2010) Simulation of SWAN and
Acknowledgements This research is funded by the grants of MOHE WAVEWATCH III in Northern South China Sea. 20th
LRGS/TD/2011/UKM/PG/01, MOSTI Science fund 04-01-02-SF0747 International Offshore and Polar Engineering Conference (Beijing,
and Universiti Kebangsaan Malaysia DIP-2012-020. China, 20–25 June 2010)
Juneng L, Tangang FT (2005) Evolution of ENSO-related rainfall
anomalies in Southeast Asia region and its relationship with atmo-
sphere–ocean variations in Indo-Pacific sector. Clim Dyn 25:337–
References 350. doi:10.1077/s00382-005-0031-6
Juneng L, Tangang FT (2010) Long-term trends of winter monsoon
synoptic circulations over the maritime continent: 1962–2007.
Ashok K, Yamagata T (2009) The El Niño with a difference. Nature Atmos Sci Lett 11(3):199–203. doi:10.1002/asl.272
461:481–484. doi:10.1038/461481a Kriiezi EE, Broman B (2008) Past and future wave climate in the Baltic
Ashok K, Behera SK, Rao SA, Weng HY, Yamagata T (2007) El Nino Sea produced by the SWAN model with forcing from the regional
Modoki and its possible teleconnection. J Geophys Res climate model RCA of the Rossby Center. In: IEEE/OES US/EU-
112(C11007). doi:10.1029/2006JC003798 Baltic International Symposium (eds); (Tallinn, Estonia, 27–29
Bidlot JR, Li JG, Wittmann P, Faucher M, Chen HS, Lefevre JM, Bruns May 2008), pp 360–366
T, Greenslade D, Ardhuin F, Kohno N, Park S (2007) Gomez M Lau KM, Wu HT, Yang S (1998) Hydrologic processes associated with
2007 Inter-comparison of operational wave forecasting systems, the first transition of the Asian Summer Monsoon: A pilot satellite
10th International workshop on wave hindcasting and forecasting. study. Bull Am Meteorol Soc 79:1871–1882
Oahu, Hawaii Lionello P, Sanna A (2005) Mediterranean wave climate variability and
Bosserelle C, Pattiaratchi CB, Haigh I (2012) Inter-annual variability its links with NAO and Indian Monsoon. Clim Dyn 25:611–623.
and longer-term changes in the wave climate of Western Australia doi:10.1007/s00382-005-0025-4
between 1970 and 2009. Ocean Dyn 62:63–76. doi:10.1007/ Mori N, Yasuda T, Mase H, Tom T, Oku Y (2010) Projection of extreme
s10236-011-0487-3 wave climate change under global warming. Hydrol Res Lett
Brown JM, Wolf J, Souza AJ (2012) Past to future extreme events in 4:15–19
Liverpool Bay: model projections from 1960–2100. Clim Chang Music S, Nikovic S (2008) 44-year wave hindcast for the Eastern
111:65–391. doi:10.1007/s10584-011-0145-2 Mediterranean. Coast Eng 55:872–880. doi:10.1016/j.coastaleng.
Chawla A, Tolman HL (2007) Automated grid generation for 2008.02.024
WAVEWATCH III. NOAA/NWS/NCEP/OMB, Technical Note, 254 NCEP (2006) National Center for Environmental Prediction,
Chawla A, Tolman HL, Lanson JL, Devaliere EM, Gerald VM (2009) Washington DC. available at http://polar.ncep.noaa.gov/waves
Validation of a Multi-Grid WAVEWATCH III™ modeling system. NGDC (2006) National Geophysical Data Center. Available at http://
NOAA/MMAB Contribution no. 281 ngdc.noaa.gov/mgg/global/etopo2.html
Chu PC, Lu S, Liu WT (1999) Uncertainty of South China Sea predic- Padilla-Hernandez R, Perrie W, Toulany B, Smith PC, Zhang W,
tion using NSCAT and national centers for environmental predic- Jimenez-Hernandez S (2004) Intercomparison of modern opera-
tion winds during tropical storm Ernie, 1996. J Geophys Res tional wave models. 8th International workshop on wave
104(C5):11273–11289. doi:10.1029/1998JC900046 hindcasting and forecasting (North Shore, Oahu, Hawaii)
Chu PC, Veneziano JM, Fan C, Carron MJ, Liu WT (2000) Response of Panchang VG, Jeong C, Li D (2008) Wave climatology in coastal
the South China Sea to tropical Cyclone Ernie 1996. J Geophys Maine for aquaculture and other applications. Estuaries and
Res 105(C6):13991–14009. doi:10.1029/2000JC900035 Coast. J CERF 31:289–299. doi:10.1007/s12237-007-9016-5
Chu PC, Qi Y, Chen Y, Shi P, Mao Q (2004) South China Sea wind- Qi YQ, Zhang ZX, SHI P (2010) Extreme wind, wave and current in deep
wave characteristics. Part I: validation of Wavewatch-III using water of South China Sea. Int J Offshore Polar Eng 20(1):18–23
TOPEX/Poseidon data. J Atmos Ocean Technol 21(11):1718– Saha S, Moorthi S, Pan HL, Wu X, Wang J, Nadiga S, Tripp P, Kistler
1733. doi:10.1175/JTECH1661.1 R, Woollen J, Behringer D, Liu H, Stokes D, Grumbine R, Gayno
Chunxia L, Qi Y, Liang J (2008) The effect of sea waves on Typhoon G, Wang J, Hou YT, Chuang H, Juang HMH, Sela J, Iredell M,
Imodo. 2nd CAWCR Modelling Workshop 2008, High Resolution Treadon R, Kleist D, van Delst P, Keyser D, Derber J, Ek M, Meng
Modelling (Australia, 25–28 Nov 2008) J, Wei H, Yang R, Lord S, van den Dool H, Kumar A, Wang W,
Cuchiara DC, Fernandes EH, Strauch JC, Winterwerp JC, Calliari LJ Long C, Chelliah M, Xue Y, Huang B, Schemm JK, Ebisuzaki W,
(2009) Determination of the wave climate for southern Brazilian Lin R, Xie P, Chen M, Zhou S, Higginsz W, Zou CZ, Liu Q, Chen
shelf. Cont Shelf Res 29:545–555. doi:10.1016/j.csr.2008.09.025 Y, Han Y, Cucurull L, Reynolds RW, Rutledge G, Goldberg M
Fang G, Kwok YK, Yu K, Zhu Y (1999) Numerical simulation of (2010) The NCEP climate forecast system reanalysis. Bull Am
principal tidal constituents in the South China Sea, Gulf of Meteorol Soc 91:1015–1057
Tonkin and Gulf of Thailand. Cont Shelf Res 19(7):845–869. Salahuddin A, Curtis S (2011) Climate extremes in Malaysia and the
doi:10.1016/s0278-4343(99)00002-3 equatorial South China Sea. Glob Planet Chang 78:83–91. doi:10.
Feng M, McPhaden MJ, Lee T (2010) Decadal variability of the Pacific 1016/j.gloplacha.2011.05.001
subtropical cells and their influence on the southeast Indian Ocean. Swail VR, Cox AT, Cardone VJ (1999) Analysis of wave climate trends
Geophys Res Lett 37, L09606. doi:10.1029/2010GL042796 and variability. WMO Workshop on Advances in Marine
Ocean Dynamics (2013) 63:961–977 977

Climatology (CLIMAR99), 8–15 September 1999, Vancouver, Weng H, Behera SK, Yamagata T (2009) Anomalous winter climate
B.C. p: 245–256 conditions in the Pacific rim during recent El Niño Modoki and El
Tolman HL (2002) Validation of WAVEWATCH III version 1.15 for a Niño events. Clim Dyn 29:113–129. doi:10.1007/s00382-008-
global domain. NCEP/NOAA/NWS, National Centers for 0394-6
Environmental Prediction, Technical note 213, Washington Wolf J, Prandle D (1999) Some observation of wave–current interac-
Tolman HL (2009) User manual and system documentation of tion. Coast Eng 37:471–485
WAVEWATCH III TM version 3.14. NOAA/NWS/NCEP/ Wright LD, Boon JD, Kim SC, List JH (1991) Modes of cross-
MMAB Technical Note 276 shore sediment transport on the shoreface of the Middle
WAMDI (1988) The WAM model—a third generation ocean wave Atlantic Bight. Mar Geol 96:19–51. doi:10.1016/0025-
prediction model. J Phys Oceanogr 18:1775–1810 3227(91)90200-N
Wang XL, Swail VR (2006) Trends of Atlantic wave extremes as Xue Y, Huang B, Hu ZZ, Kumar A, Wen C, Behringer D, Nadiga S
simulated in a 40-yr wave hindcast using kinematically reanalyzed (2010) An assessment of oceanic variability in the NCEP climate
wind fields. J Clim 15:1020–1035. doi:10.1175/1520-0442(2002) forecast system reanalysis. Clim Dyn. doi:10.1007/s00382-010-
015<1020:TOAWEA>2.0.CO;2 0954-4
Wang XL, Zwiers FW, Swail VR (2004) North Atlantic Ocean wave Young IR (1999) Seasonal variability of the global ocean wind and
climate change scenarios for the twenty-first century. J Clim. wave climate. Int J Climatol 19:931–950
doi:10.1175/1520-0442 Young IR, Zieger S, Babanin AV (2011) Global trends in wind speed
Weng H, Ashok K, Behera SK, Rao SA, Yamagata T (2007) Impacts of and wave height. Science. doi:10.1126/science.1197219
recent El Niño Modoki on dry/wet conditions in the Pacific Rim Zacharioudaki A, Pan S, Simmonds D (2011) Future wave climate over
during boreal summer. Clim Dyn 29:113–129. doi:10.1007/ the west-European shelf seas. Ocean Dyn 61:807–827. doi:10.
s00382-007-0234-0 1007/s10236-011-0395-6